{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "global_step": 5545,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.1976047904191619e-06,
      "loss": 0.2885,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.3952095808383237e-06,
      "loss": 1.4087,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.592814371257485e-06,
      "loss": 1.3698,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.7904191616766475e-06,
      "loss": 1.2934,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.9880239520958085e-06,
      "loss": 1.3185,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.18562874251497e-06,
      "loss": 0.2838,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.383233532934131e-06,
      "loss": 1.3835,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.580838323353295e-06,
      "loss": 1.2936,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.0778443113772455e-05,
      "loss": 1.2066,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.1976047904191617e-05,
      "loss": 1.2226,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.317365269461078e-05,
      "loss": 1.2099,
      "step": 11
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.437125748502994e-05,
      "loss": 1.2113,
      "step": 12
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.5568862275449103e-05,
      "loss": 1.1605,
      "step": 13
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.6766467065868263e-05,
      "loss": 1.1536,
      "step": 14
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.7964071856287426e-05,
      "loss": 1.05,
      "step": 15
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.916167664670659e-05,
      "loss": 1.1224,
      "step": 16
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.035928143712575e-05,
      "loss": 1.1326,
      "step": 17
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.155688622754491e-05,
      "loss": 1.1318,
      "step": 18
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.275449101796407e-05,
      "loss": 1.1411,
      "step": 19
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.3952095808383234e-05,
      "loss": 1.0917,
      "step": 20
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.5149700598802394e-05,
      "loss": 1.0307,
      "step": 21
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.634730538922156e-05,
      "loss": 1.087,
      "step": 22
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.754491017964072e-05,
      "loss": 1.1036,
      "step": 23
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.874251497005988e-05,
      "loss": 1.0446,
      "step": 24
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.994011976047904e-05,
      "loss": 0.3425,
      "step": 25
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.1137724550898205e-05,
      "loss": 1.0632,
      "step": 26
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.233532934131737e-05,
      "loss": 1.0997,
      "step": 27
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.3532934131736525e-05,
      "loss": 1.0095,
      "step": 28
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.473053892215569e-05,
      "loss": 1.0676,
      "step": 29
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.592814371257485e-05,
      "loss": 1.0075,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.712574850299401e-05,
      "loss": 1.0498,
      "step": 31
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.832335329341318e-05,
      "loss": 1.033,
      "step": 32
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.9520958083832336e-05,
      "loss": 1.0538,
      "step": 33
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.07185628742515e-05,
      "loss": 1.0261,
      "step": 34
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.191616766467066e-05,
      "loss": 0.9697,
      "step": 35
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.311377245508982e-05,
      "loss": 1.0,
      "step": 36
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.4311377245508984e-05,
      "loss": 1.054,
      "step": 37
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.550898203592814e-05,
      "loss": 1.1101,
      "step": 38
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.670658682634731e-05,
      "loss": 1.0584,
      "step": 39
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.790419161676647e-05,
      "loss": 1.004,
      "step": 40
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.910179640718563e-05,
      "loss": 0.9803,
      "step": 41
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.029940119760479e-05,
      "loss": 1.0521,
      "step": 42
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.149700598802395e-05,
      "loss": 1.016,
      "step": 43
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.269461077844312e-05,
      "loss": 0.9757,
      "step": 44
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.389221556886228e-05,
      "loss": 0.9601,
      "step": 45
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.508982035928144e-05,
      "loss": 1.0342,
      "step": 46
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.62874251497006e-05,
      "loss": 1.038,
      "step": 47
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.748502994011976e-05,
      "loss": 0.9976,
      "step": 48
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.868263473053892e-05,
      "loss": 0.9443,
      "step": 49
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.988023952095808e-05,
      "loss": 1.0518,
      "step": 50
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.107784431137725e-05,
      "loss": 1.0989,
      "step": 51
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.227544910179641e-05,
      "loss": 1.0203,
      "step": 52
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.347305389221557e-05,
      "loss": 0.9962,
      "step": 53
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.467065868263474e-05,
      "loss": 0.985,
      "step": 54
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.58682634730539e-05,
      "loss": 0.9819,
      "step": 55
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.706586826347305e-05,
      "loss": 0.2824,
      "step": 56
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.826347305389222e-05,
      "loss": 1.006,
      "step": 57
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.946107784431138e-05,
      "loss": 0.9949,
      "step": 58
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.065868263473055e-05,
      "loss": 0.8969,
      "step": 59
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.18562874251497e-05,
      "loss": 0.2999,
      "step": 60
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.305389221556886e-05,
      "loss": 1.0104,
      "step": 61
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.425149700598802e-05,
      "loss": 0.9798,
      "step": 62
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.544910179640719e-05,
      "loss": 0.938,
      "step": 63
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.664670658682636e-05,
      "loss": 0.9396,
      "step": 64
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.784431137724552e-05,
      "loss": 0.9832,
      "step": 65
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.904191616766467e-05,
      "loss": 0.9124,
      "step": 66
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.023952095808383e-05,
      "loss": 0.9966,
      "step": 67
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.1437125748503e-05,
      "loss": 0.927,
      "step": 68
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.263473053892216e-05,
      "loss": 0.9613,
      "step": 69
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.383233532934131e-05,
      "loss": 0.973,
      "step": 70
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.502994011976048e-05,
      "loss": 1.0117,
      "step": 71
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.622754491017964e-05,
      "loss": 1.0077,
      "step": 72
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.742514970059881e-05,
      "loss": 1.014,
      "step": 73
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.862275449101797e-05,
      "loss": 0.9429,
      "step": 74
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.982035928143712e-05,
      "loss": 1.0264,
      "step": 75
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.101796407185628e-05,
      "loss": 0.9454,
      "step": 76
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.221556886227547e-05,
      "loss": 1.0533,
      "step": 77
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.341317365269462e-05,
      "loss": 0.3204,
      "step": 78
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.461077844311378e-05,
      "loss": 0.9421,
      "step": 79
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.580838323353294e-05,
      "loss": 0.9589,
      "step": 80
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.700598802395209e-05,
      "loss": 0.9477,
      "step": 81
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.820359281437126e-05,
      "loss": 0.9845,
      "step": 82
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.940119760479042e-05,
      "loss": 0.956,
      "step": 83
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00010059880239520958,
      "loss": 1.0222,
      "step": 84
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00010179640718562875,
      "loss": 0.9604,
      "step": 85
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001029940119760479,
      "loss": 0.9721,
      "step": 86
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00010419161676646707,
      "loss": 0.9769,
      "step": 87
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00010538922155688624,
      "loss": 0.9876,
      "step": 88
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00010658682634730539,
      "loss": 0.95,
      "step": 89
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00010778443113772456,
      "loss": 0.9519,
      "step": 90
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00010898203592814371,
      "loss": 0.2964,
      "step": 91
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00011017964071856288,
      "loss": 0.9857,
      "step": 92
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00011137724550898203,
      "loss": 0.9506,
      "step": 93
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001125748502994012,
      "loss": 0.9219,
      "step": 94
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00011377245508982037,
      "loss": 0.9814,
      "step": 95
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00011497005988023953,
      "loss": 0.9038,
      "step": 96
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001161676646706587,
      "loss": 0.9216,
      "step": 97
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00011736526946107784,
      "loss": 0.9747,
      "step": 98
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00011856287425149701,
      "loss": 0.9391,
      "step": 99
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00011976047904191617,
      "loss": 0.305,
      "step": 100
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00012095808383233534,
      "loss": 0.9326,
      "step": 101
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001221556886227545,
      "loss": 0.9672,
      "step": 102
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00012335329341317365,
      "loss": 0.9274,
      "step": 103
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00012455089820359282,
      "loss": 0.9665,
      "step": 104
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00012574850299401196,
      "loss": 0.9436,
      "step": 105
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00012694610778443113,
      "loss": 0.9464,
      "step": 106
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001281437125748503,
      "loss": 0.9662,
      "step": 107
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00012934131736526947,
      "loss": 0.9607,
      "step": 108
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00013053892215568865,
      "loss": 0.9661,
      "step": 109
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001317365269461078,
      "loss": 0.9253,
      "step": 110
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00013293413173652696,
      "loss": 0.8996,
      "step": 111
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001341317365269461,
      "loss": 0.2684,
      "step": 112
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00013532934131736527,
      "loss": 1.0148,
      "step": 113
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00013652694610778444,
      "loss": 0.8979,
      "step": 114
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00013772455089820359,
      "loss": 0.9471,
      "step": 115
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00013892215568862276,
      "loss": 0.9816,
      "step": 116
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00014011976047904193,
      "loss": 0.9937,
      "step": 117
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001413173652694611,
      "loss": 0.9662,
      "step": 118
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00014251497005988024,
      "loss": 0.9524,
      "step": 119
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001437125748502994,
      "loss": 0.9613,
      "step": 120
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00014491017964071858,
      "loss": 0.9058,
      "step": 121
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00014610778443113772,
      "loss": 0.9553,
      "step": 122
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001473053892215569,
      "loss": 0.9655,
      "step": 123
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00014850299401197604,
      "loss": 0.9489,
      "step": 124
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001497005988023952,
      "loss": 0.9821,
      "step": 125
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00015089820359281438,
      "loss": 0.9748,
      "step": 126
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00015209580838323355,
      "loss": 0.9658,
      "step": 127
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00015329341317365272,
      "loss": 0.9887,
      "step": 128
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00015449101796407186,
      "loss": 0.9292,
      "step": 129
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00015568862275449103,
      "loss": 0.9391,
      "step": 130
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00015688622754491018,
      "loss": 1.0044,
      "step": 131
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00015808383233532935,
      "loss": 0.9607,
      "step": 132
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001592814371257485,
      "loss": 0.9453,
      "step": 133
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00016047904191616766,
      "loss": 0.9992,
      "step": 134
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00016167664670658683,
      "loss": 0.9574,
      "step": 135
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000162874251497006,
      "loss": 0.3056,
      "step": 136
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00016407185628742517,
      "loss": 0.9099,
      "step": 137
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00016526946107784431,
      "loss": 0.9537,
      "step": 138
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00016646706586826348,
      "loss": 0.3119,
      "step": 139
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00016766467065868263,
      "loss": 0.9403,
      "step": 140
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001688622754491018,
      "loss": 0.947,
      "step": 141
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00017005988023952097,
      "loss": 0.9288,
      "step": 142
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001712574850299401,
      "loss": 0.9635,
      "step": 143
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00017245508982035928,
      "loss": 1.0032,
      "step": 144
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00017365269461077845,
      "loss": 0.8915,
      "step": 145
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00017485029940119762,
      "loss": 0.9361,
      "step": 146
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001760479041916168,
      "loss": 0.9041,
      "step": 147
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00017724550898203594,
      "loss": 0.9476,
      "step": 148
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001784431137724551,
      "loss": 0.891,
      "step": 149
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00017964071856287425,
      "loss": 0.9032,
      "step": 150
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00018083832335329342,
      "loss": 0.9164,
      "step": 151
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00018203592814371256,
      "loss": 0.897,
      "step": 152
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00018323353293413173,
      "loss": 0.9411,
      "step": 153
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00018443113772455093,
      "loss": 0.8962,
      "step": 154
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00018562874251497007,
      "loss": 0.9226,
      "step": 155
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00018682634730538924,
      "loss": 0.9694,
      "step": 156
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001880239520958084,
      "loss": 1.0053,
      "step": 157
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00018922155688622756,
      "loss": 0.9859,
      "step": 158
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001904191616766467,
      "loss": 0.9533,
      "step": 159
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019161676646706587,
      "loss": 0.9904,
      "step": 160
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019281437125748504,
      "loss": 0.946,
      "step": 161
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019401197604790419,
      "loss": 0.9223,
      "step": 162
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019520958083832338,
      "loss": 0.9719,
      "step": 163
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019640718562874253,
      "loss": 0.2981,
      "step": 164
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001976047904191617,
      "loss": 0.946,
      "step": 165
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019880239520958084,
      "loss": 0.9167,
      "step": 166
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0002,
      "loss": 0.9058,
      "step": 167
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999998293807058,
      "loss": 0.8965,
      "step": 168
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999993175228808,
      "loss": 0.9318,
      "step": 169
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999984644267,
      "loss": 0.9286,
      "step": 170
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999972700924541,
      "loss": 0.9871,
      "step": 171
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999957345205512,
      "loss": 0.9091,
      "step": 172
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999938577115148,
      "loss": 0.9379,
      "step": 173
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999916396659857,
      "loss": 0.9666,
      "step": 174
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999890803847208,
      "loss": 0.9629,
      "step": 175
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999861798685934,
      "loss": 0.9543,
      "step": 176
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999982938118593,
      "loss": 0.9679,
      "step": 177
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999793551358255,
      "loss": 0.9326,
      "step": 178
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999754309215145,
      "loss": 0.9094,
      "step": 179
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999711654769984,
      "loss": 0.9466,
      "step": 180
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999665588037334,
      "loss": 0.9161,
      "step": 181
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999616109032907,
      "loss": 0.9899,
      "step": 182
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999563217773587,
      "loss": 0.9988,
      "step": 183
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999506914277428,
      "loss": 0.9288,
      "step": 184
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999447198563644,
      "loss": 0.9295,
      "step": 185
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999384070652605,
      "loss": 0.9632,
      "step": 186
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999931753056586,
      "loss": 0.9412,
      "step": 187
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999924757832611,
      "loss": 0.9444,
      "step": 188
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999917421395723,
      "loss": 0.9647,
      "step": 189
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999909743748425,
      "loss": 0.9128,
      "step": 190
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999901724893337,
      "loss": 0.9517,
      "step": 191
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019998933648331953,
      "loss": 0.8794,
      "step": 192
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019998846635708534,
      "loss": 0.9435,
      "step": 193
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019998756211092795,
      "loss": 0.9646,
      "step": 194
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019998662374515598,
      "loss": 0.931,
      "step": 195
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999856512600896,
      "loss": 0.9334,
      "step": 196
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999846446560607,
      "loss": 0.9275,
      "step": 197
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019998360393341276,
      "loss": 0.9084,
      "step": 198
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999825290925009,
      "loss": 0.3228,
      "step": 199
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019998142013369192,
      "loss": 0.9157,
      "step": 200
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999802770573642,
      "loss": 0.867,
      "step": 201
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019997909986390786,
      "loss": 0.9597,
      "step": 202
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019997788855372458,
      "loss": 0.9642,
      "step": 203
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019997664312722767,
      "loss": 0.9686,
      "step": 204
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019997536358484213,
      "loss": 0.9375,
      "step": 205
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019997404992700465,
      "loss": 0.3197,
      "step": 206
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019997270215416343,
      "loss": 0.8823,
      "step": 207
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019997132026677842,
      "loss": 0.9677,
      "step": 208
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019996990426532113,
      "loss": 0.95,
      "step": 209
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019996845415027483,
      "loss": 0.9523,
      "step": 210
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999669699221343,
      "loss": 0.9265,
      "step": 211
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000199965451581406,
      "loss": 0.8872,
      "step": 212
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999638991286081,
      "loss": 0.9321,
      "step": 213
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999623125642703,
      "loss": 0.8988,
      "step": 214
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019996069188893403,
      "loss": 0.9363,
      "step": 215
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019995903710315233,
      "loss": 0.9635,
      "step": 216
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999573482074899,
      "loss": 0.9328,
      "step": 217
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000199955625202523,
      "loss": 0.9104,
      "step": 218
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019995386808883964,
      "loss": 0.3386,
      "step": 219
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019995207686703933,
      "loss": 0.9456,
      "step": 220
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019995025153773339,
      "loss": 0.2754,
      "step": 221
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019994839210154468,
      "loss": 0.9716,
      "step": 222
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999464985591077,
      "loss": 0.9821,
      "step": 223
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019994457091106855,
      "loss": 0.8759,
      "step": 224
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999426091580851,
      "loss": 0.8963,
      "step": 225
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019994061330082675,
      "loss": 0.337,
      "step": 226
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019993858333997453,
      "loss": 0.9217,
      "step": 227
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019993651927622118,
      "loss": 1.0219,
      "step": 228
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019993442111027098,
      "loss": 0.9871,
      "step": 229
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019993228884284,
      "loss": 0.9053,
      "step": 230
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019993012247465576,
      "loss": 0.8881,
      "step": 231
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019992792200645756,
      "loss": 0.9035,
      "step": 232
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999256874389963,
      "loss": 0.9203,
      "step": 233
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019992341877303441,
      "loss": 0.9007,
      "step": 234
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019992111600934616,
      "loss": 0.9248,
      "step": 235
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019991877914871727,
      "loss": 0.9081,
      "step": 236
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999164081919452,
      "loss": 0.9611,
      "step": 237
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019991400313983899,
      "loss": 0.9185,
      "step": 238
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019991156399321939,
      "loss": 0.945,
      "step": 239
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019990909075291865,
      "loss": 0.9152,
      "step": 240
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019990658341978074,
      "loss": 0.9273,
      "step": 241
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019990404199466132,
      "loss": 0.9205,
      "step": 242
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001999014664784276,
      "loss": 0.9045,
      "step": 243
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019989885687195843,
      "loss": 0.898,
      "step": 244
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019989621317614432,
      "loss": 0.9219,
      "step": 245
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019989353539188743,
      "loss": 0.9823,
      "step": 246
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019989082352010144,
      "loss": 0.8548,
      "step": 247
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019988807756171184,
      "loss": 0.9361,
      "step": 248
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001998852975176556,
      "loss": 0.9586,
      "step": 249
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019988248338888135,
      "loss": 0.9192,
      "step": 250
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019987963517634948,
      "loss": 0.9393,
      "step": 251
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019987675288103183,
      "loss": 0.9225,
      "step": 252
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019987383650391197,
      "loss": 0.991,
      "step": 253
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019987088604598509,
      "loss": 0.9576,
      "step": 254
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000199867901508258,
      "loss": 0.9706,
      "step": 255
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019986488289174909,
      "loss": 0.9615,
      "step": 256
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001998618301974885,
      "loss": 0.9468,
      "step": 257
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019985874342651788,
      "loss": 0.3303,
      "step": 258
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001998556225798906,
      "loss": 0.8744,
      "step": 259
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001998524676586716,
      "loss": 0.9586,
      "step": 260
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001998492786639374,
      "loss": 0.8891,
      "step": 261
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019984605559677628,
      "loss": 0.9517,
      "step": 262
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019984279845828806,
      "loss": 0.9376,
      "step": 263
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019983950724958418,
      "loss": 0.9683,
      "step": 264
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019983618197178774,
      "loss": 0.9576,
      "step": 265
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019983282262603345,
      "loss": 0.9399,
      "step": 266
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019982942921346764,
      "loss": 0.9115,
      "step": 267
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001998260017352483,
      "loss": 0.9309,
      "step": 268
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019982254019254502,
      "loss": 0.9209,
      "step": 269
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019981904458653892,
      "loss": 0.9512,
      "step": 270
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019981551491842297,
      "loss": 0.952,
      "step": 271
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019981195118940156,
      "loss": 0.91,
      "step": 272
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001998083534006908,
      "loss": 0.9019,
      "step": 273
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019980472155351834,
      "loss": 0.9588,
      "step": 274
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001998010556491235,
      "loss": 0.9173,
      "step": 275
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019979735568875735,
      "loss": 0.8774,
      "step": 276
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019979362167368234,
      "loss": 0.92,
      "step": 277
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019978985360517273,
      "loss": 0.9629,
      "step": 278
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019978605148451427,
      "loss": 0.9299,
      "step": 279
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001997822153130044,
      "loss": 0.9228,
      "step": 280
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019977834509195223,
      "loss": 0.9465,
      "step": 281
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019977444082267836,
      "loss": 0.9499,
      "step": 282
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019977050250651512,
      "loss": 0.9117,
      "step": 283
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001997665301448064,
      "loss": 0.3191,
      "step": 284
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019976252373890771,
      "loss": 0.8806,
      "step": 285
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001997584832901862,
      "loss": 0.9467,
      "step": 286
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019975440880002062,
      "loss": 0.9162,
      "step": 287
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001997503002698014,
      "loss": 0.881,
      "step": 288
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019974615770093046,
      "loss": 0.9339,
      "step": 289
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019974198109482143,
      "loss": 0.9062,
      "step": 290
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001997377704528995,
      "loss": 0.928,
      "step": 291
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019973352577660157,
      "loss": 0.9876,
      "step": 292
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019972924706737602,
      "loss": 0.9256,
      "step": 293
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019972493432668294,
      "loss": 0.3693,
      "step": 294
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019972058755599398,
      "loss": 0.9256,
      "step": 295
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001997162067567925,
      "loss": 0.8775,
      "step": 296
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019971179193057332,
      "loss": 0.3365,
      "step": 297
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019970734307884298,
      "loss": 0.9094,
      "step": 298
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019970286020311957,
      "loss": 0.9589,
      "step": 299
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019969834330493287,
      "loss": 0.9376,
      "step": 300
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019969379238582417,
      "loss": 0.8758,
      "step": 301
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019968920744734648,
      "loss": 0.9546,
      "step": 302
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019968458849106425,
      "loss": 0.9081,
      "step": 303
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001996799355185538,
      "loss": 0.9534,
      "step": 304
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019967524853140278,
      "loss": 0.97,
      "step": 305
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001996705275312106,
      "loss": 0.8747,
      "step": 306
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019966577251958828,
      "loss": 1.0005,
      "step": 307
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001996609834981584,
      "loss": 0.9764,
      "step": 308
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019965616046855515,
      "loss": 0.8894,
      "step": 309
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019965130343242433,
      "loss": 0.9053,
      "step": 310
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019964641239142336,
      "loss": 0.9177,
      "step": 311
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019964148734722122,
      "loss": 0.9679,
      "step": 312
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019963652830149857,
      "loss": 0.9135,
      "step": 313
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001996315352559476,
      "loss": 0.9073,
      "step": 314
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019962650821227215,
      "loss": 0.966,
      "step": 315
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019962144717218766,
      "loss": 0.9064,
      "step": 316
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019961635213742106,
      "loss": 0.9903,
      "step": 317
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019961122310971106,
      "loss": 0.9139,
      "step": 318
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019960606009080786,
      "loss": 0.8449,
      "step": 319
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019960086308247327,
      "loss": 0.9108,
      "step": 320
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019959563208648073,
      "loss": 0.9302,
      "step": 321
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019959036710461524,
      "loss": 0.9523,
      "step": 322
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019958506813867342,
      "loss": 0.9232,
      "step": 323
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001995797351904635,
      "loss": 0.3026,
      "step": 324
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019957436826180524,
      "loss": 0.9283,
      "step": 325
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019956896735453007,
      "loss": 0.9685,
      "step": 326
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000199563532470481,
      "loss": 0.9512,
      "step": 327
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019955806361151264,
      "loss": 0.9398,
      "step": 328
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019955256077949112,
      "loss": 0.9148,
      "step": 329
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019954702397629426,
      "loss": 0.8659,
      "step": 330
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019954145320381139,
      "loss": 0.9739,
      "step": 331
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019953584846394353,
      "loss": 0.9041,
      "step": 332
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019953020975860318,
      "loss": 0.9296,
      "step": 333
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019952453708971455,
      "loss": 0.8918,
      "step": 334
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019951883045921332,
      "loss": 0.8351,
      "step": 335
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001995130898690468,
      "loss": 0.8815,
      "step": 336
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019950731532117393,
      "loss": 0.9493,
      "step": 337
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001995015068175652,
      "loss": 0.9376,
      "step": 338
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001994956643602027,
      "loss": 0.9572,
      "step": 339
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019948978795108013,
      "loss": 0.9779,
      "step": 340
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019948387759220269,
      "loss": 0.913,
      "step": 341
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019947793328558725,
      "loss": 0.8977,
      "step": 342
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019947195503326225,
      "loss": 0.9631,
      "step": 343
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001994659428372677,
      "loss": 0.8816,
      "step": 344
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019945989669965518,
      "loss": 0.9185,
      "step": 345
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019945381662248786,
      "loss": 0.93,
      "step": 346
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019944770260784048,
      "loss": 0.9128,
      "step": 347
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019944155465779944,
      "loss": 0.3015,
      "step": 348
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001994353727744626,
      "loss": 0.9114,
      "step": 349
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019942915695993948,
      "loss": 0.949,
      "step": 350
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019942290721635116,
      "loss": 0.3057,
      "step": 351
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019941662354583026,
      "loss": 0.977,
      "step": 352
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019941030595052107,
      "loss": 0.3302,
      "step": 353
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019940395443257934,
      "loss": 0.9125,
      "step": 354
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019939756899417247,
      "loss": 0.9255,
      "step": 355
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019939114963747945,
      "loss": 0.956,
      "step": 356
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019938469636469078,
      "loss": 0.8991,
      "step": 357
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019937820917800857,
      "loss": 0.8889,
      "step": 358
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001993716880796465,
      "loss": 0.9,
      "step": 359
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019936513307182985,
      "loss": 0.9256,
      "step": 360
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019935854415679537,
      "loss": 0.9279,
      "step": 361
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001993519213367915,
      "loss": 0.9405,
      "step": 362
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019934526461407823,
      "loss": 0.9008,
      "step": 363
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019933857399092705,
      "loss": 0.8443,
      "step": 364
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019933184946962105,
      "loss": 0.8915,
      "step": 365
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019932509105245494,
      "loss": 0.9853,
      "step": 366
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001993182987417349,
      "loss": 0.9129,
      "step": 367
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019931147253977877,
      "loss": 0.9129,
      "step": 368
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019930461244891588,
      "loss": 0.908,
      "step": 369
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019929771847148717,
      "loss": 0.9002,
      "step": 370
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019929079060984517,
      "loss": 0.8625,
      "step": 371
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019928382886635388,
      "loss": 0.8851,
      "step": 372
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019927683324338894,
      "loss": 0.8459,
      "step": 373
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019926980374333753,
      "loss": 0.9314,
      "step": 374
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019926274036859834,
      "loss": 0.925,
      "step": 375
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019925564312158175,
      "loss": 0.9303,
      "step": 376
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019924851200470959,
      "loss": 0.9193,
      "step": 377
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019924134702041518,
      "loss": 0.8801,
      "step": 378
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019923414817114363,
      "loss": 0.8531,
      "step": 379
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019922691545935136,
      "loss": 0.898,
      "step": 380
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001992196488875065,
      "loss": 0.8899,
      "step": 381
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019921234845808865,
      "loss": 0.8989,
      "step": 382
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019920501417358905,
      "loss": 0.8736,
      "step": 383
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001991976460365104,
      "loss": 0.8667,
      "step": 384
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000199190244049367,
      "loss": 0.931,
      "step": 385
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019918280821468466,
      "loss": 0.9274,
      "step": 386
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019917533853500084,
      "loss": 0.9693,
      "step": 387
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001991678350128645,
      "loss": 0.9004,
      "step": 388
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019916029765083604,
      "loss": 0.8958,
      "step": 389
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019915272645148755,
      "loss": 0.9278,
      "step": 390
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001991451214174026,
      "loss": 0.9491,
      "step": 391
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019913748255117635,
      "loss": 0.9552,
      "step": 392
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019912980985541543,
      "loss": 0.9217,
      "step": 393
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001991221033327381,
      "loss": 0.8848,
      "step": 394
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019911436298577413,
      "loss": 0.9525,
      "step": 395
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019910658881716479,
      "loss": 0.9101,
      "step": 396
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019909878082956294,
      "loss": 0.8785,
      "step": 397
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019909093902563293,
      "loss": 0.8844,
      "step": 398
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019908306340805077,
      "loss": 0.9302,
      "step": 399
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019907515397950384,
      "loss": 0.9267,
      "step": 400
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019906721074269124,
      "loss": 0.8937,
      "step": 401
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001990592337003234,
      "loss": 0.8313,
      "step": 402
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019905122285512246,
      "loss": 0.871,
      "step": 403
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019904317820982202,
      "loss": 0.9051,
      "step": 404
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019903509976716718,
      "loss": 0.9469,
      "step": 405
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019902698752991468,
      "loss": 0.9029,
      "step": 406
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019901884150083272,
      "loss": 0.8691,
      "step": 407
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000199010661682701,
      "loss": 0.9306,
      "step": 408
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001990024480783108,
      "loss": 0.9056,
      "step": 409
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019899420069046495,
      "loss": 0.9342,
      "step": 410
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019898591952197773,
      "loss": 0.9267,
      "step": 411
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019897760457567503,
      "loss": 0.8407,
      "step": 412
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019896925585439424,
      "loss": 0.9135,
      "step": 413
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019896087336098423,
      "loss": 0.9391,
      "step": 414
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019895245709830546,
      "loss": 0.9099,
      "step": 415
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019894400706922983,
      "loss": 0.933,
      "step": 416
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019893552327664092,
      "loss": 0.9021,
      "step": 417
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019892700572343364,
      "loss": 0.8709,
      "step": 418
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019891845441251453,
      "loss": 0.8602,
      "step": 419
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019890986934680162,
      "loss": 0.9202,
      "step": 420
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019890125052922448,
      "loss": 0.8993,
      "step": 421
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019889259796272417,
      "loss": 0.9498,
      "step": 422
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001988839116502533,
      "loss": 0.9078,
      "step": 423
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019887519159477596,
      "loss": 0.9603,
      "step": 424
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019886643779926778,
      "loss": 0.8907,
      "step": 425
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019885765026671584,
      "loss": 0.8898,
      "step": 426
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019884882900011888,
      "loss": 0.8384,
      "step": 427
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019883997400248698,
      "loss": 0.8335,
      "step": 428
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019883108527684186,
      "loss": 0.9511,
      "step": 429
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019882216282621665,
      "loss": 0.9052,
      "step": 430
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019881320665365607,
      "loss": 0.9466,
      "step": 431
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019880421676221627,
      "loss": 0.8598,
      "step": 432
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000198795193154965,
      "loss": 0.9417,
      "step": 433
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001987861358349814,
      "loss": 0.8406,
      "step": 434
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019877704480535625,
      "loss": 0.8625,
      "step": 435
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019876792006919168,
      "loss": 0.8947,
      "step": 436
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019875876162960153,
      "loss": 0.8995,
      "step": 437
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019874956948971083,
      "loss": 0.8765,
      "step": 438
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019874034365265645,
      "loss": 0.9205,
      "step": 439
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019873108412158655,
      "loss": 0.866,
      "step": 440
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001987217908996608,
      "loss": 0.865,
      "step": 441
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001987124639900505,
      "loss": 0.9128,
      "step": 442
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001987031033959382,
      "loss": 0.8214,
      "step": 443
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019869370912051822,
      "loss": 0.8771,
      "step": 444
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019868428116699623,
      "loss": 0.9318,
      "step": 445
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019867481953858942,
      "loss": 0.9386,
      "step": 446
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001986653242385264,
      "loss": 0.907,
      "step": 447
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019865579527004738,
      "loss": 0.9185,
      "step": 448
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019864623263640403,
      "loss": 0.8802,
      "step": 449
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019863663634085943,
      "loss": 0.3116,
      "step": 450
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019862700638668826,
      "loss": 0.9315,
      "step": 451
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001986173427771766,
      "loss": 0.8956,
      "step": 452
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019860764551562205,
      "loss": 0.9068,
      "step": 453
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019859791460533373,
      "loss": 0.8871,
      "step": 454
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019858815004963215,
      "loss": 0.9033,
      "step": 455
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019857835185184937,
      "loss": 0.9233,
      "step": 456
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019856852001532893,
      "loss": 0.929,
      "step": 457
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019855865454342583,
      "loss": 0.8704,
      "step": 458
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001985487554395065,
      "loss": 0.8917,
      "step": 459
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019853882270694896,
      "loss": 0.9507,
      "step": 460
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019852885634914262,
      "loss": 0.885,
      "step": 461
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019851885636948838,
      "loss": 0.9276,
      "step": 462
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001985088227713986,
      "loss": 0.9231,
      "step": 463
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019849875555829718,
      "loss": 0.9334,
      "step": 464
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019848865473361943,
      "loss": 0.9185,
      "step": 465
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019847852030081207,
      "loss": 0.9379,
      "step": 466
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019846835226333345,
      "loss": 0.8798,
      "step": 467
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019845815062465326,
      "loss": 0.9331,
      "step": 468
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019844791538825267,
      "loss": 0.9079,
      "step": 469
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001984376465576244,
      "loss": 0.9142,
      "step": 470
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001984273441362725,
      "loss": 0.9177,
      "step": 471
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019841700812771264,
      "loss": 0.9272,
      "step": 472
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019840663853547174,
      "loss": 0.8996,
      "step": 473
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001983962353630884,
      "loss": 0.9433,
      "step": 474
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019838579861411256,
      "loss": 0.846,
      "step": 475
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019837532829210564,
      "loss": 0.8911,
      "step": 476
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019836482440064055,
      "loss": 0.8944,
      "step": 477
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001983542869433016,
      "loss": 0.9251,
      "step": 478
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019834371592368456,
      "loss": 0.8996,
      "step": 479
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019833311134539666,
      "loss": 0.8755,
      "step": 480
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019832247321205658,
      "loss": 0.8779,
      "step": 481
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019831180152729455,
      "loss": 0.3315,
      "step": 482
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019830109629475208,
      "loss": 0.9043,
      "step": 483
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019829035751808225,
      "loss": 0.9028,
      "step": 484
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019827958520094948,
      "loss": 0.8756,
      "step": 485
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001982687793470298,
      "loss": 0.9136,
      "step": 486
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019825793996001048,
      "loss": 0.8528,
      "step": 487
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019824706704359043,
      "loss": 0.8718,
      "step": 488
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001982361606014799,
      "loss": 0.8615,
      "step": 489
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001982252206374005,
      "loss": 0.8908,
      "step": 490
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019821424715508543,
      "loss": 0.8948,
      "step": 491
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019820324015827926,
      "loss": 0.8483,
      "step": 492
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000198192199650738,
      "loss": 0.8536,
      "step": 493
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019818112563622911,
      "loss": 0.8933,
      "step": 494
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019817001811853148,
      "loss": 0.9163,
      "step": 495
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019815887710143538,
      "loss": 0.8661,
      "step": 496
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019814770258874258,
      "loss": 0.9027,
      "step": 497
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019813649458426627,
      "loss": 0.9083,
      "step": 498
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000198125253091831,
      "loss": 0.9259,
      "step": 499
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019811397811527286,
      "loss": 0.9422,
      "step": 500
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001981026696584393,
      "loss": 0.881,
      "step": 501
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001980913277251892,
      "loss": 0.9025,
      "step": 502
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001980799523193928,
      "loss": 0.9256,
      "step": 503
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001980685434449319,
      "loss": 0.8604,
      "step": 504
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019805710110569965,
      "loss": 0.3219,
      "step": 505
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019804562530560059,
      "loss": 0.942,
      "step": 506
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001980341160485507,
      "loss": 0.9316,
      "step": 507
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019802257333847737,
      "loss": 0.8756,
      "step": 508
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019801099717931951,
      "loss": 0.8887,
      "step": 509
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019799938757502722,
      "loss": 0.9273,
      "step": 510
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019798774452956224,
      "loss": 0.9497,
      "step": 511
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019797606804689761,
      "loss": 0.8419,
      "step": 512
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019796435813101778,
      "loss": 0.9074,
      "step": 513
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001979526147859186,
      "loss": 0.8576,
      "step": 514
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019794083801560737,
      "loss": 0.867,
      "step": 515
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019792902782410284,
      "loss": 0.9158,
      "step": 516
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000197917184215435,
      "loss": 0.9266,
      "step": 517
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019790530719364542,
      "loss": 0.9508,
      "step": 518
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000197893396762787,
      "loss": 0.877,
      "step": 519
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000197881452926924,
      "loss": 0.9533,
      "step": 520
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019786947569013213,
      "loss": 0.9167,
      "step": 521
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019785746505649848,
      "loss": 0.8785,
      "step": 522
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019784542103012153,
      "loss": 0.8454,
      "step": 523
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001978333436151112,
      "loss": 0.958,
      "step": 524
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001978212328155888,
      "loss": 0.8849,
      "step": 525
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019780908863568691,
      "loss": 0.8775,
      "step": 526
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019779691107954964,
      "loss": 0.9002,
      "step": 527
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019778470015133248,
      "loss": 0.9716,
      "step": 528
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019777245585520222,
      "loss": 0.8743,
      "step": 529
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019776017819533707,
      "loss": 0.9205,
      "step": 530
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001977478671759267,
      "loss": 0.9135,
      "step": 531
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019773552280117208,
      "loss": 0.9031,
      "step": 532
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019772314507528557,
      "loss": 0.9366,
      "step": 533
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019771073400249095,
      "loss": 0.8974,
      "step": 534
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001976982895870233,
      "loss": 0.8727,
      "step": 535
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019768581183312925,
      "loss": 0.8519,
      "step": 536
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019767330074506662,
      "loss": 0.855,
      "step": 537
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019766075632710463,
      "loss": 0.9156,
      "step": 538
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000197648178583524,
      "loss": 0.8368,
      "step": 539
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001976355675186167,
      "loss": 0.3309,
      "step": 540
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019762292313668612,
      "loss": 0.8335,
      "step": 541
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019761024544204703,
      "loss": 0.9016,
      "step": 542
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019759753443902552,
      "loss": 0.9304,
      "step": 543
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001975847901319591,
      "loss": 0.8774,
      "step": 544
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001975720125251966,
      "loss": 0.8899,
      "step": 545
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019755920162309822,
      "loss": 0.9147,
      "step": 546
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019754635743003557,
      "loss": 0.9177,
      "step": 547
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019753347995039158,
      "loss": 0.8273,
      "step": 548
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001975205691885605,
      "loss": 0.8628,
      "step": 549
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019750762514894802,
      "loss": 0.9127,
      "step": 550
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019749464783597116,
      "loss": 0.9152,
      "step": 551
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019748163725405823,
      "loss": 0.9351,
      "step": 552
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019746859340764896,
      "loss": 0.881,
      "step": 553
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019745551630119445,
      "loss": 0.9127,
      "step": 554
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019744240593915708,
      "loss": 0.9262,
      "step": 555
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001974292623260106,
      "loss": 0.9136,
      "step": 556
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019741608546624016,
      "loss": 0.924,
      "step": 557
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019740287536434216,
      "loss": 0.8829,
      "step": 558
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019738963202482445,
      "loss": 0.9177,
      "step": 559
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001973763554522061,
      "loss": 0.8452,
      "step": 560
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019736304565101768,
      "loss": 0.882,
      "step": 561
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019734970262580098,
      "loss": 0.8349,
      "step": 562
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001973363263811091,
      "loss": 0.8953,
      "step": 563
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019732291692150654,
      "loss": 0.9129,
      "step": 564
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019730947425156917,
      "loss": 0.8283,
      "step": 565
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019729599837588416,
      "loss": 0.938,
      "step": 566
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019728248929904997,
      "loss": 0.8881,
      "step": 567
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019726894702567637,
      "loss": 0.8785,
      "step": 568
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019725537156038456,
      "loss": 0.847,
      "step": 569
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019724176290780701,
      "loss": 0.9505,
      "step": 570
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019722812107258753,
      "loss": 0.82,
      "step": 571
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019721444605938122,
      "loss": 0.86,
      "step": 572
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001972007378728545,
      "loss": 0.8471,
      "step": 573
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019718699651768518,
      "loss": 0.9212,
      "step": 574
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001971732219985623,
      "loss": 0.9367,
      "step": 575
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001971594143201863,
      "loss": 0.9323,
      "step": 576
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019714557348726882,
      "loss": 0.9325,
      "step": 577
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019713169950453299,
      "loss": 0.8159,
      "step": 578
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001971177923767131,
      "loss": 0.9138,
      "step": 579
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019710385210855474,
      "loss": 0.9172,
      "step": 580
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019708987870481495,
      "loss": 0.9244,
      "step": 581
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019707587217026194,
      "loss": 0.9517,
      "step": 582
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019706183250967532,
      "loss": 0.8526,
      "step": 583
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019704775972784595,
      "loss": 0.8564,
      "step": 584
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019703365382957598,
      "loss": 0.8574,
      "step": 585
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019701951481967893,
      "loss": 0.8908,
      "step": 586
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019700534270297956,
      "loss": 0.8891,
      "step": 587
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019699113748431393,
      "loss": 0.872,
      "step": 588
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019697689916852942,
      "loss": 0.8576,
      "step": 589
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001969626277604847,
      "loss": 0.9476,
      "step": 590
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001969483232650497,
      "loss": 0.8745,
      "step": 591
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001969339856871057,
      "loss": 0.9562,
      "step": 592
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019691961503154519,
      "loss": 0.913,
      "step": 593
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019690521130327204,
      "loss": 0.8275,
      "step": 594
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019689077450720131,
      "loss": 0.8748,
      "step": 595
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019687630464825946,
      "loss": 0.8956,
      "step": 596
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019686180173138408,
      "loss": 0.3463,
      "step": 597
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019684726576152422,
      "loss": 0.8967,
      "step": 598
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019683269674364003,
      "loss": 0.8992,
      "step": 599
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019681809468270306,
      "loss": 0.9446,
      "step": 600
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019680345958369606,
      "loss": 0.902,
      "step": 601
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019678879145161316,
      "loss": 0.9274,
      "step": 602
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019677409029145962,
      "loss": 0.8541,
      "step": 603
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001967593561082521,
      "loss": 0.3444,
      "step": 604
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019674458890701845,
      "loss": 0.8494,
      "step": 605
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001967297886927978,
      "loss": 0.8758,
      "step": 606
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019671495547064053,
      "loss": 0.87,
      "step": 607
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001967000892456084,
      "loss": 0.9183,
      "step": 608
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019668519002277423,
      "loss": 0.8879,
      "step": 609
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001966702578072223,
      "loss": 0.8584,
      "step": 610
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019665529260404796,
      "loss": 0.8307,
      "step": 611
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019664029441835798,
      "loss": 0.887,
      "step": 612
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019662526325527036,
      "loss": 0.8584,
      "step": 613
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019661019911991422,
      "loss": 0.8665,
      "step": 614
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019659510201743005,
      "loss": 0.9343,
      "step": 615
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019657997195296964,
      "loss": 0.8963,
      "step": 616
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019656480893169586,
      "loss": 0.8949,
      "step": 617
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019654961295878295,
      "loss": 0.8724,
      "step": 618
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019653438403941637,
      "loss": 0.8819,
      "step": 619
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001965191221787928,
      "loss": 0.9024,
      "step": 620
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001965038273821202,
      "loss": 0.8999,
      "step": 621
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019648849965461774,
      "loss": 0.8969,
      "step": 622
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001964731390015158,
      "loss": 0.8972,
      "step": 623
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019645774542805607,
      "loss": 0.8934,
      "step": 624
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001964423189394914,
      "loss": 0.9282,
      "step": 625
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001964268595410859,
      "loss": 0.8745,
      "step": 626
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019641136723811493,
      "loss": 0.9103,
      "step": 627
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019639584203586506,
      "loss": 0.9337,
      "step": 628
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019638028393963408,
      "loss": 0.3133,
      "step": 629
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019636469295473106,
      "loss": 0.9229,
      "step": 630
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019634906908647616,
      "loss": 0.9057,
      "step": 631
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019633341234020092,
      "loss": 0.8755,
      "step": 632
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019631772272124797,
      "loss": 0.8392,
      "step": 633
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019630200023497128,
      "loss": 0.3056,
      "step": 634
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001962862448867359,
      "loss": 0.9097,
      "step": 635
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019627045668191818,
      "loss": 0.9183,
      "step": 636
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001962546356259057,
      "loss": 0.8808,
      "step": 637
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001962387817240972,
      "loss": 0.8626,
      "step": 638
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019622289498190263,
      "loss": 0.8731,
      "step": 639
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019620697540474318,
      "loss": 0.8963,
      "step": 640
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019619102299805116,
      "loss": 0.8849,
      "step": 641
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019617503776727025,
      "loss": 0.8986,
      "step": 642
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019615901971785515,
      "loss": 0.8595,
      "step": 643
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019614296885527187,
      "loss": 0.3302,
      "step": 644
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019612688518499757,
      "loss": 0.3018,
      "step": 645
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001961107687125206,
      "loss": 0.9344,
      "step": 646
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019609461944334057,
      "loss": 0.825,
      "step": 647
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001960784373829682,
      "loss": 0.9352,
      "step": 648
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019606222253692543,
      "loss": 0.9319,
      "step": 649
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019604597491074545,
      "loss": 0.8763,
      "step": 650
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019602969450997247,
      "loss": 0.3185,
      "step": 651
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001960133813401621,
      "loss": 0.9191,
      "step": 652
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019599703540688096,
      "loss": 0.8785,
      "step": 653
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019598065671570694,
      "loss": 0.9284,
      "step": 654
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019596424527222904,
      "loss": 0.908,
      "step": 655
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001959478010820475,
      "loss": 0.8864,
      "step": 656
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019593132415077376,
      "loss": 0.909,
      "step": 657
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019591481448403032,
      "loss": 0.306,
      "step": 658
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019589827208745092,
      "loss": 0.8805,
      "step": 659
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019588169696668054,
      "loss": 0.8765,
      "step": 660
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019586508912737517,
      "loss": 0.9131,
      "step": 661
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019584844857520203,
      "loss": 0.9132,
      "step": 662
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019583177531583962,
      "loss": 0.9393,
      "step": 663
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001958150693549774,
      "loss": 0.9062,
      "step": 664
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019579833069831613,
      "loss": 0.9076,
      "step": 665
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001957815593515677,
      "loss": 0.8436,
      "step": 666
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001957647553204551,
      "loss": 0.8842,
      "step": 667
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019574791861071256,
      "loss": 0.8979,
      "step": 668
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019573104922808538,
      "loss": 0.3075,
      "step": 669
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019571414717833006,
      "loss": 0.8617,
      "step": 670
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019569721246721423,
      "loss": 0.8388,
      "step": 671
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019568024510051665,
      "loss": 0.8813,
      "step": 672
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019566324508402726,
      "loss": 0.803,
      "step": 673
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019564621242354713,
      "loss": 0.8794,
      "step": 674
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019562914712488842,
      "loss": 0.9609,
      "step": 675
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019561204919387452,
      "loss": 0.9316,
      "step": 676
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001955949186363399,
      "loss": 0.8449,
      "step": 677
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001955777554581301,
      "loss": 0.8994,
      "step": 678
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019556055966510195,
      "loss": 0.9351,
      "step": 679
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019554333126312322,
      "loss": 0.869,
      "step": 680
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019552607025807303,
      "loss": 0.8535,
      "step": 681
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019550877665584136,
      "loss": 0.9308,
      "step": 682
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019549145046232958,
      "loss": 0.8722,
      "step": 683
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019547409168344995,
      "loss": 0.9242,
      "step": 684
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019545670032512606,
      "loss": 0.8533,
      "step": 685
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019543927639329245,
      "loss": 0.8643,
      "step": 686
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001954218198938948,
      "loss": 0.9058,
      "step": 687
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019540433083289002,
      "loss": 0.9193,
      "step": 688
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019538680921624604,
      "loss": 0.8939,
      "step": 689
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001953692550499419,
      "loss": 0.8654,
      "step": 690
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001953516683399677,
      "loss": 0.941,
      "step": 691
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019533404909232477,
      "loss": 0.8463,
      "step": 692
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019531639731302545,
      "loss": 0.8314,
      "step": 693
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019529871300809327,
      "loss": 0.9452,
      "step": 694
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001952809961835627,
      "loss": 0.8655,
      "step": 695
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019526324684547949,
      "loss": 0.8721,
      "step": 696
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001952454649999003,
      "loss": 0.8908,
      "step": 697
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001952276506528931,
      "loss": 0.8443,
      "step": 698
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019520980381053677,
      "loss": 0.8854,
      "step": 699
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001951919244789213,
      "loss": 0.8639,
      "step": 700
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001951740126641479,
      "loss": 0.275,
      "step": 701
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019515606837232875,
      "loss": 0.9239,
      "step": 702
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019513809160958708,
      "loss": 0.8269,
      "step": 703
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019512008238205727,
      "loss": 0.9569,
      "step": 704
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019510204069588482,
      "loss": 0.8847,
      "step": 705
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019508396655722619,
      "loss": 0.8583,
      "step": 706
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019506585997224897,
      "loss": 0.9038,
      "step": 707
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001950477209471319,
      "loss": 0.3492,
      "step": 708
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019502954948806463,
      "loss": 0.9088,
      "step": 709
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019501134560124798,
      "loss": 0.9793,
      "step": 710
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019499310929289388,
      "loss": 0.8972,
      "step": 711
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001949748405692252,
      "loss": 0.8289,
      "step": 712
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019495653943647594,
      "loss": 0.8885,
      "step": 713
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019493820590089117,
      "loss": 0.843,
      "step": 714
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000194919839968727,
      "loss": 0.8156,
      "step": 715
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019490144164625059,
      "loss": 0.8999,
      "step": 716
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019488301093974016,
      "loss": 0.8792,
      "step": 717
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019486454785548496,
      "loss": 0.8608,
      "step": 718
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019484605239978534,
      "loss": 0.85,
      "step": 719
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019482752457895263,
      "loss": 0.8912,
      "step": 720
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019480896439930928,
      "loss": 0.8548,
      "step": 721
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019479037186718868,
      "loss": 0.8509,
      "step": 722
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019477174698893536,
      "loss": 0.8403,
      "step": 723
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019475308977090486,
      "loss": 0.9145,
      "step": 724
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019473440021946367,
      "loss": 0.9112,
      "step": 725
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019471567834098948,
      "loss": 0.9095,
      "step": 726
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001946969241418709,
      "loss": 0.9368,
      "step": 727
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019467813762850752,
      "loss": 0.9369,
      "step": 728
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019465931880731005,
      "loss": 0.8665,
      "step": 729
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019464046768470023,
      "loss": 0.8651,
      "step": 730
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019462158426711079,
      "loss": 0.8572,
      "step": 731
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019460266856098545,
      "loss": 0.9316,
      "step": 732
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000194583720572779,
      "loss": 0.9058,
      "step": 733
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019456474030895722,
      "loss": 0.8571,
      "step": 734
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019454572777599687,
      "loss": 0.9324,
      "step": 735
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019452668298038583,
      "loss": 0.9289,
      "step": 736
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019450760592862288,
      "loss": 0.8805,
      "step": 737
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019448849662721785,
      "loss": 0.9344,
      "step": 738
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001944693550826916,
      "loss": 0.8343,
      "step": 739
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019445018130157587,
      "loss": 0.8837,
      "step": 740
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001944309752904136,
      "loss": 0.8584,
      "step": 741
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001944117370557586,
      "loss": 0.8143,
      "step": 742
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019439246660417564,
      "loss": 0.8628,
      "step": 743
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019437316394224058,
      "loss": 0.9108,
      "step": 744
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019435382907654028,
      "loss": 0.8239,
      "step": 745
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019433446201367246,
      "loss": 0.8757,
      "step": 746
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019431506276024595,
      "loss": 0.933,
      "step": 747
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019429563132288057,
      "loss": 0.3464,
      "step": 748
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019427616770820695,
      "loss": 0.8539,
      "step": 749
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019425667192286697,
      "loss": 0.8401,
      "step": 750
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019423714397351325,
      "loss": 0.9108,
      "step": 751
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019421758386680947,
      "loss": 0.9334,
      "step": 752
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001941979916094304,
      "loss": 0.8979,
      "step": 753
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019417836720806155,
      "loss": 0.9735,
      "step": 754
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019415871066939962,
      "loss": 0.9018,
      "step": 755
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019413902200015207,
      "loss": 0.8241,
      "step": 756
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019411930120703756,
      "loss": 0.9246,
      "step": 757
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001940995482967855,
      "loss": 0.8737,
      "step": 758
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001940797632761364,
      "loss": 0.8602,
      "step": 759
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019405994615184161,
      "loss": 0.8974,
      "step": 760
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001940400969306636,
      "loss": 0.8338,
      "step": 761
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019402021561937556,
      "loss": 0.9215,
      "step": 762
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019400030222476182,
      "loss": 0.8946,
      "step": 763
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001939803567536176,
      "loss": 0.8873,
      "step": 764
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019396037921274909,
      "loss": 0.8622,
      "step": 765
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019394036960897334,
      "loss": 0.8218,
      "step": 766
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001939203279491185,
      "loss": 0.9189,
      "step": 767
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019390025424002343,
      "loss": 0.8732,
      "step": 768
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019388014848853812,
      "loss": 0.8592,
      "step": 769
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019386001070152345,
      "loss": 0.8775,
      "step": 770
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019383984088585114,
      "loss": 0.8353,
      "step": 771
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019381963904840398,
      "loss": 0.8891,
      "step": 772
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001937994051960756,
      "loss": 0.8681,
      "step": 773
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019377913933577053,
      "loss": 0.3416,
      "step": 774
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019375884147440433,
      "loss": 0.8552,
      "step": 775
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019373851161890335,
      "loss": 0.8371,
      "step": 776
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019371814977620498,
      "loss": 0.9098,
      "step": 777
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019369775595325737,
      "loss": 0.9145,
      "step": 778
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019367733015701982,
      "loss": 0.9259,
      "step": 779
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019365687239446228,
      "loss": 0.3169,
      "step": 780
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001936363826725658,
      "loss": 0.8511,
      "step": 781
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019361586099832218,
      "loss": 0.8788,
      "step": 782
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001935953073787343,
      "loss": 0.8683,
      "step": 783
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001935747218208158,
      "loss": 0.9018,
      "step": 784
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001935541043315913,
      "loss": 0.8483,
      "step": 785
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019353345491809624,
      "loss": 0.8845,
      "step": 786
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019351277358737706,
      "loss": 0.9044,
      "step": 787
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019349206034649093,
      "loss": 0.8863,
      "step": 788
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019347131520250607,
      "loss": 0.838,
      "step": 789
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019345053816250153,
      "loss": 0.9025,
      "step": 790
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019342972923356723,
      "loss": 0.9009,
      "step": 791
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019340888842280397,
      "loss": 0.8439,
      "step": 792
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019338801573732343,
      "loss": 0.8601,
      "step": 793
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001933671111842482,
      "loss": 0.9282,
      "step": 794
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001933461747707117,
      "loss": 0.9176,
      "step": 795
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019332520650385825,
      "loss": 0.9529,
      "step": 796
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019330420639084305,
      "loss": 0.8898,
      "step": 797
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019328317443883213,
      "loss": 0.8722,
      "step": 798
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001932621106550024,
      "loss": 0.8929,
      "step": 799
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019324101504654162,
      "loss": 0.8834,
      "step": 800
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019321988762064846,
      "loss": 0.8857,
      "step": 801
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001931987283845324,
      "loss": 0.8743,
      "step": 802
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001931775373454138,
      "loss": 0.8513,
      "step": 803
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019315631451052382,
      "loss": 0.8992,
      "step": 804
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019313505988710457,
      "loss": 0.9115,
      "step": 805
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001931137734824089,
      "loss": 0.9302,
      "step": 806
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019309245530370057,
      "loss": 0.875,
      "step": 807
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019307110535825417,
      "loss": 0.8394,
      "step": 808
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001930497236533551,
      "loss": 0.8963,
      "step": 809
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019302831019629965,
      "loss": 0.9033,
      "step": 810
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019300686499439492,
      "loss": 0.8689,
      "step": 811
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001929853880549588,
      "loss": 0.856,
      "step": 812
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001929638793853201,
      "loss": 0.8418,
      "step": 813
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001929423389928184,
      "loss": 0.8928,
      "step": 814
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001929207668848041,
      "loss": 0.9332,
      "step": 815
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001928991630686384,
      "loss": 0.8884,
      "step": 816
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019287752755169345,
      "loss": 0.8641,
      "step": 817
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019285586034135203,
      "loss": 0.3271,
      "step": 818
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019283416144500788,
      "loss": 0.9204,
      "step": 819
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019281243087006548,
      "loss": 0.841,
      "step": 820
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019279066862394015,
      "loss": 0.9076,
      "step": 821
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019276887471405798,
      "loss": 0.9198,
      "step": 822
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019274704914785593,
      "loss": 0.8501,
      "step": 823
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019272519193278172,
      "loss": 0.8904,
      "step": 824
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019270330307629387,
      "loss": 0.8761,
      "step": 825
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001926813825858617,
      "loss": 0.8626,
      "step": 826
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019265943046896532,
      "loss": 0.9,
      "step": 827
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019263744673309564,
      "loss": 0.8849,
      "step": 828
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001926154313857544,
      "loss": 0.8597,
      "step": 829
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019259338443445401,
      "loss": 0.8455,
      "step": 830
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019257130588671778,
      "loss": 0.8934,
      "step": 831
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001925491957500798,
      "loss": 0.9288,
      "step": 832
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019252705403208486,
      "loss": 0.8425,
      "step": 833
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001925048807402886,
      "loss": 0.8561,
      "step": 834
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001924826758822573,
      "loss": 0.8597,
      "step": 835
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019246043946556825,
      "loss": 0.8919,
      "step": 836
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019243817149780932,
      "loss": 0.8173,
      "step": 837
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019241587198657917,
      "loss": 0.9327,
      "step": 838
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019239354093948728,
      "loss": 0.9005,
      "step": 839
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019237117836415387,
      "loss": 0.3364,
      "step": 840
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019234878426820995,
      "loss": 0.8708,
      "step": 841
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019232635865929716,
      "loss": 0.8463,
      "step": 842
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019230390154506804,
      "loss": 0.8299,
      "step": 843
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001922814129331858,
      "loss": 0.8967,
      "step": 844
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019225889283132447,
      "loss": 0.8867,
      "step": 845
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019223634124716874,
      "loss": 0.8549,
      "step": 846
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019221375818841407,
      "loss": 0.8274,
      "step": 847
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019219114366276672,
      "loss": 0.882,
      "step": 848
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019216849767794355,
      "loss": 0.858,
      "step": 849
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019214582024167234,
      "loss": 0.8626,
      "step": 850
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019212311136169145,
      "loss": 0.8413,
      "step": 851
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019210037104575007,
      "loss": 0.8753,
      "step": 852
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000192077599301608,
      "loss": 0.8805,
      "step": 853
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019205479613703592,
      "loss": 0.8559,
      "step": 854
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001920319615598151,
      "loss": 0.8308,
      "step": 855
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001920090955777376,
      "loss": 0.8834,
      "step": 856
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019198619819860616,
      "loss": 0.8556,
      "step": 857
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019196326943023425,
      "loss": 0.8448,
      "step": 858
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019194030928044607,
      "loss": 0.8609,
      "step": 859
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019191731775707648,
      "loss": 0.8216,
      "step": 860
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019189429486797112,
      "loss": 0.9069,
      "step": 861
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019187124062098625,
      "loss": 0.8826,
      "step": 862
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019184815502398884,
      "loss": 0.8256,
      "step": 863
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019182503808485665,
      "loss": 0.8881,
      "step": 864
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001918018898114781,
      "loss": 0.9312,
      "step": 865
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019177871021175213,
      "loss": 0.8765,
      "step": 866
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019175549929358865,
      "loss": 0.8086,
      "step": 867
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019173225706490808,
      "loss": 0.8622,
      "step": 868
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019170898353364156,
      "loss": 0.3785,
      "step": 869
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019168567870773092,
      "loss": 0.8299,
      "step": 870
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019166234259512864,
      "loss": 0.8784,
      "step": 871
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019163897520379797,
      "loss": 0.8567,
      "step": 872
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001916155765417127,
      "loss": 0.8218,
      "step": 873
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019159214661685736,
      "loss": 0.3345,
      "step": 874
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019156868543722717,
      "loss": 0.9117,
      "step": 875
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000191545193010828,
      "loss": 0.8632,
      "step": 876
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019152166934567632,
      "loss": 0.8976,
      "step": 877
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019149811444979937,
      "loss": 0.8806,
      "step": 878
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019147452833123495,
      "loss": 0.8871,
      "step": 879
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001914509109980316,
      "loss": 0.8972,
      "step": 880
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019142726245824842,
      "loss": 0.8923,
      "step": 881
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019140358271995523,
      "loss": 0.8592,
      "step": 882
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019137987179123245,
      "loss": 0.8596,
      "step": 883
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019135612968017117,
      "loss": 0.8693,
      "step": 884
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019133235639487312,
      "loss": 0.9169,
      "step": 885
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019130855194345067,
      "loss": 0.3038,
      "step": 886
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001912847163340268,
      "loss": 0.8985,
      "step": 887
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019126084957473517,
      "loss": 0.3241,
      "step": 888
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019123695167372,
      "loss": 0.8868,
      "step": 889
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001912130226391362,
      "loss": 0.8585,
      "step": 890
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019118906247914924,
      "loss": 0.8555,
      "step": 891
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019116507120193533,
      "loss": 0.8125,
      "step": 892
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019114104881568116,
      "loss": 0.8548,
      "step": 893
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001911169953285841,
      "loss": 0.9317,
      "step": 894
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019109291074885217,
      "loss": 0.8884,
      "step": 895
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019106879508470388,
      "loss": 0.864,
      "step": 896
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001910446483443685,
      "loss": 0.9095,
      "step": 897
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001910204705360858,
      "loss": 0.8944,
      "step": 898
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019099626166810615,
      "loss": 0.8739,
      "step": 899
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001909720217486906,
      "loss": 0.8547,
      "step": 900
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019094775078611073,
      "loss": 0.9086,
      "step": 901
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019092344878864873,
      "loss": 0.8514,
      "step": 902
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019089911576459741,
      "loss": 0.3866,
      "step": 903
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019087475172226004,
      "loss": 0.91,
      "step": 904
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019085035666995066,
      "loss": 0.9193,
      "step": 905
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001908259306159938,
      "loss": 0.8683,
      "step": 906
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001908014735687245,
      "loss": 0.8378,
      "step": 907
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019077698553648852,
      "loss": 0.3251,
      "step": 908
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019075246652764213,
      "loss": 0.8998,
      "step": 909
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019072791655055208,
      "loss": 0.83,
      "step": 910
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019070333561359588,
      "loss": 0.8164,
      "step": 911
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001906787237251614,
      "loss": 0.9219,
      "step": 912
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019065408089364718,
      "loss": 0.9046,
      "step": 913
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019062940712746235,
      "loss": 0.9024,
      "step": 914
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019060470243502655,
      "loss": 0.9123,
      "step": 915
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001905799668247699,
      "loss": 0.9203,
      "step": 916
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019055520030513323,
      "loss": 0.8903,
      "step": 917
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001905304028845678,
      "loss": 0.8924,
      "step": 918
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019050557457153544,
      "loss": 0.8564,
      "step": 919
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019048071537450852,
      "loss": 0.8215,
      "step": 920
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019045582530196999,
      "loss": 0.9176,
      "step": 921
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019043090436241328,
      "loss": 0.8852,
      "step": 922
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019040595256434236,
      "loss": 0.875,
      "step": 923
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001903809699162718,
      "loss": 0.3346,
      "step": 924
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001903559564267266,
      "loss": 0.8645,
      "step": 925
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019033091210424234,
      "loss": 0.8371,
      "step": 926
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001903058369573651,
      "loss": 0.916,
      "step": 927
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001902807309946515,
      "loss": 0.9106,
      "step": 928
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019025559422466864,
      "loss": 0.9,
      "step": 929
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019023042665599423,
      "loss": 0.9065,
      "step": 930
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001902052282972163,
      "loss": 0.8925,
      "step": 931
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019017999915693358,
      "loss": 0.8549,
      "step": 932
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001901547392437552,
      "loss": 0.9453,
      "step": 933
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019012944856630087,
      "loss": 0.8756,
      "step": 934
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019010412713320064,
      "loss": 0.8509,
      "step": 935
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019007877495309528,
      "loss": 0.8428,
      "step": 936
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019005339203463582,
      "loss": 0.8566,
      "step": 937
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019002797838648397,
      "loss": 0.8992,
      "step": 938
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019000253401731186,
      "loss": 0.8328,
      "step": 939
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018997705893580199,
      "loss": 0.8611,
      "step": 940
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018995155315064754,
      "loss": 0.9113,
      "step": 941
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018992601667055202,
      "loss": 0.3357,
      "step": 942
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018990044950422945,
      "loss": 0.7908,
      "step": 943
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001898748516604044,
      "loss": 0.8637,
      "step": 944
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018984922314781177,
      "loss": 0.8553,
      "step": 945
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018982356397519704,
      "loss": 0.9095,
      "step": 946
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001897978741513161,
      "loss": 0.8806,
      "step": 947
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001897721536849353,
      "loss": 0.9207,
      "step": 948
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018974640258483146,
      "loss": 0.9215,
      "step": 949
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018972062085979187,
      "loss": 0.9114,
      "step": 950
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018969480851861422,
      "loss": 0.8113,
      "step": 951
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018966896557010666,
      "loss": 0.9182,
      "step": 952
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018964309202308788,
      "loss": 0.8817,
      "step": 953
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018961718788638687,
      "loss": 0.889,
      "step": 954
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018959125316884312,
      "loss": 0.8668,
      "step": 955
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018956528787930657,
      "loss": 0.891,
      "step": 956
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018953929202663756,
      "loss": 0.8539,
      "step": 957
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001895132656197069,
      "loss": 0.9011,
      "step": 958
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018948720866739582,
      "loss": 0.8804,
      "step": 959
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018946112117859595,
      "loss": 0.8366,
      "step": 960
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001894350031622093,
      "loss": 0.9133,
      "step": 961
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001894088546271484,
      "loss": 0.3365,
      "step": 962
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018938267558233612,
      "loss": 0.8924,
      "step": 963
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018935646603670576,
      "loss": 0.8481,
      "step": 964
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000189330225999201,
      "loss": 0.8435,
      "step": 965
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018930395547877603,
      "loss": 0.8441,
      "step": 966
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018927765448439527,
      "loss": 0.8883,
      "step": 967
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018925132302503373,
      "loss": 0.8127,
      "step": 968
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018922496110967664,
      "loss": 0.8358,
      "step": 969
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00018919856874731976,
      "loss": 0.9309,
      "step": 970
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018917214594696912,
      "loss": 0.8999,
      "step": 971
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018914569271764127,
      "loss": 0.8732,
      "step": 972
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018911920906836303,
      "loss": 0.8712,
      "step": 973
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018909269500817164,
      "loss": 0.324,
      "step": 974
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018906615054611474,
      "loss": 0.8713,
      "step": 975
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018903957569125032,
      "loss": 0.9129,
      "step": 976
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018901297045264675,
      "loss": 0.9059,
      "step": 977
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018898633483938274,
      "loss": 0.8503,
      "step": 978
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001889596688605474,
      "loss": 0.8419,
      "step": 979
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018893297252524022,
      "loss": 0.3472,
      "step": 980
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018890624584257096,
      "loss": 0.9227,
      "step": 981
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018887948882165986,
      "loss": 0.9134,
      "step": 982
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018885270147163744,
      "loss": 0.9394,
      "step": 983
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018882588380164452,
      "loss": 0.8842,
      "step": 984
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018879903582083238,
      "loss": 0.861,
      "step": 985
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018877215753836256,
      "loss": 0.8689,
      "step": 986
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000188745248963407,
      "loss": 0.8526,
      "step": 987
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001887183101051479,
      "loss": 0.9096,
      "step": 988
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018869134097277787,
      "loss": 0.888,
      "step": 989
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001886643415754998,
      "loss": 0.8513,
      "step": 990
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018863731192252691,
      "loss": 0.8708,
      "step": 991
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018861025202308282,
      "loss": 0.9001,
      "step": 992
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018858316188640135,
      "loss": 0.8353,
      "step": 993
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018855604152172675,
      "loss": 0.8502,
      "step": 994
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001885288909383135,
      "loss": 0.8731,
      "step": 995
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018850171014542645,
      "loss": 0.8285,
      "step": 996
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001884744991523407,
      "loss": 0.9314,
      "step": 997
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001884472579683417,
      "loss": 0.3169,
      "step": 998
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018841998660272524,
      "loss": 0.8513,
      "step": 999
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018839268506479732,
      "loss": 0.8654,
      "step": 1000
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018836535336387428,
      "loss": 0.9252,
      "step": 1001
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018833799150928278,
      "loss": 0.8847,
      "step": 1002
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001883105995103597,
      "loss": 0.8193,
      "step": 1003
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018828317737645226,
      "loss": 0.3161,
      "step": 1004
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018825572511691795,
      "loss": 0.9138,
      "step": 1005
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001882282427411246,
      "loss": 0.8567,
      "step": 1006
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018820073025845017,
      "loss": 0.9199,
      "step": 1007
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000188173187678283,
      "loss": 0.8488,
      "step": 1008
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001881456150100217,
      "loss": 0.9025,
      "step": 1009
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018811801226307513,
      "loss": 0.8809,
      "step": 1010
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018809037944686244,
      "loss": 0.8918,
      "step": 1011
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001880627165708129,
      "loss": 0.8806,
      "step": 1012
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001880350236443663,
      "loss": 0.8461,
      "step": 1013
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018800730067697243,
      "loss": 0.9014,
      "step": 1014
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001879795476780915,
      "loss": 0.8685,
      "step": 1015
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018795176465719385,
      "loss": 0.7983,
      "step": 1016
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018792395162376018,
      "loss": 0.8038,
      "step": 1017
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018789610858728132,
      "loss": 0.8871,
      "step": 1018
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018786823555725838,
      "loss": 0.8802,
      "step": 1019
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018784033254320276,
      "loss": 0.9084,
      "step": 1020
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000187812399554636,
      "loss": 0.8684,
      "step": 1021
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018778443660108997,
      "loss": 0.8682,
      "step": 1022
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018775644369210665,
      "loss": 0.8799,
      "step": 1023
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00018772842083723834,
      "loss": 0.9033,
      "step": 1024
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001877003680460475,
      "loss": 0.8709,
      "step": 1025
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018767228532810685,
      "loss": 0.8232,
      "step": 1026
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018764417269299925,
      "loss": 0.3025,
      "step": 1027
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018761603015031786,
      "loss": 0.8684,
      "step": 1028
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018758785770966595,
      "loss": 0.8921,
      "step": 1029
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018755965538065712,
      "loss": 0.8208,
      "step": 1030
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018753142317291503,
      "loss": 0.9352,
      "step": 1031
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018750316109607363,
      "loss": 0.8739,
      "step": 1032
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000187474869159777,
      "loss": 0.9015,
      "step": 1033
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018744654737367947,
      "loss": 0.8809,
      "step": 1034
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018741819574744556,
      "loss": 0.8681,
      "step": 1035
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018738981429074984,
      "loss": 0.8686,
      "step": 1036
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018736140301327722,
      "loss": 0.8692,
      "step": 1037
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018733296192472275,
      "loss": 0.841,
      "step": 1038
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018730449103479158,
      "loss": 0.8689,
      "step": 1039
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001872759903531991,
      "loss": 0.869,
      "step": 1040
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018724745988967084,
      "loss": 0.8554,
      "step": 1041
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018721889965394247,
      "loss": 0.8452,
      "step": 1042
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001871903096557599,
      "loss": 0.775,
      "step": 1043
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018716168990487905,
      "loss": 0.881,
      "step": 1044
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018713304041106618,
      "loss": 0.8571,
      "step": 1045
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018710436118409755,
      "loss": 0.8341,
      "step": 1046
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018707565223375966,
      "loss": 0.9099,
      "step": 1047
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000187046913569849,
      "loss": 0.8058,
      "step": 1048
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018701814520217247,
      "loss": 0.8749,
      "step": 1049
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018698934714054687,
      "loss": 0.855,
      "step": 1050
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018696051939479918,
      "loss": 0.9107,
      "step": 1051
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018693166197476658,
      "loss": 0.8839,
      "step": 1052
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018690277489029632,
      "loss": 0.9632,
      "step": 1053
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018687385815124578,
      "loss": 0.8881,
      "step": 1054
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001868449117674825,
      "loss": 0.9057,
      "step": 1055
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018681593574888407,
      "loss": 0.8673,
      "step": 1056
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018678693010533824,
      "loss": 0.8431,
      "step": 1057
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001867578948467429,
      "loss": 0.9048,
      "step": 1058
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001867288299830059,
      "loss": 0.8468,
      "step": 1059
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018669973552404537,
      "loss": 0.9279,
      "step": 1060
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001866706114797894,
      "loss": 0.8875,
      "step": 1061
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018664145786017632,
      "loss": 0.8627,
      "step": 1062
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018661227467515443,
      "loss": 0.91,
      "step": 1063
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018658306193468214,
      "loss": 0.8267,
      "step": 1064
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000186553819648728,
      "loss": 0.8728,
      "step": 1065
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018652454782727059,
      "loss": 0.8477,
      "step": 1066
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018649524648029856,
      "loss": 0.8697,
      "step": 1067
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018646591561781065,
      "loss": 0.8475,
      "step": 1068
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018643655524981577,
      "loss": 0.8682,
      "step": 1069
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018640716538633273,
      "loss": 0.8592,
      "step": 1070
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018637774603739052,
      "loss": 0.882,
      "step": 1071
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018634829721302817,
      "loss": 0.8606,
      "step": 1072
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018631881892329467,
      "loss": 0.8729,
      "step": 1073
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018628931117824928,
      "loss": 0.9118,
      "step": 1074
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018625977398796105,
      "loss": 0.9144,
      "step": 1075
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018623020736250934,
      "loss": 0.7951,
      "step": 1076
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018620061131198332,
      "loss": 0.862,
      "step": 1077
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018617098584648237,
      "loss": 0.8376,
      "step": 1078
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018614133097611576,
      "loss": 0.9666,
      "step": 1079
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018611164671100295,
      "loss": 0.8684,
      "step": 1080
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00018608193306127334,
      "loss": 0.9368,
      "step": 1081
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001860521900370664,
      "loss": 0.8771,
      "step": 1082
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018602241764853155,
      "loss": 0.8214,
      "step": 1083
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001859926159058283,
      "loss": 0.8701,
      "step": 1084
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018596278481912613,
      "loss": 0.3415,
      "step": 1085
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001859329243986046,
      "loss": 0.8218,
      "step": 1086
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001859030346544532,
      "loss": 0.7754,
      "step": 1087
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018587311559687146,
      "loss": 0.8399,
      "step": 1088
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018584316723606898,
      "loss": 0.8395,
      "step": 1089
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018581318958226523,
      "loss": 0.824,
      "step": 1090
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018578318264568977,
      "loss": 0.8725,
      "step": 1091
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018575314643658213,
      "loss": 0.9016,
      "step": 1092
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001857230809651918,
      "loss": 0.8657,
      "step": 1093
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018569298624177832,
      "loss": 0.9135,
      "step": 1094
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018566286227661113,
      "loss": 0.8982,
      "step": 1095
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018563270907996965,
      "loss": 0.8322,
      "step": 1096
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018560252666214344,
      "loss": 0.9047,
      "step": 1097
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018557231503343183,
      "loss": 0.9152,
      "step": 1098
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018554207420414415,
      "loss": 0.8625,
      "step": 1099
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018551180418459982,
      "loss": 0.8703,
      "step": 1100
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018548150498512808,
      "loss": 0.9233,
      "step": 1101
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018545117661606823,
      "loss": 0.8414,
      "step": 1102
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018542081908776945,
      "loss": 0.9538,
      "step": 1103
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001853904324105909,
      "loss": 0.8525,
      "step": 1104
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018536001659490172,
      "loss": 0.9143,
      "step": 1105
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001853295716510809,
      "loss": 0.8335,
      "step": 1106
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001852990975895175,
      "loss": 0.8716,
      "step": 1107
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018526859442061043,
      "loss": 0.9113,
      "step": 1108
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001852380621547685,
      "loss": 0.9605,
      "step": 1109
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001852075008024105,
      "loss": 0.8922,
      "step": 1110
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018517691037396524,
      "loss": 0.8393,
      "step": 1111
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018514629087987123,
      "loss": 0.8219,
      "step": 1112
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001851156423305771,
      "loss": 0.8423,
      "step": 1113
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001850849647365413,
      "loss": 0.8947,
      "step": 1114
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018505425810823217,
      "loss": 0.8868,
      "step": 1115
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018502352245612806,
      "loss": 0.8312,
      "step": 1116
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018499275779071712,
      "loss": 0.8494,
      "step": 1117
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018496196412249747,
      "loss": 0.8493,
      "step": 1118
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018493114146197708,
      "loss": 0.8395,
      "step": 1119
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001849002898196738,
      "loss": 0.8691,
      "step": 1120
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018486940920611547,
      "loss": 0.867,
      "step": 1121
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001848384996318397,
      "loss": 0.8047,
      "step": 1122
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018480756110739402,
      "loss": 0.8772,
      "step": 1123
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018477659364333587,
      "loss": 0.8666,
      "step": 1124
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018474559725023255,
      "loss": 0.9301,
      "step": 1125
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018471457193866125,
      "loss": 0.8887,
      "step": 1126
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018468351771920891,
      "loss": 0.8736,
      "step": 1127
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018465243460247252,
      "loss": 0.9223,
      "step": 1128
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001846213225990588,
      "loss": 0.8725,
      "step": 1129
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001845901817195844,
      "loss": 0.868,
      "step": 1130
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018455901197467572,
      "loss": 0.8324,
      "step": 1131
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018452781337496913,
      "loss": 0.8942,
      "step": 1132
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018449658593111082,
      "loss": 0.9034,
      "step": 1133
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018446532965375674,
      "loss": 0.8697,
      "step": 1134
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00018443404455357278,
      "loss": 0.8744,
      "step": 1135
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001844027306412346,
      "loss": 0.8405,
      "step": 1136
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018437138792742772,
      "loss": 0.8521,
      "step": 1137
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001843400164228475,
      "loss": 0.8783,
      "step": 1138
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001843086161381991,
      "loss": 0.8765,
      "step": 1139
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001842771870841975,
      "loss": 0.8346,
      "step": 1140
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001842457292715675,
      "loss": 0.8399,
      "step": 1141
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018421424271104374,
      "loss": 0.8257,
      "step": 1142
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018418272741337063,
      "loss": 0.8954,
      "step": 1143
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001841511833893024,
      "loss": 0.8528,
      "step": 1144
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018411961064960315,
      "loss": 0.8211,
      "step": 1145
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018408800920504664,
      "loss": 0.9132,
      "step": 1146
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018405637906641654,
      "loss": 0.8441,
      "step": 1147
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001840247202445063,
      "loss": 0.8945,
      "step": 1148
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018399303275011902,
      "loss": 0.8466,
      "step": 1149
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018396131659406784,
      "loss": 0.8572,
      "step": 1150
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018392957178717544,
      "loss": 0.8869,
      "step": 1151
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001838977983402744,
      "loss": 0.8414,
      "step": 1152
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018386599626420708,
      "loss": 0.9077,
      "step": 1153
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018383416556982552,
      "loss": 0.8374,
      "step": 1154
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018380230626799163,
      "loss": 0.8427,
      "step": 1155
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000183770418369577,
      "loss": 0.8674,
      "step": 1156
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018373850188546303,
      "loss": 0.8388,
      "step": 1157
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001837065568265408,
      "loss": 0.9035,
      "step": 1158
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001836745832037113,
      "loss": 0.8683,
      "step": 1159
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001836425810278851,
      "loss": 0.3369,
      "step": 1160
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018361055030998257,
      "loss": 0.8495,
      "step": 1161
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018357849106093386,
      "loss": 0.857,
      "step": 1162
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001835464032916788,
      "loss": 0.909,
      "step": 1163
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018351428701316694,
      "loss": 0.8902,
      "step": 1164
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018348214223635768,
      "loss": 0.865,
      "step": 1165
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018344996897222,
      "loss": 0.8526,
      "step": 1166
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018341776723173266,
      "loss": 0.868,
      "step": 1167
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018338553702588418,
      "loss": 0.9114,
      "step": 1168
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018335327836567272,
      "loss": 0.8128,
      "step": 1169
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018332099126210614,
      "loss": 0.7852,
      "step": 1170
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001832886757262021,
      "loss": 0.9299,
      "step": 1171
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001832563317689879,
      "loss": 0.8432,
      "step": 1172
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001832239594015005,
      "loss": 0.9003,
      "step": 1173
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001831915586347867,
      "loss": 0.8123,
      "step": 1174
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018315912947990278,
      "loss": 0.7916,
      "step": 1175
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001831266719479149,
      "loss": 0.8427,
      "step": 1176
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001830941860498988,
      "loss": 0.8267,
      "step": 1177
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018306167179693987,
      "loss": 0.8302,
      "step": 1178
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001830291292001333,
      "loss": 0.8484,
      "step": 1179
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018299655827058386,
      "loss": 0.3759,
      "step": 1180
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000182963959019406,
      "loss": 0.8727,
      "step": 1181
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018293133145772382,
      "loss": 0.8863,
      "step": 1182
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018289867559667116,
      "loss": 0.7889,
      "step": 1183
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018286599144739138,
      "loss": 0.8935,
      "step": 1184
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018283327902103764,
      "loss": 0.9466,
      "step": 1185
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018280053832877267,
      "loss": 0.8567,
      "step": 1186
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018276776938176882,
      "loss": 0.8822,
      "step": 1187
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001827349721912082,
      "loss": 0.8684,
      "step": 1188
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001827021467682824,
      "loss": 0.917,
      "step": 1189
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018266929312419272,
      "loss": 0.828,
      "step": 1190
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018263641127015014,
      "loss": 0.8742,
      "step": 1191
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00018260350121737518,
      "loss": 0.8417,
      "step": 1192
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018257056297709805,
      "loss": 0.8946,
      "step": 1193
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001825375965605585,
      "loss": 0.8627,
      "step": 1194
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000182504601979006,
      "loss": 0.8533,
      "step": 1195
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018247157924369955,
      "loss": 0.919,
      "step": 1196
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001824385283659078,
      "loss": 0.902,
      "step": 1197
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018240544935690895,
      "loss": 0.3657,
      "step": 1198
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018237234222799085,
      "loss": 0.8326,
      "step": 1199
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001823392069904509,
      "loss": 0.8882,
      "step": 1200
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001823060436555962,
      "loss": 0.8455,
      "step": 1201
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001822728522347433,
      "loss": 0.8655,
      "step": 1202
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001822396327392184,
      "loss": 0.8587,
      "step": 1203
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001822063851803573,
      "loss": 0.8397,
      "step": 1204
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018217310956950532,
      "loss": 0.8577,
      "step": 1205
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018213980591801735,
      "loss": 0.8901,
      "step": 1206
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018210647423725796,
      "loss": 0.8363,
      "step": 1207
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001820731145386012,
      "loss": 0.8784,
      "step": 1208
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018203972683343062,
      "loss": 0.8277,
      "step": 1209
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018200631113313943,
      "loss": 0.8632,
      "step": 1210
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018197286744913035,
      "loss": 0.9129,
      "step": 1211
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018193939579281567,
      "loss": 0.7932,
      "step": 1212
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018190589617561722,
      "loss": 0.8192,
      "step": 1213
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018187236860896632,
      "loss": 0.8537,
      "step": 1214
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001818388131043039,
      "loss": 0.3712,
      "step": 1215
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018180522967308038,
      "loss": 0.893,
      "step": 1216
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018177161832675574,
      "loss": 0.8541,
      "step": 1217
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018173797907679946,
      "loss": 0.9283,
      "step": 1218
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018170431193469054,
      "loss": 0.8915,
      "step": 1219
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001816706169119175,
      "loss": 0.8876,
      "step": 1220
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001816368940199784,
      "loss": 0.9012,
      "step": 1221
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018160314327038082,
      "loss": 0.9371,
      "step": 1222
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018156936467464177,
      "loss": 0.3159,
      "step": 1223
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018153555824428782,
      "loss": 0.333,
      "step": 1224
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018150172399085504,
      "loss": 0.8821,
      "step": 1225
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000181467861925889,
      "loss": 0.8755,
      "step": 1226
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018143397206094473,
      "loss": 0.8325,
      "step": 1227
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018140005440758672,
      "loss": 0.8367,
      "step": 1228
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018136610897738904,
      "loss": 0.8385,
      "step": 1229
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018133213578193513,
      "loss": 0.8935,
      "step": 1230
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000181298134832818,
      "loss": 0.8947,
      "step": 1231
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018126410614164006,
      "loss": 0.8419,
      "step": 1232
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018123004972001323,
      "loss": 0.8569,
      "step": 1233
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018119596557955884,
      "loss": 0.9125,
      "step": 1234
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018116185373190776,
      "loss": 0.902,
      "step": 1235
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018112771418870026,
      "loss": 0.8556,
      "step": 1236
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018109354696158603,
      "loss": 0.8482,
      "step": 1237
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018105935206222425,
      "loss": 0.8441,
      "step": 1238
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018102512950228362,
      "loss": 0.879,
      "step": 1239
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001809908792934421,
      "loss": 0.8469,
      "step": 1240
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018095660144738728,
      "loss": 0.8828,
      "step": 1241
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018092229597581599,
      "loss": 0.3365,
      "step": 1242
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00018088796289043459,
      "loss": 0.8402,
      "step": 1243
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001808536022029589,
      "loss": 0.3391,
      "step": 1244
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001808192139251141,
      "loss": 0.8861,
      "step": 1245
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001807847980686348,
      "loss": 0.8825,
      "step": 1246
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000180750354645265,
      "loss": 0.8763,
      "step": 1247
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018071588366675814,
      "loss": 0.8425,
      "step": 1248
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018068138514487704,
      "loss": 0.8668,
      "step": 1249
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018064685909139393,
      "loss": 0.8544,
      "step": 1250
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018061230551809043,
      "loss": 0.8627,
      "step": 1251
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018057772443675756,
      "loss": 0.887,
      "step": 1252
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001805431158591957,
      "loss": 0.902,
      "step": 1253
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018050847979721467,
      "loss": 0.8729,
      "step": 1254
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001804738162626336,
      "loss": 0.8628,
      "step": 1255
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018043912526728102,
      "loss": 0.3294,
      "step": 1256
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018040440682299486,
      "loss": 0.8714,
      "step": 1257
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018036966094162236,
      "loss": 0.8395,
      "step": 1258
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001803348876350202,
      "loss": 0.9105,
      "step": 1259
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018030008691505436,
      "loss": 0.8918,
      "step": 1260
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018026525879360014,
      "loss": 0.8581,
      "step": 1261
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018023040328254232,
      "loss": 0.8498,
      "step": 1262
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018019552039377488,
      "loss": 0.8844,
      "step": 1263
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018016061013920124,
      "loss": 0.8422,
      "step": 1264
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018012567253073415,
      "loss": 0.9353,
      "step": 1265
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001800907075802956,
      "loss": 0.893,
      "step": 1266
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00018005571529981703,
      "loss": 0.8987,
      "step": 1267
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001800206957012391,
      "loss": 0.9154,
      "step": 1268
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017998564879651195,
      "loss": 0.8744,
      "step": 1269
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017995057459759486,
      "loss": 0.8811,
      "step": 1270
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017991547311645652,
      "loss": 0.8582,
      "step": 1271
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017988034436507491,
      "loss": 0.8528,
      "step": 1272
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017984518835543732,
      "loss": 0.887,
      "step": 1273
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017981000509954034,
      "loss": 0.825,
      "step": 1274
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017977479460938985,
      "loss": 0.7989,
      "step": 1275
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000179739556897001,
      "loss": 0.9111,
      "step": 1276
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001797042919743983,
      "loss": 0.8581,
      "step": 1277
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017966899985361547,
      "loss": 0.8349,
      "step": 1278
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017963368054669556,
      "loss": 0.9263,
      "step": 1279
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001795983340656909,
      "loss": 0.8991,
      "step": 1280
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017956296042266303,
      "loss": 0.3301,
      "step": 1281
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017952755962968284,
      "loss": 0.8557,
      "step": 1282
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017949213169883043,
      "loss": 0.9151,
      "step": 1283
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017945667664219517,
      "loss": 0.9016,
      "step": 1284
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001794211944718757,
      "loss": 0.8374,
      "step": 1285
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001793856851999799,
      "loss": 0.8117,
      "step": 1286
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017935014883862493,
      "loss": 0.8642,
      "step": 1287
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017931458539993713,
      "loss": 0.8649,
      "step": 1288
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017927899489605213,
      "loss": 0.8665,
      "step": 1289
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001792433773391148,
      "loss": 0.367,
      "step": 1290
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017920773274127924,
      "loss": 0.8509,
      "step": 1291
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001791720611147087,
      "loss": 0.8425,
      "step": 1292
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001791363624715758,
      "loss": 0.8484,
      "step": 1293
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001791006368240622,
      "loss": 0.8448,
      "step": 1294
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001790648841843589,
      "loss": 0.9218,
      "step": 1295
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017902910456466615,
      "loss": 0.8057,
      "step": 1296
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017899329797719325,
      "loss": 0.8369,
      "step": 1297
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017895746443415887,
      "loss": 0.8188,
      "step": 1298
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001789216039477907,
      "loss": 0.9183,
      "step": 1299
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001788857165303258,
      "loss": 0.8865,
      "step": 1300
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001788498021940103,
      "loss": 0.8827,
      "step": 1301
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017881386095109955,
      "loss": 0.8627,
      "step": 1302
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00017877789281385814,
      "loss": 0.9231,
      "step": 1303
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017874189779455973,
      "loss": 0.8614,
      "step": 1304
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017870587590548725,
      "loss": 0.8836,
      "step": 1305
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017866982715893274,
      "loss": 0.8624,
      "step": 1306
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017863375156719746,
      "loss": 0.8347,
      "step": 1307
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017859764914259177,
      "loss": 0.8757,
      "step": 1308
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017856151989743515,
      "loss": 0.885,
      "step": 1309
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017852536384405635,
      "loss": 0.8953,
      "step": 1310
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017848918099479323,
      "loss": 0.8467,
      "step": 1311
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017845297136199277,
      "loss": 0.8777,
      "step": 1312
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017841673495801104,
      "loss": 0.861,
      "step": 1313
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017838047179521334,
      "loss": 0.8578,
      "step": 1314
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001783441818859741,
      "loss": 0.8981,
      "step": 1315
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001783078652426767,
      "loss": 0.882,
      "step": 1316
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017827152187771395,
      "loss": 0.915,
      "step": 1317
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017823515180348748,
      "loss": 0.895,
      "step": 1318
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017819875503240825,
      "loss": 0.8759,
      "step": 1319
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001781623315768962,
      "loss": 0.9024,
      "step": 1320
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017812588144938042,
      "loss": 0.8277,
      "step": 1321
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001780894046622991,
      "loss": 0.3216,
      "step": 1322
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017805290122809955,
      "loss": 0.8546,
      "step": 1323
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001780163711592381,
      "loss": 0.7957,
      "step": 1324
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017797981446818028,
      "loss": 0.9106,
      "step": 1325
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017794323116740058,
      "loss": 0.9033,
      "step": 1326
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001779066212693827,
      "loss": 0.8916,
      "step": 1327
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017786998478661934,
      "loss": 0.8767,
      "step": 1328
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001778333217316122,
      "loss": 0.8312,
      "step": 1329
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017779663211687225,
      "loss": 0.8695,
      "step": 1330
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001777599159549193,
      "loss": 0.8354,
      "step": 1331
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001777231732582824,
      "loss": 0.8512,
      "step": 1332
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001776864040394995,
      "loss": 0.3417,
      "step": 1333
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017764960831111775,
      "loss": 0.8443,
      "step": 1334
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017761278608569324,
      "loss": 0.8657,
      "step": 1335
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017757593737579108,
      "loss": 0.8545,
      "step": 1336
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001775390621939856,
      "loss": 0.906,
      "step": 1337
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017750216055285992,
      "loss": 0.9145,
      "step": 1338
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017746523246500633,
      "loss": 0.8141,
      "step": 1339
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017742827794302613,
      "loss": 0.9179,
      "step": 1340
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017739129699952965,
      "loss": 0.8878,
      "step": 1341
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001773542896471362,
      "loss": 0.8431,
      "step": 1342
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001773172558984741,
      "loss": 0.9036,
      "step": 1343
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001772801957661807,
      "loss": 0.824,
      "step": 1344
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017724310926290238,
      "loss": 0.8922,
      "step": 1345
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017720599640129445,
      "loss": 0.8294,
      "step": 1346
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017716885719402126,
      "loss": 0.9065,
      "step": 1347
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017713169165375613,
      "loss": 0.8226,
      "step": 1348
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017709449979318138,
      "loss": 0.8223,
      "step": 1349
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017705728162498833,
      "loss": 0.8984,
      "step": 1350
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017702003716187726,
      "loss": 0.8342,
      "step": 1351
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017698276641655737,
      "loss": 0.8512,
      "step": 1352
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017694546940174693,
      "loss": 0.8906,
      "step": 1353
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017690814613017307,
      "loss": 0.8382,
      "step": 1354
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017687079661457195,
      "loss": 0.8826,
      "step": 1355
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017683342086768873,
      "loss": 0.8463,
      "step": 1356
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017679601890227735,
      "loss": 0.894,
      "step": 1357
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00017675859073110087,
      "loss": 0.8605,
      "step": 1358
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001767211363669312,
      "loss": 0.8991,
      "step": 1359
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017668365582254922,
      "loss": 0.832,
      "step": 1360
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017664614911074475,
      "loss": 0.8869,
      "step": 1361
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017660861624431654,
      "loss": 0.8898,
      "step": 1362
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017657105723607218,
      "loss": 0.9004,
      "step": 1363
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017653347209882836,
      "loss": 0.8937,
      "step": 1364
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001764958608454105,
      "loss": 0.8617,
      "step": 1365
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017645822348865304,
      "loss": 0.8882,
      "step": 1366
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001764205600413993,
      "loss": 0.8928,
      "step": 1367
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017638287051650146,
      "loss": 0.8889,
      "step": 1368
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017634515492682073,
      "loss": 0.8346,
      "step": 1369
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017630741328522704,
      "loss": 0.8562,
      "step": 1370
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017626964560459932,
      "loss": 0.8776,
      "step": 1371
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001762318518978254,
      "loss": 0.8804,
      "step": 1372
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017619403217780186,
      "loss": 0.8403,
      "step": 1373
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017615618645743436,
      "loss": 0.8536,
      "step": 1374
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017611831474963723,
      "loss": 0.8587,
      "step": 1375
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001760804170673338,
      "loss": 0.8546,
      "step": 1376
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017604249342345623,
      "loss": 0.8565,
      "step": 1377
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017600454383094548,
      "loss": 0.8719,
      "step": 1378
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017596656830275146,
      "loss": 0.8291,
      "step": 1379
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001759285668518329,
      "loss": 0.8522,
      "step": 1380
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001758905394911573,
      "loss": 0.855,
      "step": 1381
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017585248623370114,
      "loss": 0.8515,
      "step": 1382
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017581440709244962,
      "loss": 0.8785,
      "step": 1383
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017577630208039677,
      "loss": 0.3427,
      "step": 1384
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017573817121054557,
      "loss": 0.8802,
      "step": 1385
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001757000144959077,
      "loss": 0.8181,
      "step": 1386
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001756618319495037,
      "loss": 0.8329,
      "step": 1387
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017562362358436296,
      "loss": 0.3566,
      "step": 1388
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017558538941352358,
      "loss": 0.8607,
      "step": 1389
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017554712945003263,
      "loss": 0.8655,
      "step": 1390
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017550884370694583,
      "loss": 0.9268,
      "step": 1391
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017547053219732774,
      "loss": 0.8036,
      "step": 1392
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017543219493425176,
      "loss": 0.8848,
      "step": 1393
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017539383193080003,
      "loss": 0.914,
      "step": 1394
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017535544320006349,
      "loss": 0.8225,
      "step": 1395
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017531702875514188,
      "loss": 0.8639,
      "step": 1396
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017527858860914362,
      "loss": 0.8128,
      "step": 1397
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000175240122775186,
      "loss": 0.8259,
      "step": 1398
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001752016312663951,
      "loss": 0.9148,
      "step": 1399
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017516311409590567,
      "loss": 0.8743,
      "step": 1400
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017512457127686126,
      "loss": 0.8067,
      "step": 1401
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017508600282241415,
      "loss": 0.8753,
      "step": 1402
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017504740874572537,
      "loss": 0.3556,
      "step": 1403
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001750087890599648,
      "loss": 0.8943,
      "step": 1404
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017497014377831086,
      "loss": 0.9124,
      "step": 1405
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017493147291395083,
      "loss": 0.8786,
      "step": 1406
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017489277648008079,
      "loss": 0.8695,
      "step": 1407
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017485405448990532,
      "loss": 0.8722,
      "step": 1408
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000174815306956638,
      "loss": 0.8817,
      "step": 1409
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017477653389350088,
      "loss": 0.8834,
      "step": 1410
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00017473773531372487,
      "loss": 0.9007,
      "step": 1411
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001746989112305495,
      "loss": 0.8615,
      "step": 1412
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001746600616572231,
      "loss": 0.3392,
      "step": 1413
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017462118660700263,
      "loss": 0.8345,
      "step": 1414
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017458228609315373,
      "loss": 0.8706,
      "step": 1415
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017454336012895076,
      "loss": 0.913,
      "step": 1416
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001745044087276768,
      "loss": 0.3464,
      "step": 1417
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017446543190262353,
      "loss": 0.893,
      "step": 1418
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017442642966709136,
      "loss": 0.8624,
      "step": 1419
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001743874020343894,
      "loss": 0.766,
      "step": 1420
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001743483490178353,
      "loss": 0.8625,
      "step": 1421
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001743092706307555,
      "loss": 0.8305,
      "step": 1422
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017427016688648507,
      "loss": 0.8899,
      "step": 1423
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001742310377983677,
      "loss": 0.895,
      "step": 1424
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017419188337975577,
      "loss": 0.8276,
      "step": 1425
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017415270364401023,
      "loss": 0.3139,
      "step": 1426
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017411349860450072,
      "loss": 0.8597,
      "step": 1427
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017407426827460557,
      "loss": 0.8344,
      "step": 1428
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017403501266771162,
      "loss": 0.8432,
      "step": 1429
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017399573179721446,
      "loss": 0.8647,
      "step": 1430
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017395642567651816,
      "loss": 0.8758,
      "step": 1431
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017391709431903557,
      "loss": 0.8774,
      "step": 1432
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017387773773818802,
      "loss": 0.8837,
      "step": 1433
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001738383559474055,
      "loss": 0.8361,
      "step": 1434
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017379894896012656,
      "loss": 0.8319,
      "step": 1435
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017375951678979846,
      "loss": 0.862,
      "step": 1436
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017372005944987692,
      "loss": 0.8296,
      "step": 1437
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017368057695382633,
      "loss": 0.878,
      "step": 1438
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017364106931511963,
      "loss": 0.8825,
      "step": 1439
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017360153654723837,
      "loss": 0.844,
      "step": 1440
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017356197866367262,
      "loss": 0.8506,
      "step": 1441
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017352239567792116,
      "loss": 0.8655,
      "step": 1442
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017348278760349106,
      "loss": 0.9022,
      "step": 1443
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017344315445389828,
      "loss": 0.8796,
      "step": 1444
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017340349624266708,
      "loss": 0.3626,
      "step": 1445
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001733638129833304,
      "loss": 0.9272,
      "step": 1446
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017332410468942974,
      "loss": 0.9106,
      "step": 1447
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017328437137451505,
      "loss": 0.8045,
      "step": 1448
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017324461305214488,
      "loss": 0.9113,
      "step": 1449
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017320482973588635,
      "loss": 0.8227,
      "step": 1450
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000173165021439315,
      "loss": 0.8759,
      "step": 1451
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000173125188176015,
      "loss": 0.8803,
      "step": 1452
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000173085329959579,
      "loss": 0.8565,
      "step": 1453
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001730454468036081,
      "loss": 0.8469,
      "step": 1454
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017300553872171202,
      "loss": 0.8376,
      "step": 1455
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017296560572750892,
      "loss": 0.8464,
      "step": 1456
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001729256478346255,
      "loss": 0.9038,
      "step": 1457
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017288566505669698,
      "loss": 0.7957,
      "step": 1458
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017284565740736688,
      "loss": 0.8167,
      "step": 1459
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017280562490028752,
      "loss": 0.8233,
      "step": 1460
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001727655675491194,
      "loss": 0.7831,
      "step": 1461
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017272548536753174,
      "loss": 0.8606,
      "step": 1462
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017268537836920208,
      "loss": 0.8326,
      "step": 1463
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017264524656781646,
      "loss": 0.9301,
      "step": 1464
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017260508997706943,
      "loss": 0.8627,
      "step": 1465
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017256490861066394,
      "loss": 0.8356,
      "step": 1466
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017252470248231146,
      "loss": 0.8953,
      "step": 1467
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017248447160573185,
      "loss": 0.3255,
      "step": 1468
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00017244421599465343,
      "loss": 0.8776,
      "step": 1469
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000172403935662813,
      "loss": 0.9173,
      "step": 1470
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017236363062395567,
      "loss": 0.9164,
      "step": 1471
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017232330089183523,
      "loss": 0.9029,
      "step": 1472
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017228294648021362,
      "loss": 0.82,
      "step": 1473
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017224256740286134,
      "loss": 0.8723,
      "step": 1474
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017220216367355734,
      "loss": 0.8948,
      "step": 1475
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017216173530608888,
      "loss": 0.8733,
      "step": 1476
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017212128231425172,
      "loss": 0.8979,
      "step": 1477
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017208080471184994,
      "loss": 0.8692,
      "step": 1478
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017204030251269609,
      "loss": 0.8507,
      "step": 1479
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017199977573061106,
      "loss": 0.3308,
      "step": 1480
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017195922437942415,
      "loss": 0.877,
      "step": 1481
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017191864847297306,
      "loss": 0.8419,
      "step": 1482
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017187804802510392,
      "loss": 0.8709,
      "step": 1483
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017183742304967105,
      "loss": 0.8814,
      "step": 1484
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017179677356053734,
      "loss": 0.8962,
      "step": 1485
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001717560995715739,
      "loss": 0.8555,
      "step": 1486
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017171540109666035,
      "loss": 0.8852,
      "step": 1487
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017167467814968452,
      "loss": 0.8268,
      "step": 1488
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017163393074454266,
      "loss": 0.9106,
      "step": 1489
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017159315889513934,
      "loss": 0.8237,
      "step": 1490
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017155236261538753,
      "loss": 0.8436,
      "step": 1491
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001715115419192085,
      "loss": 0.903,
      "step": 1492
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001714706968205318,
      "loss": 0.8698,
      "step": 1493
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017142982733329537,
      "loss": 0.8663,
      "step": 1494
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017138893347144545,
      "loss": 0.8353,
      "step": 1495
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017134801524893665,
      "loss": 0.8854,
      "step": 1496
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017130707267973177,
      "loss": 0.86,
      "step": 1497
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001712661057778021,
      "loss": 0.8078,
      "step": 1498
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000171225114557127,
      "loss": 0.8567,
      "step": 1499
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017118409903169436,
      "loss": 0.8439,
      "step": 1500
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001711430592155002,
      "loss": 0.8202,
      "step": 1501
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017110199512254893,
      "loss": 0.8371,
      "step": 1502
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017106090676685315,
      "loss": 0.8661,
      "step": 1503
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017101979416243385,
      "loss": 0.8316,
      "step": 1504
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017097865732332022,
      "loss": 0.8424,
      "step": 1505
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017093749626354972,
      "loss": 0.8486,
      "step": 1506
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001708963109971681,
      "loss": 0.8786,
      "step": 1507
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017085510153822938,
      "loss": 0.8443,
      "step": 1508
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017081386790079574,
      "loss": 0.8248,
      "step": 1509
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001707726100989378,
      "loss": 0.3633,
      "step": 1510
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017073132814673426,
      "loss": 0.8142,
      "step": 1511
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017069002205827213,
      "loss": 0.3336,
      "step": 1512
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001706486918476466,
      "loss": 0.9683,
      "step": 1513
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001706073375289612,
      "loss": 0.8659,
      "step": 1514
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017056595911632756,
      "loss": 0.8032,
      "step": 1515
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017052455662386562,
      "loss": 0.8666,
      "step": 1516
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001704831300657035,
      "loss": 0.8583,
      "step": 1517
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001704416794559775,
      "loss": 0.9067,
      "step": 1518
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017040020480883227,
      "loss": 0.3405,
      "step": 1519
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017035870613842045,
      "loss": 0.8413,
      "step": 1520
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017031718345890306,
      "loss": 0.8748,
      "step": 1521
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017027563678444923,
      "loss": 0.8287,
      "step": 1522
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00017023406612923627,
      "loss": 0.8311,
      "step": 1523
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001701924715074497,
      "loss": 0.8588,
      "step": 1524
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00017015085293328318,
      "loss": 0.8744,
      "step": 1525
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00017010921042093867,
      "loss": 0.8767,
      "step": 1526
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00017006754398462605,
      "loss": 0.838,
      "step": 1527
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00017002585363856364,
      "loss": 0.8515,
      "step": 1528
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016998413939697774,
      "loss": 0.8782,
      "step": 1529
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001699424012741029,
      "loss": 0.9223,
      "step": 1530
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016990063928418173,
      "loss": 0.8506,
      "step": 1531
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016985885344146507,
      "loss": 0.8377,
      "step": 1532
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016981704376021185,
      "loss": 0.8418,
      "step": 1533
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016977521025468909,
      "loss": 0.9222,
      "step": 1534
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001697333529391721,
      "loss": 0.866,
      "step": 1535
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016969147182794415,
      "loss": 0.8551,
      "step": 1536
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016964956693529668,
      "loss": 0.921,
      "step": 1537
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001696076382755293,
      "loss": 0.8339,
      "step": 1538
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016956568586294965,
      "loss": 0.8778,
      "step": 1539
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016952370971187354,
      "loss": 0.8703,
      "step": 1540
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016948170983662484,
      "loss": 0.8243,
      "step": 1541
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001694396862515355,
      "loss": 0.8531,
      "step": 1542
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016939763897094566,
      "loss": 0.8703,
      "step": 1543
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016935556800920341,
      "loss": 0.848,
      "step": 1544
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000169313473380665,
      "loss": 0.8371,
      "step": 1545
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016927135509969477,
      "loss": 0.882,
      "step": 1546
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016922921318066506,
      "loss": 0.8737,
      "step": 1547
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016918704763795636,
      "loss": 0.8327,
      "step": 1548
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016914485848595713,
      "loss": 0.3227,
      "step": 1549
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016910264573906397,
      "loss": 0.3378,
      "step": 1550
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016906040941168152,
      "loss": 0.8466,
      "step": 1551
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001690181495182224,
      "loss": 0.8669,
      "step": 1552
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016897586607310733,
      "loss": 0.8522,
      "step": 1553
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016893355909076508,
      "loss": 0.8354,
      "step": 1554
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016889122858563242,
      "loss": 0.834,
      "step": 1555
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001688488745721541,
      "loss": 0.8586,
      "step": 1556
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000168806497064783,
      "loss": 0.8971,
      "step": 1557
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016876409607797995,
      "loss": 0.8399,
      "step": 1558
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016872167162621375,
      "loss": 0.9424,
      "step": 1559
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016867922372396134,
      "loss": 0.9092,
      "step": 1560
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016863675238570753,
      "loss": 0.8064,
      "step": 1561
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001685942576259452,
      "loss": 0.7882,
      "step": 1562
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016855173945917523,
      "loss": 0.8683,
      "step": 1563
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016850919789990635,
      "loss": 0.9151,
      "step": 1564
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016846663296265555,
      "loss": 0.887,
      "step": 1565
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016842404466194748,
      "loss": 0.8611,
      "step": 1566
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016838143301231497,
      "loss": 0.8141,
      "step": 1567
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016833879802829877,
      "loss": 0.854,
      "step": 1568
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016829613972444758,
      "loss": 0.8753,
      "step": 1569
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016825345811531802,
      "loss": 0.8456,
      "step": 1570
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016821075321547476,
      "loss": 0.8187,
      "step": 1571
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016816802503949032,
      "loss": 0.8811,
      "step": 1572
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001681252736019452,
      "loss": 0.8747,
      "step": 1573
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001680824989174279,
      "loss": 0.8111,
      "step": 1574
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001680397010005347,
      "loss": 0.7925,
      "step": 1575
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016799687986586995,
      "loss": 0.8374,
      "step": 1576
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001679540355280459,
      "loss": 0.8947,
      "step": 1577
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016791116800168266,
      "loss": 0.8692,
      "step": 1578
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00016786827730140826,
      "loss": 0.7845,
      "step": 1579
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001678253634418587,
      "loss": 0.8908,
      "step": 1580
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016778242643767784,
      "loss": 0.9041,
      "step": 1581
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016773946630351745,
      "loss": 0.7735,
      "step": 1582
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016769648305403715,
      "loss": 0.9169,
      "step": 1583
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001676534767039045,
      "loss": 0.887,
      "step": 1584
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016761044726779493,
      "loss": 0.8534,
      "step": 1585
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016756739476039175,
      "loss": 0.8951,
      "step": 1586
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016752431919638612,
      "loss": 0.8764,
      "step": 1587
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016748122059047715,
      "loss": 0.8763,
      "step": 1588
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016743809895737165,
      "loss": 0.8908,
      "step": 1589
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016739495431178445,
      "loss": 0.8498,
      "step": 1590
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016735178666843817,
      "loss": 0.8019,
      "step": 1591
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001673085960420632,
      "loss": 0.8073,
      "step": 1592
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001672653824473979,
      "loss": 0.908,
      "step": 1593
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016722214589918845,
      "loss": 0.9377,
      "step": 1594
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016717888641218883,
      "loss": 0.917,
      "step": 1595
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016713560400116072,
      "loss": 0.842,
      "step": 1596
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016709229868087393,
      "loss": 0.8196,
      "step": 1597
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016704897046610577,
      "loss": 0.9015,
      "step": 1598
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016700561937164156,
      "loss": 0.8942,
      "step": 1599
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016696224541227434,
      "loss": 0.8508,
      "step": 1600
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016691884860280503,
      "loss": 0.8857,
      "step": 1601
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016687542895804221,
      "loss": 0.8953,
      "step": 1602
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016683198649280242,
      "loss": 0.3329,
      "step": 1603
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016678852122190986,
      "loss": 0.8103,
      "step": 1604
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016674503316019656,
      "loss": 0.8638,
      "step": 1605
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016670152232250236,
      "loss": 0.8244,
      "step": 1606
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016665798872367478,
      "loss": 0.9107,
      "step": 1607
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016661443237856922,
      "loss": 0.8212,
      "step": 1608
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016657085330204876,
      "loss": 0.8843,
      "step": 1609
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016652725150898424,
      "loss": 0.8584,
      "step": 1610
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016648362701425433,
      "loss": 0.8511,
      "step": 1611
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016643997983274535,
      "loss": 0.8403,
      "step": 1612
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016639630997935138,
      "loss": 0.8903,
      "step": 1613
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016635261746897432,
      "loss": 0.8184,
      "step": 1614
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001663089023165237,
      "loss": 0.8433,
      "step": 1615
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016626516453691685,
      "loss": 0.9334,
      "step": 1616
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016622140414507874,
      "loss": 0.8695,
      "step": 1617
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016617762115594216,
      "loss": 0.8664,
      "step": 1618
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001661338155844475,
      "loss": 0.8239,
      "step": 1619
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016608998744554294,
      "loss": 0.8783,
      "step": 1620
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016604613675418435,
      "loss": 0.7964,
      "step": 1621
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016600226352533523,
      "loss": 0.8731,
      "step": 1622
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016595836777396684,
      "loss": 0.8442,
      "step": 1623
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016591444951505815,
      "loss": 0.8655,
      "step": 1624
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016587050876359571,
      "loss": 0.8527,
      "step": 1625
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001658265455345738,
      "loss": 0.8414,
      "step": 1626
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016578255984299437,
      "loss": 0.8838,
      "step": 1627
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016573855170386713,
      "loss": 0.8471,
      "step": 1628
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016569452113220917,
      "loss": 0.8749,
      "step": 1629
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016565046814304558,
      "loss": 0.9031,
      "step": 1630
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016560639275140892,
      "loss": 0.882,
      "step": 1631
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016556229497233932,
      "loss": 0.8803,
      "step": 1632
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016551817482088473,
      "loss": 0.9294,
      "step": 1633
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016547403231210062,
      "loss": 0.9196,
      "step": 1634
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00016542986746105013,
      "loss": 0.8527,
      "step": 1635
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016538568028280398,
      "loss": 0.8382,
      "step": 1636
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001653414707924406,
      "loss": 0.8263,
      "step": 1637
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001652972390050459,
      "loss": 0.8538,
      "step": 1638
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016525298493571354,
      "loss": 0.7862,
      "step": 1639
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016520870859954468,
      "loss": 0.8622,
      "step": 1640
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016516441001164812,
      "loss": 0.8766,
      "step": 1641
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001651200891871402,
      "loss": 0.8339,
      "step": 1642
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016507574614114495,
      "loss": 0.8891,
      "step": 1643
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016503138088879397,
      "loss": 0.8525,
      "step": 1644
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016498699344522633,
      "loss": 0.8197,
      "step": 1645
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016494258382558874,
      "loss": 0.91,
      "step": 1646
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016489815204503552,
      "loss": 0.8195,
      "step": 1647
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016485369811872845,
      "loss": 0.8111,
      "step": 1648
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016480922206183696,
      "loss": 0.7831,
      "step": 1649
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016476472388953802,
      "loss": 0.8455,
      "step": 1650
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016472020361701608,
      "loss": 0.7967,
      "step": 1651
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016467566125946318,
      "loss": 0.8609,
      "step": 1652
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016463109683207888,
      "loss": 0.854,
      "step": 1653
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016458651035007032,
      "loss": 0.9131,
      "step": 1654
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016454190182865216,
      "loss": 0.8624,
      "step": 1655
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016449727128304645,
      "loss": 0.9076,
      "step": 1656
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001644526187284829,
      "loss": 0.832,
      "step": 1657
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016440794418019875,
      "loss": 0.8657,
      "step": 1658
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016436324765343855,
      "loss": 0.8664,
      "step": 1659
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016431852916345459,
      "loss": 0.3101,
      "step": 1660
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001642737887255065,
      "loss": 0.8513,
      "step": 1661
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001642290263548614,
      "loss": 0.8485,
      "step": 1662
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016418424206679407,
      "loss": 0.7916,
      "step": 1663
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001641394358765865,
      "loss": 0.8896,
      "step": 1664
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016409460779952836,
      "loss": 0.3482,
      "step": 1665
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016404975785091667,
      "loss": 0.8215,
      "step": 1666
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016400488604605603,
      "loss": 0.8015,
      "step": 1667
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001639599924002584,
      "loss": 0.8294,
      "step": 1668
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016391507692884322,
      "loss": 0.8852,
      "step": 1669
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016387013964713737,
      "loss": 0.844,
      "step": 1670
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016382518057047522,
      "loss": 0.8668,
      "step": 1671
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001637801997141985,
      "loss": 0.8463,
      "step": 1672
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016373519709365648,
      "loss": 0.8138,
      "step": 1673
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001636901727242057,
      "loss": 0.3413,
      "step": 1674
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016364512662121028,
      "loss": 0.8446,
      "step": 1675
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016360005880004165,
      "loss": 0.8488,
      "step": 1676
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016355496927607875,
      "loss": 0.8501,
      "step": 1677
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016350985806470777,
      "loss": 0.8596,
      "step": 1678
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016346472518132248,
      "loss": 0.8797,
      "step": 1679
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016341957064132392,
      "loss": 0.8973,
      "step": 1680
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016337439446012058,
      "loss": 0.3452,
      "step": 1681
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016332919665312828,
      "loss": 0.7983,
      "step": 1682
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001632839772357703,
      "loss": 0.8564,
      "step": 1683
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016323873622347724,
      "loss": 0.8723,
      "step": 1684
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001631934736316871,
      "loss": 0.7954,
      "step": 1685
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016314818947584515,
      "loss": 0.8609,
      "step": 1686
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016310288377140413,
      "loss": 0.8346,
      "step": 1687
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016305755653382416,
      "loss": 0.8933,
      "step": 1688
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016301220777857254,
      "loss": 0.8217,
      "step": 1689
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016296683752112408,
      "loss": 0.8715,
      "step": 1690
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00016292144577696082,
      "loss": 0.8684,
      "step": 1691
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016287603256157223,
      "loss": 0.8384,
      "step": 1692
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016283059789045498,
      "loss": 0.8265,
      "step": 1693
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016278514177911317,
      "loss": 0.8535,
      "step": 1694
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016273966424305822,
      "loss": 0.8421,
      "step": 1695
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016269416529780876,
      "loss": 0.8548,
      "step": 1696
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016264864495889084,
      "loss": 0.8922,
      "step": 1697
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016260310324183767,
      "loss": 0.8596,
      "step": 1698
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001625575401621899,
      "loss": 0.7905,
      "step": 1699
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016251195573549544,
      "loss": 0.837,
      "step": 1700
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001624663499773094,
      "loss": 0.8408,
      "step": 1701
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016242072290319426,
      "loss": 0.7793,
      "step": 1702
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001623750745287197,
      "loss": 0.8478,
      "step": 1703
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016232940486946275,
      "loss": 0.9303,
      "step": 1704
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016228371394100765,
      "loss": 0.33,
      "step": 1705
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016223800175894587,
      "loss": 0.86,
      "step": 1706
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001621922683388762,
      "loss": 0.8487,
      "step": 1707
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001621465136964047,
      "loss": 0.3423,
      "step": 1708
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016210073784714453,
      "loss": 0.8003,
      "step": 1709
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016205494080671621,
      "loss": 0.9128,
      "step": 1710
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016200912259074746,
      "loss": 0.8772,
      "step": 1711
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016196328321487323,
      "loss": 0.7988,
      "step": 1712
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016191742269473566,
      "loss": 0.833,
      "step": 1713
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016187154104598417,
      "loss": 0.7971,
      "step": 1714
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016182563828427533,
      "loss": 0.8107,
      "step": 1715
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001617797144252729,
      "loss": 0.8782,
      "step": 1716
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016173376948464793,
      "loss": 0.8261,
      "step": 1717
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016168780347807857,
      "loss": 0.8724,
      "step": 1718
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001616418164212502,
      "loss": 0.864,
      "step": 1719
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001615958083298554,
      "loss": 0.8375,
      "step": 1720
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016154977921959386,
      "loss": 0.3278,
      "step": 1721
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016150372910617256,
      "loss": 0.8189,
      "step": 1722
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001614576580053055,
      "loss": 0.8439,
      "step": 1723
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016141156593271397,
      "loss": 0.8474,
      "step": 1724
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016136545290412632,
      "loss": 0.7886,
      "step": 1725
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016131931893527814,
      "loss": 0.8556,
      "step": 1726
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016127316404191208,
      "loss": 0.8555,
      "step": 1727
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000161226988239778,
      "loss": 0.8445,
      "step": 1728
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016118079154463287,
      "loss": 0.7991,
      "step": 1729
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016113457397224075,
      "loss": 0.8185,
      "step": 1730
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016108833553837287,
      "loss": 0.8757,
      "step": 1731
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001610420762588076,
      "loss": 0.8922,
      "step": 1732
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016099579614933037,
      "loss": 0.8869,
      "step": 1733
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016094949522573376,
      "loss": 0.8713,
      "step": 1734
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016090317350381738,
      "loss": 0.8864,
      "step": 1735
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016085683099938806,
      "loss": 0.842,
      "step": 1736
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016081046772825955,
      "loss": 0.8812,
      "step": 1737
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016076408370625287,
      "loss": 0.8813,
      "step": 1738
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016071767894919602,
      "loss": 0.8207,
      "step": 1739
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016067125347292408,
      "loss": 0.8613,
      "step": 1740
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016062480729327924,
      "loss": 0.3383,
      "step": 1741
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001605783404261107,
      "loss": 0.8636,
      "step": 1742
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001605318528872748,
      "loss": 0.8801,
      "step": 1743
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001604853446926348,
      "loss": 0.8429,
      "step": 1744
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016043881585806112,
      "loss": 0.8804,
      "step": 1745
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00016039226639943125,
      "loss": 0.8211,
      "step": 1746
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001603456963326296,
      "loss": 0.8869,
      "step": 1747
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00016029910567354766,
      "loss": 0.8276,
      "step": 1748
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00016025249443808406,
      "loss": 0.87,
      "step": 1749
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00016020586264214423,
      "loss": 0.8712,
      "step": 1750
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00016015921030164082,
      "loss": 0.8583,
      "step": 1751
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00016011253743249338,
      "loss": 0.8913,
      "step": 1752
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001600658440506285,
      "loss": 0.8218,
      "step": 1753
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00016001913017197974,
      "loss": 0.873,
      "step": 1754
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015997239581248772,
      "loss": 0.8717,
      "step": 1755
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001599256409881,
      "loss": 0.8949,
      "step": 1756
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001598788657147711,
      "loss": 0.8619,
      "step": 1757
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001598320700084625,
      "loss": 0.8467,
      "step": 1758
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015978525388514288,
      "loss": 0.8484,
      "step": 1759
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001597384173607875,
      "loss": 0.8699,
      "step": 1760
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015969156045137894,
      "loss": 0.9346,
      "step": 1761
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001596446831729065,
      "loss": 0.8691,
      "step": 1762
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015959778554136658,
      "loss": 0.3688,
      "step": 1763
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015955086757276239,
      "loss": 0.8769,
      "step": 1764
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015950392928310416,
      "loss": 0.8425,
      "step": 1765
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015945697068840911,
      "loss": 0.3328,
      "step": 1766
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001594099918047013,
      "loss": 0.8432,
      "step": 1767
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015936299264801172,
      "loss": 0.8725,
      "step": 1768
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001593159732343783,
      "loss": 0.7568,
      "step": 1769
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015926893357984584,
      "loss": 0.8444,
      "step": 1770
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015922187370046617,
      "loss": 0.8417,
      "step": 1771
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015917479361229786,
      "loss": 0.7837,
      "step": 1772
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015912769333140654,
      "loss": 0.8242,
      "step": 1773
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015908057287386453,
      "loss": 0.8523,
      "step": 1774
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015903343225575123,
      "loss": 0.882,
      "step": 1775
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015898627149315281,
      "loss": 0.8118,
      "step": 1776
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015893909060216233,
      "loss": 0.8777,
      "step": 1777
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015889188959887978,
      "loss": 0.8261,
      "step": 1778
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001588446684994119,
      "loss": 0.7948,
      "step": 1779
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001587974273198724,
      "loss": 0.8402,
      "step": 1780
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001587501660763818,
      "loss": 0.8948,
      "step": 1781
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015870288478506738,
      "loss": 0.8357,
      "step": 1782
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015865558346206347,
      "loss": 0.862,
      "step": 1783
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015860826212351096,
      "loss": 0.377,
      "step": 1784
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015856092078555783,
      "loss": 0.8834,
      "step": 1785
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015851355946435873,
      "loss": 0.8551,
      "step": 1786
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015846617817607517,
      "loss": 0.7849,
      "step": 1787
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015841877693687544,
      "loss": 0.8358,
      "step": 1788
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015837135576293477,
      "loss": 0.8777,
      "step": 1789
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000158323914670435,
      "loss": 0.8594,
      "step": 1790
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001582764536755649,
      "loss": 0.8374,
      "step": 1791
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015822897279452,
      "loss": 0.9013,
      "step": 1792
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015818147204350256,
      "loss": 0.8988,
      "step": 1793
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015813395143872172,
      "loss": 0.8469,
      "step": 1794
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015808641099639332,
      "loss": 0.8691,
      "step": 1795
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015803885073274003,
      "loss": 0.8226,
      "step": 1796
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015799127066399123,
      "loss": 0.9167,
      "step": 1797
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000157943670806383,
      "loss": 0.8312,
      "step": 1798
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015789605117615837,
      "loss": 0.8359,
      "step": 1799
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015784841178956692,
      "loss": 0.8964,
      "step": 1800
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015780075266286507,
      "loss": 0.3425,
      "step": 1801
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00015775307381231597,
      "loss": 0.8158,
      "step": 1802
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015770537525418944,
      "loss": 0.8978,
      "step": 1803
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001576576570047621,
      "loss": 0.8411,
      "step": 1804
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001576099190803172,
      "loss": 0.8101,
      "step": 1805
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001575621614971449,
      "loss": 0.3215,
      "step": 1806
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015751438427154176,
      "loss": 0.8524,
      "step": 1807
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015746658741981135,
      "loss": 0.8119,
      "step": 1808
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015741877095826374,
      "loss": 0.8533,
      "step": 1809
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015737093490321576,
      "loss": 0.8204,
      "step": 1810
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015732307927099085,
      "loss": 0.8467,
      "step": 1811
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015727520407791935,
      "loss": 0.7851,
      "step": 1812
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015722730934033798,
      "loss": 0.8332,
      "step": 1813
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015717939507459033,
      "loss": 0.8778,
      "step": 1814
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015713146129702662,
      "loss": 0.864,
      "step": 1815
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015708350802400372,
      "loss": 0.9077,
      "step": 1816
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015703553527188503,
      "loss": 0.8768,
      "step": 1817
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001569875430570408,
      "loss": 0.8508,
      "step": 1818
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001569395313958478,
      "loss": 0.8552,
      "step": 1819
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015689150030468945,
      "loss": 0.8538,
      "step": 1820
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001568434497999558,
      "loss": 0.8309,
      "step": 1821
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001567953798980436,
      "loss": 0.8136,
      "step": 1822
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001567472906153561,
      "loss": 0.8438,
      "step": 1823
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001566991819683032,
      "loss": 0.8047,
      "step": 1824
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001566510539733015,
      "loss": 0.8497,
      "step": 1825
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015660290664677406,
      "loss": 0.8165,
      "step": 1826
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015655474000515065,
      "loss": 0.8547,
      "step": 1827
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015650655406486755,
      "loss": 0.8702,
      "step": 1828
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015645834884236772,
      "loss": 0.889,
      "step": 1829
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015641012435410056,
      "loss": 0.8848,
      "step": 1830
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015636188061652218,
      "loss": 0.3536,
      "step": 1831
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015631361764609523,
      "loss": 0.8276,
      "step": 1832
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015626533545928884,
      "loss": 0.8574,
      "step": 1833
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015621703407257877,
      "loss": 0.8777,
      "step": 1834
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015616871350244735,
      "loss": 0.8682,
      "step": 1835
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001561203737653834,
      "loss": 0.8388,
      "step": 1836
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015607201487788226,
      "loss": 0.8221,
      "step": 1837
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015602363685644592,
      "loss": 0.8381,
      "step": 1838
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015597523971758284,
      "loss": 0.3644,
      "step": 1839
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015592682347780786,
      "loss": 0.7879,
      "step": 1840
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001558783881536426,
      "loss": 0.8682,
      "step": 1841
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015582993376161508,
      "loss": 0.9095,
      "step": 1842
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015578146031825968,
      "loss": 0.8142,
      "step": 1843
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015573296784011752,
      "loss": 0.8696,
      "step": 1844
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015568445634373603,
      "loss": 0.4016,
      "step": 1845
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015563592584566925,
      "loss": 0.8581,
      "step": 1846
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015558737636247764,
      "loss": 0.8176,
      "step": 1847
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001555388079107282,
      "loss": 0.833,
      "step": 1848
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001554902205069943,
      "loss": 0.8689,
      "step": 1849
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015544161416785585,
      "loss": 0.8716,
      "step": 1850
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015539298890989925,
      "loss": 0.7987,
      "step": 1851
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015534434474971724,
      "loss": 0.8778,
      "step": 1852
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015529568170390915,
      "loss": 0.8517,
      "step": 1853
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015524699978908068,
      "loss": 0.8296,
      "step": 1854
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015519829902184396,
      "loss": 0.8173,
      "step": 1855
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015514957941881757,
      "loss": 0.8101,
      "step": 1856
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00015510084099662654,
      "loss": 0.8698,
      "step": 1857
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001550520837719023,
      "loss": 0.7843,
      "step": 1858
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015500330776128265,
      "loss": 0.861,
      "step": 1859
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015495451298141189,
      "loss": 0.841,
      "step": 1860
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015490569944894066,
      "loss": 0.8558,
      "step": 1861
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015485686718052605,
      "loss": 0.8656,
      "step": 1862
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015480801619283149,
      "loss": 0.866,
      "step": 1863
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015475914650252683,
      "loss": 0.8831,
      "step": 1864
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001547102581262883,
      "loss": 0.8612,
      "step": 1865
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001546613510807985,
      "loss": 0.8758,
      "step": 1866
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015461242538274635,
      "loss": 0.835,
      "step": 1867
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001545634810488273,
      "loss": 0.8722,
      "step": 1868
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015451451809574292,
      "loss": 0.9146,
      "step": 1869
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015446553654020133,
      "loss": 0.8138,
      "step": 1870
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015441653639891692,
      "loss": 0.8852,
      "step": 1871
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015436751768861044,
      "loss": 0.8801,
      "step": 1872
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001543184804260089,
      "loss": 0.889,
      "step": 1873
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015426942462784578,
      "loss": 0.8547,
      "step": 1874
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001542203503108608,
      "loss": 0.8222,
      "step": 1875
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015417125749179998,
      "loss": 0.8509,
      "step": 1876
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001541221461874157,
      "loss": 0.8489,
      "step": 1877
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015407301641446664,
      "loss": 0.8796,
      "step": 1878
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015402386818971772,
      "loss": 0.8199,
      "step": 1879
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015397470152994029,
      "loss": 0.8629,
      "step": 1880
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001539255164519119,
      "loss": 0.3536,
      "step": 1881
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015387631297241635,
      "loss": 0.8052,
      "step": 1882
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015382709110824377,
      "loss": 0.8832,
      "step": 1883
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015377785087619062,
      "loss": 0.9278,
      "step": 1884
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015372859229305954,
      "loss": 0.826,
      "step": 1885
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015367931537565942,
      "loss": 0.887,
      "step": 1886
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001536300201408055,
      "loss": 0.81,
      "step": 1887
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015358070660531915,
      "loss": 0.8219,
      "step": 1888
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015353137478602808,
      "loss": 0.8316,
      "step": 1889
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015348202469976627,
      "loss": 0.3498,
      "step": 1890
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001534326563633738,
      "loss": 0.8641,
      "step": 1891
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015338326979369709,
      "loss": 0.8878,
      "step": 1892
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015333386500758871,
      "loss": 0.8687,
      "step": 1893
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001532844420219075,
      "loss": 0.8829,
      "step": 1894
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001532350008535185,
      "loss": 0.8517,
      "step": 1895
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001531855415192929,
      "loss": 0.8886,
      "step": 1896
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015313606403610824,
      "loss": 0.7818,
      "step": 1897
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015308656842084802,
      "loss": 0.8878,
      "step": 1898
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001530370546904021,
      "loss": 0.8839,
      "step": 1899
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001529875228616665,
      "loss": 0.7757,
      "step": 1900
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001529379729515434,
      "loss": 0.8797,
      "step": 1901
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001528884049769411,
      "loss": 0.8219,
      "step": 1902
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001528388189547741,
      "loss": 0.8361,
      "step": 1903
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001527892149019631,
      "loss": 0.8912,
      "step": 1904
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001527395928354349,
      "loss": 0.8697,
      "step": 1905
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015268995277212244,
      "loss": 0.7738,
      "step": 1906
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001526402947289649,
      "loss": 0.8693,
      "step": 1907
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015259061872290746,
      "loss": 0.8229,
      "step": 1908
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015254092477090148,
      "loss": 0.8447,
      "step": 1909
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015249121288990448,
      "loss": 0.8534,
      "step": 1910
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015244148309688009,
      "loss": 0.8509,
      "step": 1911
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000152391735408798,
      "loss": 0.894,
      "step": 1912
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00015234196984263406,
      "loss": 0.828,
      "step": 1913
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015229218641537017,
      "loss": 0.8151,
      "step": 1914
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001522423851439944,
      "loss": 0.8625,
      "step": 1915
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015219256604550086,
      "loss": 0.7952,
      "step": 1916
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015214272913688972,
      "loss": 0.854,
      "step": 1917
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015209287443516728,
      "loss": 0.8904,
      "step": 1918
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015204300195734586,
      "loss": 0.8406,
      "step": 1919
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015199311172044394,
      "loss": 0.8686,
      "step": 1920
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015194320374148593,
      "loss": 0.8435,
      "step": 1921
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015189327803750234,
      "loss": 0.803,
      "step": 1922
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015184333462552982,
      "loss": 0.8281,
      "step": 1923
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015179337352261096,
      "loss": 0.8676,
      "step": 1924
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001517433947457944,
      "loss": 0.8878,
      "step": 1925
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015169339831213482,
      "loss": 0.3411,
      "step": 1926
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015164338423869294,
      "loss": 0.8905,
      "step": 1927
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015159335254253555,
      "loss": 0.3342,
      "step": 1928
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001515433032407353,
      "loss": 0.8504,
      "step": 1929
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015149323635037102,
      "loss": 0.8262,
      "step": 1930
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015144315188852744,
      "loss": 0.7618,
      "step": 1931
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001513930498722953,
      "loss": 0.8472,
      "step": 1932
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015134293031877132,
      "loss": 0.9349,
      "step": 1933
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015129279324505826,
      "loss": 0.8208,
      "step": 1934
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001512426386682648,
      "loss": 0.8133,
      "step": 1935
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001511924666055057,
      "loss": 0.85,
      "step": 1936
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001511422770739015,
      "loss": 0.865,
      "step": 1937
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015109207009057884,
      "loss": 0.9118,
      "step": 1938
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015104184567267027,
      "loss": 0.8192,
      "step": 1939
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015099160383731436,
      "loss": 0.8984,
      "step": 1940
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015094134460165548,
      "loss": 0.823,
      "step": 1941
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015089106798284405,
      "loss": 0.9153,
      "step": 1942
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015084077399803638,
      "loss": 0.834,
      "step": 1943
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001507904626643948,
      "loss": 0.8985,
      "step": 1944
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015074013399908736,
      "loss": 0.8323,
      "step": 1945
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015068978801928822,
      "loss": 0.8285,
      "step": 1946
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015063942474217735,
      "loss": 0.8181,
      "step": 1947
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015058904418494064,
      "loss": 0.814,
      "step": 1948
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015053864636476987,
      "loss": 0.8472,
      "step": 1949
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015048823129886274,
      "loss": 0.3651,
      "step": 1950
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015043779900442283,
      "loss": 0.8656,
      "step": 1951
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015038734949865957,
      "loss": 0.7984,
      "step": 1952
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015033688279878825,
      "loss": 0.8265,
      "step": 1953
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015028639892203008,
      "loss": 0.892,
      "step": 1954
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001502358978856121,
      "loss": 0.8214,
      "step": 1955
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015018537970676722,
      "loss": 0.8309,
      "step": 1956
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001501348444027342,
      "loss": 0.8414,
      "step": 1957
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015008429199075764,
      "loss": 0.7839,
      "step": 1958
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00015003372248808793,
      "loss": 0.3133,
      "step": 1959
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001499831359119814,
      "loss": 0.8045,
      "step": 1960
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00014993253227970008,
      "loss": 0.8152,
      "step": 1961
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00014988191160851191,
      "loss": 0.8531,
      "step": 1962
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00014983127391569062,
      "loss": 0.8507,
      "step": 1963
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00014978061921851573,
      "loss": 0.8486,
      "step": 1964
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00014972994753427262,
      "loss": 0.8736,
      "step": 1965
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00014967925888025236,
      "loss": 0.9094,
      "step": 1966
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00014962855327375194,
      "loss": 0.9016,
      "step": 1967
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000149577830732074,
      "loss": 0.7728,
      "step": 1968
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014952709127252707,
      "loss": 0.8248,
      "step": 1969
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001494763349124254,
      "loss": 0.8575,
      "step": 1970
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014942556166908907,
      "loss": 0.8308,
      "step": 1971
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001493747715598438,
      "loss": 0.8323,
      "step": 1972
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014932396460202115,
      "loss": 0.8477,
      "step": 1973
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001492731408129584,
      "loss": 0.8537,
      "step": 1974
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014922230020999867,
      "loss": 0.8065,
      "step": 1975
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014917144281049064,
      "loss": 0.8071,
      "step": 1976
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014912056863178888,
      "loss": 0.7929,
      "step": 1977
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014906967769125358,
      "loss": 0.8921,
      "step": 1978
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001490187700062507,
      "loss": 0.8458,
      "step": 1979
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014896784559415192,
      "loss": 0.8244,
      "step": 1980
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014891690447233467,
      "loss": 0.8419,
      "step": 1981
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014886594665818192,
      "loss": 0.8296,
      "step": 1982
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001488149721690825,
      "loss": 0.8197,
      "step": 1983
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014876398102243088,
      "loss": 0.8604,
      "step": 1984
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014871297323562716,
      "loss": 0.8069,
      "step": 1985
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014866194882607722,
      "loss": 0.3083,
      "step": 1986
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001486109078111925,
      "loss": 0.8213,
      "step": 1987
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014855985020839022,
      "loss": 0.8204,
      "step": 1988
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001485087760350932,
      "loss": 0.796,
      "step": 1989
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014845768530872988,
      "loss": 0.8564,
      "step": 1990
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001484065780467344,
      "loss": 0.814,
      "step": 1991
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014835545426654655,
      "loss": 0.8939,
      "step": 1992
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014830431398561174,
      "loss": 0.8945,
      "step": 1993
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014825315722138098,
      "loss": 0.8222,
      "step": 1994
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014820198399131094,
      "loss": 0.8295,
      "step": 1995
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001481507943128639,
      "loss": 0.8556,
      "step": 1996
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014809958820350777,
      "loss": 0.8964,
      "step": 1997
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014804836568071603,
      "loss": 0.8811,
      "step": 1998
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014799712676196778,
      "loss": 0.7779,
      "step": 1999
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001479458714647478,
      "loss": 0.8366,
      "step": 2000
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014789459980654623,
      "loss": 0.8234,
      "step": 2001
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000147843311804859,
      "loss": 0.9439,
      "step": 2002
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001477920074771876,
      "loss": 0.8698,
      "step": 2003
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000147740686841039,
      "loss": 0.868,
      "step": 2004
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001476893499139258,
      "loss": 0.857,
      "step": 2005
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014763799671336613,
      "loss": 0.3447,
      "step": 2006
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001475866272568837,
      "loss": 0.8107,
      "step": 2007
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014753524156200772,
      "loss": 0.8488,
      "step": 2008
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014748383964627299,
      "loss": 0.847,
      "step": 2009
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014743242152721986,
      "loss": 0.8708,
      "step": 2010
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001473809872223941,
      "loss": 0.8299,
      "step": 2011
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014732953674934714,
      "loss": 0.8215,
      "step": 2012
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014727807012563586,
      "loss": 0.8275,
      "step": 2013
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014722658736882264,
      "loss": 0.8146,
      "step": 2014
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001471750884964754,
      "loss": 0.8169,
      "step": 2015
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014712357352616754,
      "loss": 0.8165,
      "step": 2016
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014707204247547793,
      "loss": 0.8281,
      "step": 2017
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000147020495361991,
      "loss": 0.8257,
      "step": 2018
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014696893220329655,
      "loss": 0.343,
      "step": 2019
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014691735301699,
      "loss": 0.3641,
      "step": 2020
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014686575782067208,
      "loss": 0.8498,
      "step": 2021
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014681414663194912,
      "loss": 0.8607,
      "step": 2022
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00014676251946843283,
      "loss": 0.8757,
      "step": 2023
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014671087634774037,
      "loss": 0.36,
      "step": 2024
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001466592172874944,
      "loss": 0.8519,
      "step": 2025
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014660754230532295,
      "loss": 0.9057,
      "step": 2026
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014655585141885955,
      "loss": 0.8426,
      "step": 2027
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001465041446457431,
      "loss": 0.8286,
      "step": 2028
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014645242200361796,
      "loss": 0.806,
      "step": 2029
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001464006835101339,
      "loss": 0.8224,
      "step": 2030
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014634892918294608,
      "loss": 0.8252,
      "step": 2031
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014629715903971503,
      "loss": 0.837,
      "step": 2032
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014624537309810683,
      "loss": 0.8822,
      "step": 2033
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014619357137579274,
      "loss": 0.8501,
      "step": 2034
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014614175389044952,
      "loss": 0.3398,
      "step": 2035
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014608992065975936,
      "loss": 0.8476,
      "step": 2036
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014603807170140968,
      "loss": 0.7859,
      "step": 2037
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001459862070330934,
      "loss": 0.8531,
      "step": 2038
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014593432667250872,
      "loss": 0.9,
      "step": 2039
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014588243063735925,
      "loss": 0.8696,
      "step": 2040
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001458305189453539,
      "loss": 0.811,
      "step": 2041
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001457785916142069,
      "loss": 0.3283,
      "step": 2042
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014572664866163793,
      "loss": 0.8683,
      "step": 2043
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001456746901053719,
      "loss": 0.7818,
      "step": 2044
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014562271596313905,
      "loss": 0.8005,
      "step": 2045
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000145570726252675,
      "loss": 0.8393,
      "step": 2046
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014551872099172064,
      "loss": 0.8225,
      "step": 2047
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014546670019802214,
      "loss": 0.8474,
      "step": 2048
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014541466388933108,
      "loss": 0.8257,
      "step": 2049
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014536261208340415,
      "loss": 0.8583,
      "step": 2050
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014531054479800353,
      "loss": 0.8198,
      "step": 2051
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001452584620508965,
      "loss": 0.8619,
      "step": 2052
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001452063638598558,
      "loss": 0.7648,
      "step": 2053
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014515425024265926,
      "loss": 0.7861,
      "step": 2054
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014510212121709008,
      "loss": 0.8555,
      "step": 2055
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014504997680093673,
      "loss": 0.8827,
      "step": 2056
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014499781701199284,
      "loss": 0.3505,
      "step": 2057
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014494564186805738,
      "loss": 0.7918,
      "step": 2058
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014489345138693453,
      "loss": 0.7806,
      "step": 2059
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001448412455864337,
      "loss": 0.7541,
      "step": 2060
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014478902448436946,
      "loss": 0.8571,
      "step": 2061
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014473678809856172,
      "loss": 0.8188,
      "step": 2062
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014468453644683556,
      "loss": 0.7977,
      "step": 2063
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014463226954702124,
      "loss": 0.8334,
      "step": 2064
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014457998741695423,
      "loss": 0.8853,
      "step": 2065
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014452769007447527,
      "loss": 0.7979,
      "step": 2066
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001444753775374301,
      "loss": 0.8686,
      "step": 2067
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014442304982366995,
      "loss": 0.8535,
      "step": 2068
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014437070695105092,
      "loss": 0.7835,
      "step": 2069
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001443183489374345,
      "loss": 0.8343,
      "step": 2070
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001442659758006872,
      "loss": 0.8306,
      "step": 2071
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014421358755868077,
      "loss": 0.7933,
      "step": 2072
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014416118422929216,
      "loss": 0.8523,
      "step": 2073
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014410876583040335,
      "loss": 0.8197,
      "step": 2074
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001440563323799015,
      "loss": 0.8185,
      "step": 2075
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000144003883895679,
      "loss": 0.8,
      "step": 2076
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00014395142039563323,
      "loss": 0.7996,
      "step": 2077
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001438989418976668,
      "loss": 0.8348,
      "step": 2078
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001438464484196874,
      "loss": 0.8257,
      "step": 2079
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014379393997960782,
      "loss": 0.8269,
      "step": 2080
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001437414165953459,
      "loss": 0.8583,
      "step": 2081
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014368887828482478,
      "loss": 0.897,
      "step": 2082
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014363632506597248,
      "loss": 0.8582,
      "step": 2083
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014358375695672218,
      "loss": 0.8251,
      "step": 2084
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014353117397501215,
      "loss": 0.8409,
      "step": 2085
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014347857613878574,
      "loss": 0.7942,
      "step": 2086
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001434259634659914,
      "loss": 0.8108,
      "step": 2087
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014337333597458254,
      "loss": 0.8414,
      "step": 2088
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014332069368251773,
      "loss": 0.8336,
      "step": 2089
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014326803660776055,
      "loss": 0.7775,
      "step": 2090
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001432153647682796,
      "loss": 0.8104,
      "step": 2091
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014316267818204853,
      "loss": 0.8808,
      "step": 2092
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014310997686704612,
      "loss": 0.8141,
      "step": 2093
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014305726084125605,
      "loss": 0.8602,
      "step": 2094
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014300453012266702,
      "loss": 0.8528,
      "step": 2095
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014295178472927282,
      "loss": 0.7884,
      "step": 2096
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014289902467907222,
      "loss": 0.8617,
      "step": 2097
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014284624999006898,
      "loss": 0.8479,
      "step": 2098
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014279346068027187,
      "loss": 0.8018,
      "step": 2099
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001427406567676946,
      "loss": 0.7918,
      "step": 2100
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014268783827035592,
      "loss": 0.8403,
      "step": 2101
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014263500520627956,
      "loss": 0.8452,
      "step": 2102
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014258215759349423,
      "loss": 0.8831,
      "step": 2103
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014252929545003351,
      "loss": 0.8637,
      "step": 2104
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000142476418793936,
      "loss": 0.8192,
      "step": 2105
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014242352764324531,
      "loss": 0.8334,
      "step": 2106
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014237062201600991,
      "loss": 0.8323,
      "step": 2107
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014231770193028325,
      "loss": 0.8466,
      "step": 2108
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014226476740412372,
      "loss": 0.863,
      "step": 2109
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014221181845559456,
      "loss": 0.8864,
      "step": 2110
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014215885510276405,
      "loss": 0.9082,
      "step": 2111
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014210587736370535,
      "loss": 0.8551,
      "step": 2112
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014205288525649646,
      "loss": 0.8975,
      "step": 2113
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014199987879922035,
      "loss": 0.8994,
      "step": 2114
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014194685800996486,
      "loss": 0.8363,
      "step": 2115
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014189382290682274,
      "loss": 0.8504,
      "step": 2116
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014184077350789162,
      "loss": 0.749,
      "step": 2117
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000141787709831274,
      "loss": 0.8524,
      "step": 2118
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014173463189507722,
      "loss": 0.9121,
      "step": 2119
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014168153971741356,
      "loss": 0.8344,
      "step": 2120
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001416284333164001,
      "loss": 0.892,
      "step": 2121
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014157531271015886,
      "loss": 0.8599,
      "step": 2122
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001415221779168165,
      "loss": 0.7843,
      "step": 2123
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014146902895450478,
      "loss": 0.8369,
      "step": 2124
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014141586584136013,
      "loss": 0.8999,
      "step": 2125
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014136268859552388,
      "loss": 0.8747,
      "step": 2126
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014130949723514214,
      "loss": 0.8391,
      "step": 2127
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014125629177836585,
      "loss": 0.3527,
      "step": 2128
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001412030722433508,
      "loss": 0.876,
      "step": 2129
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001411498386482575,
      "loss": 0.901,
      "step": 2130
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001410965910112513,
      "loss": 0.8501,
      "step": 2131
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014104332935050248,
      "loss": 0.8344,
      "step": 2132
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00014099005368418578,
      "loss": 0.8793,
      "step": 2133
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000140936764030481,
      "loss": 0.8475,
      "step": 2134
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014088346040757268,
      "loss": 0.3631,
      "step": 2135
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014083014283365,
      "loss": 0.8312,
      "step": 2136
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000140776811326907,
      "loss": 0.8517,
      "step": 2137
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014072346590554246,
      "loss": 0.3517,
      "step": 2138
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001406701065877599,
      "loss": 0.8186,
      "step": 2139
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014061673339176752,
      "loss": 0.8029,
      "step": 2140
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014056334633577838,
      "loss": 0.8097,
      "step": 2141
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001405099454380102,
      "loss": 0.8407,
      "step": 2142
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014045653071668537,
      "loss": 0.7838,
      "step": 2143
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014040310219003109,
      "loss": 0.8716,
      "step": 2144
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014034965987627925,
      "loss": 0.8327,
      "step": 2145
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001402962037936664,
      "loss": 0.8043,
      "step": 2146
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014024273396043383,
      "loss": 0.8338,
      "step": 2147
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014018925039482752,
      "loss": 0.8466,
      "step": 2148
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014013575311509814,
      "loss": 0.7894,
      "step": 2149
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00014008224213950097,
      "loss": 0.852,
      "step": 2150
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001400287174862961,
      "loss": 0.8584,
      "step": 2151
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013997517917374813,
      "loss": 0.8089,
      "step": 2152
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013992162722012644,
      "loss": 0.8185,
      "step": 2153
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000139868061643705,
      "loss": 0.8559,
      "step": 2154
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013981448246276252,
      "loss": 0.879,
      "step": 2155
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013976088969558218,
      "loss": 0.8062,
      "step": 2156
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013970728336045195,
      "loss": 0.8396,
      "step": 2157
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001396536634756644,
      "loss": 0.8236,
      "step": 2158
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001396000300595167,
      "loss": 0.8515,
      "step": 2159
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001395463831303106,
      "loss": 0.8859,
      "step": 2160
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001394927227063525,
      "loss": 0.8348,
      "step": 2161
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001394390488059535,
      "loss": 0.8453,
      "step": 2162
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001393853614474291,
      "loss": 0.8177,
      "step": 2163
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013933166064909952,
      "loss": 0.834,
      "step": 2164
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001392779464292896,
      "loss": 0.8698,
      "step": 2165
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013922421880632868,
      "loss": 0.8413,
      "step": 2166
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013917047779855065,
      "loss": 0.829,
      "step": 2167
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001391167234242941,
      "loss": 0.7848,
      "step": 2168
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013906295570190202,
      "loss": 0.8535,
      "step": 2169
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013900917464972206,
      "loss": 0.8429,
      "step": 2170
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013895538028610638,
      "loss": 0.8266,
      "step": 2171
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013890157262941175,
      "loss": 0.8748,
      "step": 2172
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013884775169799933,
      "loss": 0.8737,
      "step": 2173
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013879391751023495,
      "loss": 0.8343,
      "step": 2174
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013874007008448893,
      "loss": 0.8274,
      "step": 2175
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013868620943913604,
      "loss": 0.8257,
      "step": 2176
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013863233559255564,
      "loss": 0.8346,
      "step": 2177
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013857844856313159,
      "loss": 0.8628,
      "step": 2178
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013852454836925214,
      "loss": 0.8315,
      "step": 2179
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001384706350293102,
      "loss": 0.8047,
      "step": 2180
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013841670856170303,
      "loss": 0.7989,
      "step": 2181
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013836276898483246,
      "loss": 0.8111,
      "step": 2182
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013830881631710475,
      "loss": 0.8211,
      "step": 2183
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001382548505769306,
      "loss": 0.8297,
      "step": 2184
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013820087178272524,
      "loss": 0.8404,
      "step": 2185
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001381468799529083,
      "loss": 0.8499,
      "step": 2186
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013809287510590384,
      "loss": 0.8388,
      "step": 2187
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001380388572601405,
      "loss": 0.8136,
      "step": 2188
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013798482643405114,
      "loss": 0.9104,
      "step": 2189
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00013793078264607325,
      "loss": 0.7864,
      "step": 2190
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001378767259146486,
      "loss": 0.8025,
      "step": 2191
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013782265625822345,
      "loss": 0.8705,
      "step": 2192
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013776857369524842,
      "loss": 0.83,
      "step": 2193
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013771447824417865,
      "loss": 0.8184,
      "step": 2194
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013766036992347353,
      "loss": 0.8688,
      "step": 2195
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001376062487515969,
      "loss": 0.8579,
      "step": 2196
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013755211474701702,
      "loss": 0.8599,
      "step": 2197
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013749796792820654,
      "loss": 0.8333,
      "step": 2198
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013744380831364237,
      "loss": 0.8156,
      "step": 2199
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001373896359218059,
      "loss": 0.8182,
      "step": 2200
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013733545077118282,
      "loss": 0.8667,
      "step": 2201
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001372812528802632,
      "loss": 0.847,
      "step": 2202
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013722704226754148,
      "loss": 0.8525,
      "step": 2203
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001371728189515164,
      "loss": 0.8891,
      "step": 2204
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000137118582950691,
      "loss": 0.9026,
      "step": 2205
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013706433428357277,
      "loss": 0.8307,
      "step": 2206
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013701007296867339,
      "loss": 0.9273,
      "step": 2207
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013695579902450896,
      "loss": 0.8186,
      "step": 2208
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001369015124695998,
      "loss": 0.7951,
      "step": 2209
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001368472133224706,
      "loss": 0.3362,
      "step": 2210
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013679290160165038,
      "loss": 0.8853,
      "step": 2211
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013673857732567229,
      "loss": 0.3742,
      "step": 2212
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001366842405130739,
      "loss": 0.8808,
      "step": 2213
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013662989118239707,
      "loss": 0.8896,
      "step": 2214
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001365755293521879,
      "loss": 0.8255,
      "step": 2215
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013652115504099666,
      "loss": 0.8774,
      "step": 2216
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000136466768267378,
      "loss": 0.7937,
      "step": 2217
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013641236904989086,
      "loss": 0.3472,
      "step": 2218
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013635795740709822,
      "loss": 0.8455,
      "step": 2219
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013630353335756753,
      "loss": 0.8106,
      "step": 2220
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013624909691987037,
      "loss": 0.856,
      "step": 2221
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001361946481125825,
      "loss": 0.8811,
      "step": 2222
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000136140186954284,
      "loss": 0.822,
      "step": 2223
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013608571346355913,
      "loss": 0.8479,
      "step": 2224
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001360312276589963,
      "loss": 0.8619,
      "step": 2225
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013597672955918818,
      "loss": 0.844,
      "step": 2226
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013592221918273163,
      "loss": 0.8693,
      "step": 2227
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013586769654822773,
      "loss": 0.8872,
      "step": 2228
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013581316167428166,
      "loss": 0.8115,
      "step": 2229
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013575861457950283,
      "loss": 0.8298,
      "step": 2230
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013570405528250484,
      "loss": 0.8376,
      "step": 2231
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013564948380190539,
      "loss": 0.8988,
      "step": 2232
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001355949001563264,
      "loss": 0.8443,
      "step": 2233
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013554030436439392,
      "loss": 0.86,
      "step": 2234
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013548569644473814,
      "loss": 0.8057,
      "step": 2235
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013543107641599336,
      "loss": 0.8522,
      "step": 2236
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001353764442967981,
      "loss": 0.8241,
      "step": 2237
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001353218001057949,
      "loss": 0.9104,
      "step": 2238
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013526714386163046,
      "loss": 0.3088,
      "step": 2239
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013521247558295562,
      "loss": 0.8637,
      "step": 2240
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013515779528842532,
      "loss": 0.3253,
      "step": 2241
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013510310299669854,
      "loss": 0.8426,
      "step": 2242
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013504839872643848,
      "loss": 0.8205,
      "step": 2243
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00013499368249631226,
      "loss": 0.7914,
      "step": 2244
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001349389543249912,
      "loss": 0.837,
      "step": 2245
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001348842142311507,
      "loss": 0.8977,
      "step": 2246
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013482946223347013,
      "loss": 0.8108,
      "step": 2247
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013477469835063307,
      "loss": 0.8162,
      "step": 2248
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013471992260132698,
      "loss": 0.8756,
      "step": 2249
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001346651350042435,
      "loss": 0.8683,
      "step": 2250
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001346103355780783,
      "loss": 0.8447,
      "step": 2251
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013455552434153098,
      "loss": 0.8803,
      "step": 2252
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013450070131330532,
      "loss": 0.8139,
      "step": 2253
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013444586651210905,
      "loss": 0.8296,
      "step": 2254
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013439101995665388,
      "loss": 0.8194,
      "step": 2255
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001343361616656556,
      "loss": 0.3627,
      "step": 2256
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013428129165783394,
      "loss": 0.8575,
      "step": 2257
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013422640995191273,
      "loss": 0.8313,
      "step": 2258
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013417151656661968,
      "loss": 0.3615,
      "step": 2259
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013411661152068653,
      "loss": 0.8649,
      "step": 2260
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013406169483284902,
      "loss": 0.8165,
      "step": 2261
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013400676652184683,
      "loss": 0.8476,
      "step": 2262
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013395182660642363,
      "loss": 0.8503,
      "step": 2263
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013389687510532704,
      "loss": 0.8316,
      "step": 2264
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013384191203730857,
      "loss": 0.8517,
      "step": 2265
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013378693742112385,
      "loss": 0.8329,
      "step": 2266
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013373195127553225,
      "loss": 0.826,
      "step": 2267
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013367695361929722,
      "loss": 0.8435,
      "step": 2268
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013362194447118605,
      "loss": 0.9105,
      "step": 2269
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013356692384997,
      "loss": 0.8578,
      "step": 2270
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013351189177442425,
      "loss": 0.828,
      "step": 2271
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001334568482633278,
      "loss": 0.7994,
      "step": 2272
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013340179333546368,
      "loss": 0.8668,
      "step": 2273
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013334672700961875,
      "loss": 0.8535,
      "step": 2274
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013329164930458374,
      "loss": 0.852,
      "step": 2275
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013323656023915328,
      "loss": 0.8116,
      "step": 2276
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013318145983212595,
      "loss": 0.8298,
      "step": 2277
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001331263481023041,
      "loss": 0.8511,
      "step": 2278
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001330712250684939,
      "loss": 0.8598,
      "step": 2279
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013301609074950558,
      "loss": 0.7791,
      "step": 2280
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013296094516415305,
      "loss": 0.3255,
      "step": 2281
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001329057883312541,
      "loss": 0.8328,
      "step": 2282
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013285062026963037,
      "loss": 0.8107,
      "step": 2283
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013279544099810734,
      "loss": 0.8342,
      "step": 2284
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001327402505355143,
      "loss": 0.8433,
      "step": 2285
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013268504890068437,
      "loss": 0.8045,
      "step": 2286
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001326298361124545,
      "loss": 0.7912,
      "step": 2287
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013257461218966536,
      "loss": 0.831,
      "step": 2288
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013251937715116156,
      "loss": 0.8977,
      "step": 2289
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013246413101579137,
      "loss": 0.8376,
      "step": 2290
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013240887380240693,
      "loss": 0.8991,
      "step": 2291
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013235360552986414,
      "loss": 0.8077,
      "step": 2292
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013229832621702265,
      "loss": 0.862,
      "step": 2293
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013224303588274593,
      "loss": 0.8804,
      "step": 2294
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001321877345459011,
      "loss": 0.8632,
      "step": 2295
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013213242222535914,
      "loss": 0.8308,
      "step": 2296
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013207709893999482,
      "loss": 0.8122,
      "step": 2297
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001320217647086865,
      "loss": 0.8173,
      "step": 2298
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013196641955031638,
      "loss": 0.8069,
      "step": 2299
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00013191106348377037,
      "loss": 0.7788,
      "step": 2300
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001318556965279381,
      "loss": 0.8387,
      "step": 2301
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013180031870171287,
      "loss": 0.8485,
      "step": 2302
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013174493002399178,
      "loss": 0.7609,
      "step": 2303
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001316895305136756,
      "loss": 0.8413,
      "step": 2304
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001316341201896687,
      "loss": 0.8218,
      "step": 2305
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013157869907087926,
      "loss": 0.8204,
      "step": 2306
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013152326717621917,
      "loss": 0.9039,
      "step": 2307
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013146782452460383,
      "loss": 0.7882,
      "step": 2308
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013141237113495244,
      "loss": 0.8227,
      "step": 2309
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013135690702618792,
      "loss": 0.8791,
      "step": 2310
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013130143221723662,
      "loss": 0.846,
      "step": 2311
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001312459467270288,
      "loss": 0.8499,
      "step": 2312
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013119045057449823,
      "loss": 0.8064,
      "step": 2313
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013113494377858228,
      "loss": 0.8599,
      "step": 2314
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013107942635822207,
      "loss": 0.8438,
      "step": 2315
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013102389833236227,
      "loss": 0.8858,
      "step": 2316
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001309683597199512,
      "loss": 0.8323,
      "step": 2317
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013091281053994076,
      "loss": 0.8734,
      "step": 2318
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013085725081128646,
      "loss": 0.909,
      "step": 2319
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013080168055294746,
      "loss": 0.839,
      "step": 2320
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013074609978388642,
      "loss": 0.814,
      "step": 2321
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013069050852306967,
      "loss": 0.8503,
      "step": 2322
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013063490678946713,
      "loss": 0.3343,
      "step": 2323
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013057929460205225,
      "loss": 0.8015,
      "step": 2324
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000130523671979802,
      "loss": 0.8736,
      "step": 2325
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013046803894169702,
      "loss": 0.8093,
      "step": 2326
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013041239550672142,
      "loss": 0.8799,
      "step": 2327
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013035674169386286,
      "loss": 0.7987,
      "step": 2328
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013030107752211266,
      "loss": 0.8444,
      "step": 2329
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013024540301046552,
      "loss": 0.8544,
      "step": 2330
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001301897181779197,
      "loss": 0.811,
      "step": 2331
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013013402304347705,
      "loss": 0.8352,
      "step": 2332
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013007831762614294,
      "loss": 0.8242,
      "step": 2333
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00013002260194492618,
      "loss": 0.8861,
      "step": 2334
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012996687601883906,
      "loss": 0.7698,
      "step": 2335
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012991113986689744,
      "loss": 0.8457,
      "step": 2336
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012985539350812074,
      "loss": 0.8296,
      "step": 2337
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001297996369615316,
      "loss": 0.8578,
      "step": 2338
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012974387024615645,
      "loss": 0.3477,
      "step": 2339
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012968809338102498,
      "loss": 0.8958,
      "step": 2340
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012963230638517043,
      "loss": 0.8517,
      "step": 2341
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012957650927762946,
      "loss": 0.8669,
      "step": 2342
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001295207020774422,
      "loss": 0.8743,
      "step": 2343
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012946488480365223,
      "loss": 0.3732,
      "step": 2344
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001294090574753065,
      "loss": 0.8212,
      "step": 2345
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012935322011145556,
      "loss": 0.8517,
      "step": 2346
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001292973727311532,
      "loss": 0.8371,
      "step": 2347
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001292415153534567,
      "loss": 0.8765,
      "step": 2348
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012918564799742677,
      "loss": 0.8246,
      "step": 2349
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012912977068212753,
      "loss": 0.8422,
      "step": 2350
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001290738834266264,
      "loss": 0.9068,
      "step": 2351
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012901798624999434,
      "loss": 0.841,
      "step": 2352
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012896207917130562,
      "loss": 0.8522,
      "step": 2353
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012890616220963787,
      "loss": 0.8976,
      "step": 2354
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001288502353840721,
      "loss": 0.7613,
      "step": 2355
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00012879429871369272,
      "loss": 0.8572,
      "step": 2356
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012873835221758752,
      "loss": 0.8003,
      "step": 2357
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012868239591484754,
      "loss": 0.7857,
      "step": 2358
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012862642982456723,
      "loss": 0.7834,
      "step": 2359
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012857045396584443,
      "loss": 0.8625,
      "step": 2360
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012851446835778025,
      "loss": 0.8349,
      "step": 2361
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001284584730194791,
      "loss": 0.8741,
      "step": 2362
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001284024679700488,
      "loss": 0.8424,
      "step": 2363
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001283464532286004,
      "loss": 0.8419,
      "step": 2364
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012829042881424828,
      "loss": 0.801,
      "step": 2365
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012823439474611017,
      "loss": 0.8091,
      "step": 2366
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012817835104330706,
      "loss": 0.3521,
      "step": 2367
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001281222977249632,
      "loss": 0.8147,
      "step": 2368
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012806623481020614,
      "loss": 0.8301,
      "step": 2369
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012801016231816674,
      "loss": 0.8106,
      "step": 2370
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012795408026797903,
      "loss": 0.8568,
      "step": 2371
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012789798867878047,
      "loss": 0.8551,
      "step": 2372
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012784188756971161,
      "loss": 0.8205,
      "step": 2373
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001277857769599163,
      "loss": 0.8178,
      "step": 2374
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001277296568685417,
      "loss": 0.836,
      "step": 2375
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012767352731473812,
      "loss": 0.8635,
      "step": 2376
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001276173883176591,
      "loss": 0.8224,
      "step": 2377
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001275612398964615,
      "loss": 0.7684,
      "step": 2378
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012750508207030526,
      "loss": 0.9033,
      "step": 2379
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012744891485835365,
      "loss": 0.8503,
      "step": 2380
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012739273827977308,
      "loss": 0.872,
      "step": 2381
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012733655235373312,
      "loss": 0.8007,
      "step": 2382
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012728035709940665,
      "loss": 0.854,
      "step": 2383
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001272241525359696,
      "loss": 0.7287,
      "step": 2384
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012716793868260115,
      "loss": 0.8387,
      "step": 2385
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012711171555848368,
      "loss": 0.8608,
      "step": 2386
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012705548318280263,
      "loss": 0.8583,
      "step": 2387
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012699924157474665,
      "loss": 0.8403,
      "step": 2388
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012694299075350761,
      "loss": 0.7954,
      "step": 2389
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012688673073828042,
      "loss": 0.833,
      "step": 2390
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012683046154826318,
      "loss": 0.8446,
      "step": 2391
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001267741832026571,
      "loss": 0.8884,
      "step": 2392
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012671789572066654,
      "loss": 0.8023,
      "step": 2393
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012666159912149894,
      "loss": 0.8156,
      "step": 2394
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012660529342436487,
      "loss": 0.8229,
      "step": 2395
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012654897864847805,
      "loss": 0.8254,
      "step": 2396
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012649265481305517,
      "loss": 0.7955,
      "step": 2397
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012643632193731617,
      "loss": 0.8059,
      "step": 2398
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012637998004048396,
      "loss": 0.8346,
      "step": 2399
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012632362914178464,
      "loss": 0.8403,
      "step": 2400
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012626726926044722,
      "loss": 0.8991,
      "step": 2401
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001262109004157039,
      "loss": 0.8849,
      "step": 2402
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012615452262678995,
      "loss": 0.7797,
      "step": 2403
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012609813591294358,
      "loss": 0.8203,
      "step": 2404
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001260417402934061,
      "loss": 0.8503,
      "step": 2405
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000125985335787422,
      "loss": 0.8045,
      "step": 2406
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012592892241423856,
      "loss": 0.3329,
      "step": 2407
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012587250019310616,
      "loss": 0.8498,
      "step": 2408
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001258160691432784,
      "loss": 0.3692,
      "step": 2409
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012575962928401156,
      "loss": 0.8252,
      "step": 2410
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001257031806345652,
      "loss": 0.7907,
      "step": 2411
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00012564672321420178,
      "loss": 0.8098,
      "step": 2412
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001255902570421867,
      "loss": 0.8449,
      "step": 2413
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012553378213778842,
      "loss": 0.8861,
      "step": 2414
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012547729852027835,
      "loss": 0.8299,
      "step": 2415
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012542080620893092,
      "loss": 0.3629,
      "step": 2416
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001253643052230234,
      "loss": 0.729,
      "step": 2417
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012530779558183618,
      "loss": 0.8234,
      "step": 2418
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012525127730465255,
      "loss": 0.8583,
      "step": 2419
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012519475041075865,
      "loss": 0.8105,
      "step": 2420
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012513821491944363,
      "loss": 0.8562,
      "step": 2421
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001250816708499997,
      "loss": 0.8505,
      "step": 2422
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001250251182217218,
      "loss": 0.8292,
      "step": 2423
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012496855705390782,
      "loss": 0.8124,
      "step": 2424
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001249119873658587,
      "loss": 0.8234,
      "step": 2425
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012485540917687817,
      "loss": 0.8653,
      "step": 2426
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012479882250627288,
      "loss": 0.8281,
      "step": 2427
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012474222737335238,
      "loss": 0.89,
      "step": 2428
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012468562379742916,
      "loss": 0.8896,
      "step": 2429
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012462901179781846,
      "loss": 0.8619,
      "step": 2430
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012457239139383858,
      "loss": 0.8447,
      "step": 2431
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012451576260481053,
      "loss": 0.7739,
      "step": 2432
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012445912545005824,
      "loss": 0.8486,
      "step": 2433
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001244024799489085,
      "loss": 0.8813,
      "step": 2434
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012434582612069091,
      "loss": 0.8097,
      "step": 2435
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000124289163984738,
      "loss": 0.8454,
      "step": 2436
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012423249356038505,
      "loss": 0.8395,
      "step": 2437
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012417581486697017,
      "loss": 0.8528,
      "step": 2438
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012411912792383438,
      "loss": 0.7847,
      "step": 2439
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001240624327503214,
      "loss": 0.7762,
      "step": 2440
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001240057293657778,
      "loss": 0.7681,
      "step": 2441
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012394901778955302,
      "loss": 0.8142,
      "step": 2442
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012389229804099917,
      "loss": 0.8475,
      "step": 2443
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001238355701394713,
      "loss": 0.849,
      "step": 2444
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012377883410432706,
      "loss": 0.8217,
      "step": 2445
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012372208995492707,
      "loss": 0.8746,
      "step": 2446
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012366533771063457,
      "loss": 0.7853,
      "step": 2447
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001236085773908156,
      "loss": 0.8506,
      "step": 2448
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012355180901483902,
      "loss": 0.7972,
      "step": 2449
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001234950326020764,
      "loss": 0.8338,
      "step": 2450
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012343824817190194,
      "loss": 0.7932,
      "step": 2451
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001233814557436928,
      "loss": 0.8215,
      "step": 2452
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012332465533682868,
      "loss": 0.8519,
      "step": 2453
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012326784697069208,
      "loss": 0.8132,
      "step": 2454
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012321103066466823,
      "loss": 0.9215,
      "step": 2455
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012315420643814503,
      "loss": 0.8558,
      "step": 2456
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001230973743105131,
      "loss": 0.3773,
      "step": 2457
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012304053430116574,
      "loss": 0.8571,
      "step": 2458
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000122983686429499,
      "loss": 0.3525,
      "step": 2459
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001229268307149115,
      "loss": 0.8363,
      "step": 2460
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001228699671768047,
      "loss": 0.8654,
      "step": 2461
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001228130958345825,
      "loss": 0.8582,
      "step": 2462
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012275621670765173,
      "loss": 0.8487,
      "step": 2463
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012269932981542165,
      "loss": 0.3332,
      "step": 2464
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001226424351773043,
      "loss": 0.9057,
      "step": 2465
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012258553281271434,
      "loss": 0.7671,
      "step": 2466
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00012252862274106903,
      "loss": 0.8583,
      "step": 2467
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012247170498178828,
      "loss": 0.8108,
      "step": 2468
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012241477955429465,
      "loss": 0.8219,
      "step": 2469
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012235784647801325,
      "loss": 0.338,
      "step": 2470
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012230090577237187,
      "loss": 0.8126,
      "step": 2471
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001222439574568009,
      "loss": 0.3381,
      "step": 2472
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012218700155073328,
      "loss": 0.8137,
      "step": 2473
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012213003807360452,
      "loss": 0.8387,
      "step": 2474
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012207306704485285,
      "loss": 0.8116,
      "step": 2475
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001220160884839189,
      "loss": 0.8421,
      "step": 2476
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012195910241024599,
      "loss": 0.8454,
      "step": 2477
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012190210884327996,
      "loss": 0.8144,
      "step": 2478
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012184510780246923,
      "loss": 0.8311,
      "step": 2479
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012178809930726472,
      "loss": 0.8509,
      "step": 2480
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012173108337711996,
      "loss": 0.8722,
      "step": 2481
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012167406003149096,
      "loss": 0.8636,
      "step": 2482
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012161702928983634,
      "loss": 0.8658,
      "step": 2483
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012155999117161712,
      "loss": 0.8492,
      "step": 2484
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012150294569629692,
      "loss": 0.925,
      "step": 2485
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012144589288334192,
      "loss": 0.8043,
      "step": 2486
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012138883275222067,
      "loss": 0.8575,
      "step": 2487
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012133176532240431,
      "loss": 0.867,
      "step": 2488
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012127469061336647,
      "loss": 0.8285,
      "step": 2489
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012121760864458322,
      "loss": 0.8726,
      "step": 2490
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012116051943553315,
      "loss": 0.8123,
      "step": 2491
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012110342300569728,
      "loss": 0.8155,
      "step": 2492
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012104631937455915,
      "loss": 0.7717,
      "step": 2493
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012098920856160465,
      "loss": 0.8665,
      "step": 2494
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012093209058632227,
      "loss": 0.7804,
      "step": 2495
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012087496546820284,
      "loss": 0.844,
      "step": 2496
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012081783322673963,
      "loss": 0.8765,
      "step": 2497
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001207606938814284,
      "loss": 0.9293,
      "step": 2498
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012070354745176728,
      "loss": 0.7539,
      "step": 2499
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012064639395725688,
      "loss": 0.829,
      "step": 2500
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001205892334174001,
      "loss": 0.3674,
      "step": 2501
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012053206585170236,
      "loss": 0.8493,
      "step": 2502
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012047489127967148,
      "loss": 0.8555,
      "step": 2503
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012041770972081758,
      "loss": 0.34,
      "step": 2504
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012036052119465322,
      "loss": 0.8038,
      "step": 2505
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012030332572069335,
      "loss": 0.8484,
      "step": 2506
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012024612331845526,
      "loss": 0.8713,
      "step": 2507
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001201889140074586,
      "loss": 0.7692,
      "step": 2508
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012013169780722542,
      "loss": 0.808,
      "step": 2509
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012007447473728013,
      "loss": 0.8338,
      "step": 2510
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00012001724481714938,
      "loss": 0.8231,
      "step": 2511
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011996000806636223,
      "loss": 0.8359,
      "step": 2512
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011990276450445015,
      "loss": 0.8023,
      "step": 2513
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011984551415094675,
      "loss": 0.7833,
      "step": 2514
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011978825702538814,
      "loss": 0.7995,
      "step": 2515
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011973099314731259,
      "loss": 0.8795,
      "step": 2516
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011967372253626083,
      "loss": 0.8183,
      "step": 2517
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011961644521177573,
      "loss": 0.863,
      "step": 2518
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011955916119340254,
      "loss": 0.8413,
      "step": 2519
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011950187050068882,
      "loss": 0.8183,
      "step": 2520
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011944457315318429,
      "loss": 0.3505,
      "step": 2521
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011938726917044108,
      "loss": 0.8277,
      "step": 2522
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001193299585720135,
      "loss": 0.8149,
      "step": 2523
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011927264137745812,
      "loss": 0.8721,
      "step": 2524
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001192153176063338,
      "loss": 0.8735,
      "step": 2525
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001191579872782016,
      "loss": 0.8179,
      "step": 2526
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001191006504126249,
      "loss": 0.8161,
      "step": 2527
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011904330702916917,
      "loss": 0.3905,
      "step": 2528
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011898595714740224,
      "loss": 0.8513,
      "step": 2529
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001189286007868941,
      "loss": 0.3604,
      "step": 2530
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001188712379672169,
      "loss": 0.8251,
      "step": 2531
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011881386870794511,
      "loss": 0.8124,
      "step": 2532
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011875649302865532,
      "loss": 0.8175,
      "step": 2533
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001186991109489263,
      "loss": 0.8217,
      "step": 2534
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011864172248833904,
      "loss": 0.8055,
      "step": 2535
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011858432766647672,
      "loss": 0.7886,
      "step": 2536
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011852692650292467,
      "loss": 0.8576,
      "step": 2537
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011846951901727034,
      "loss": 0.8621,
      "step": 2538
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001184121052291034,
      "loss": 0.901,
      "step": 2539
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011835468515801569,
      "loss": 0.915,
      "step": 2540
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001182972588236011,
      "loss": 0.8059,
      "step": 2541
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011823982624545571,
      "loss": 0.8588,
      "step": 2542
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011818238744317778,
      "loss": 0.8397,
      "step": 2543
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001181249424363676,
      "loss": 0.8381,
      "step": 2544
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011806749124462766,
      "loss": 0.7887,
      "step": 2545
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011801003388756246,
      "loss": 0.7777,
      "step": 2546
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011795257038477875,
      "loss": 0.8324,
      "step": 2547
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011789510075588525,
      "loss": 0.7985,
      "step": 2548
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011783762502049283,
      "loss": 0.8457,
      "step": 2549
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011778014319821442,
      "loss": 0.8062,
      "step": 2550
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011772265530866502,
      "loss": 0.8243,
      "step": 2551
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011766516137146176,
      "loss": 0.8524,
      "step": 2552
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011760766140622372,
      "loss": 0.8045,
      "step": 2553
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011755015543257219,
      "loss": 0.7853,
      "step": 2554
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011749264347013036,
      "loss": 0.8205,
      "step": 2555
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011743512553852356,
      "loss": 0.8308,
      "step": 2556
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011737760165737915,
      "loss": 0.8202,
      "step": 2557
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011732007184632645,
      "loss": 0.3464,
      "step": 2558
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011726253612499687,
      "loss": 0.7977,
      "step": 2559
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011720499451302385,
      "loss": 0.8572,
      "step": 2560
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011714744703004277,
      "loss": 0.8056,
      "step": 2561
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011708989369569105,
      "loss": 0.8032,
      "step": 2562
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011703233452960812,
      "loss": 0.842,
      "step": 2563
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011697476955143543,
      "loss": 0.8198,
      "step": 2564
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011691719878081629,
      "loss": 0.782,
      "step": 2565
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011685962223739613,
      "loss": 0.8425,
      "step": 2566
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011680203994082228,
      "loss": 0.8219,
      "step": 2567
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011674445191074401,
      "loss": 0.8382,
      "step": 2568
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001166868581668126,
      "loss": 0.8168,
      "step": 2569
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011662925872868127,
      "loss": 0.8157,
      "step": 2570
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011657165361600517,
      "loss": 0.8699,
      "step": 2571
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011651404284844134,
      "loss": 0.3449,
      "step": 2572
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011645642644564885,
      "loss": 0.8672,
      "step": 2573
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011639880442728865,
      "loss": 0.8026,
      "step": 2574
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011634117681302355,
      "loss": 0.8093,
      "step": 2575
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001162835436225183,
      "loss": 0.8034,
      "step": 2576
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011622590487543965,
      "loss": 0.8119,
      "step": 2577
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011616826059145612,
      "loss": 0.847,
      "step": 2578
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011611061079023813,
      "loss": 0.8688,
      "step": 2579
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011605295549145804,
      "loss": 0.8056,
      "step": 2580
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011599529471479012,
      "loss": 0.3348,
      "step": 2581
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011593762847991036,
      "loss": 0.7987,
      "step": 2582
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011587995680649676,
      "loss": 0.7659,
      "step": 2583
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011582227971422914,
      "loss": 0.8273,
      "step": 2584
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001157645972227891,
      "loss": 0.8754,
      "step": 2585
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011570690935186012,
      "loss": 0.8264,
      "step": 2586
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001156492161211276,
      "loss": 0.8309,
      "step": 2587
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011559151755027863,
      "loss": 0.7918,
      "step": 2588
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011553381365900221,
      "loss": 0.8071,
      "step": 2589
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011547610446698915,
      "loss": 0.8394,
      "step": 2590
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011541838999393206,
      "loss": 0.8168,
      "step": 2591
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011536067025952531,
      "loss": 0.8789,
      "step": 2592
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011530294528346513,
      "loss": 0.8589,
      "step": 2593
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011524521508544948,
      "loss": 0.8078,
      "step": 2594
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011518747968517816,
      "loss": 0.8483,
      "step": 2595
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011512973910235267,
      "loss": 0.7986,
      "step": 2596
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001150719933566764,
      "loss": 0.8076,
      "step": 2597
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011501424246785437,
      "loss": 0.8474,
      "step": 2598
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011495648645559342,
      "loss": 0.8671,
      "step": 2599
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011489872533960215,
      "loss": 0.3603,
      "step": 2600
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001148409591395909,
      "loss": 0.8112,
      "step": 2601
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011478318787527164,
      "loss": 0.8624,
      "step": 2602
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011472541156635823,
      "loss": 0.7931,
      "step": 2603
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011466763023256619,
      "loss": 0.7945,
      "step": 2604
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011460984389361266,
      "loss": 0.8357,
      "step": 2605
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011455205256921666,
      "loss": 0.807,
      "step": 2606
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011449425627909875,
      "loss": 0.8725,
      "step": 2607
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011443645504298128,
      "loss": 0.7504,
      "step": 2608
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001143786488805883,
      "loss": 0.8536,
      "step": 2609
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011432083781164542,
      "loss": 0.8794,
      "step": 2610
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011426302185588009,
      "loss": 0.9141,
      "step": 2611
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011420520103302128,
      "loss": 0.8665,
      "step": 2612
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011414737536279971,
      "loss": 0.8094,
      "step": 2613
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011408954486494778,
      "loss": 0.8521,
      "step": 2614
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011403170955919942,
      "loss": 0.8115,
      "step": 2615
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011397386946529027,
      "loss": 0.8361,
      "step": 2616
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011391602460295768,
      "loss": 0.801,
      "step": 2617
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011385817499194042,
      "loss": 0.7752,
      "step": 2618
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011380032065197912,
      "loss": 0.7776,
      "step": 2619
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011374246160281588,
      "loss": 0.8534,
      "step": 2620
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011368459786419445,
      "loss": 0.8709,
      "step": 2621
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011362672945586012,
      "loss": 0.824,
      "step": 2622
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011356885639755988,
      "loss": 0.8351,
      "step": 2623
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011351097870904224,
      "loss": 0.8146,
      "step": 2624
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011345309641005726,
      "loss": 0.8674,
      "step": 2625
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011339520952035667,
      "loss": 0.8531,
      "step": 2626
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001133373180596937,
      "loss": 0.8802,
      "step": 2627
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011327942204782312,
      "loss": 0.8533,
      "step": 2628
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011322152150450128,
      "loss": 0.7961,
      "step": 2629
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011316361644948614,
      "loss": 0.8414,
      "step": 2630
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011310570690253706,
      "loss": 0.8925,
      "step": 2631
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011304779288341507,
      "loss": 0.8168,
      "step": 2632
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00011298987441188265,
      "loss": 0.8291,
      "step": 2633
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011293195150770381,
      "loss": 0.8386,
      "step": 2634
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011287402419064411,
      "loss": 0.8068,
      "step": 2635
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011281609248047053,
      "loss": 0.7855,
      "step": 2636
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011275815639695167,
      "loss": 0.8431,
      "step": 2637
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011270021595985749,
      "loss": 0.7435,
      "step": 2638
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011264227118895956,
      "loss": 0.8487,
      "step": 2639
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011258432210403087,
      "loss": 0.8434,
      "step": 2640
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011252636872484584,
      "loss": 0.8493,
      "step": 2641
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001124684110711804,
      "loss": 0.8848,
      "step": 2642
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011241044916281198,
      "loss": 0.8248,
      "step": 2643
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011235248301951944,
      "loss": 0.8882,
      "step": 2644
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011229451266108297,
      "loss": 0.8634,
      "step": 2645
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011223653810728439,
      "loss": 0.8369,
      "step": 2646
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001121785593779068,
      "loss": 0.8199,
      "step": 2647
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011212057649273478,
      "loss": 0.8378,
      "step": 2648
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011206258947155434,
      "loss": 0.8062,
      "step": 2649
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011200459833415293,
      "loss": 0.8093,
      "step": 2650
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011194660310031931,
      "loss": 0.8015,
      "step": 2651
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011188860378984367,
      "loss": 0.8405,
      "step": 2652
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011183060042251767,
      "loss": 0.8738,
      "step": 2653
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001117725930181343,
      "loss": 0.8661,
      "step": 2654
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011171458159648788,
      "loss": 0.844,
      "step": 2655
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011165656617737417,
      "loss": 0.7832,
      "step": 2656
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011159854678059027,
      "loss": 0.3319,
      "step": 2657
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011154052342593464,
      "loss": 0.9074,
      "step": 2658
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011148249613320707,
      "loss": 0.7717,
      "step": 2659
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011142446492220875,
      "loss": 0.8311,
      "step": 2660
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011136642981274212,
      "loss": 0.8258,
      "step": 2661
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011130839082461102,
      "loss": 0.8367,
      "step": 2662
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011125034797762062,
      "loss": 0.8083,
      "step": 2663
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011119230129157731,
      "loss": 0.8414,
      "step": 2664
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011113425078628892,
      "loss": 0.8105,
      "step": 2665
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011107619648156449,
      "loss": 0.8643,
      "step": 2666
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011101813839721444,
      "loss": 0.3772,
      "step": 2667
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011096007655305037,
      "loss": 0.8304,
      "step": 2668
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011090201096888524,
      "loss": 0.836,
      "step": 2669
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011084394166453329,
      "loss": 0.8562,
      "step": 2670
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011078586865980995,
      "loss": 0.7853,
      "step": 2671
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011072779197453204,
      "loss": 0.789,
      "step": 2672
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011066971162851754,
      "loss": 0.853,
      "step": 2673
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001106116276415857,
      "loss": 0.8288,
      "step": 2674
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000110553540033557,
      "loss": 0.3356,
      "step": 2675
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001104954488242532,
      "loss": 0.8537,
      "step": 2676
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011043735403349726,
      "loss": 0.8073,
      "step": 2677
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011037925568111336,
      "loss": 0.8699,
      "step": 2678
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011032115378692687,
      "loss": 0.3536,
      "step": 2679
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011026304837076446,
      "loss": 0.8408,
      "step": 2680
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011020493945245392,
      "loss": 0.8618,
      "step": 2681
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001101468270518242,
      "loss": 0.8561,
      "step": 2682
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011008871118870556,
      "loss": 0.7806,
      "step": 2683
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00011003059188292935,
      "loss": 0.8956,
      "step": 2684
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010997246915432811,
      "loss": 0.8124,
      "step": 2685
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010991434302273557,
      "loss": 0.8371,
      "step": 2686
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010985621350798663,
      "loss": 0.8613,
      "step": 2687
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010979808062991727,
      "loss": 0.8303,
      "step": 2688
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010973994440836473,
      "loss": 0.8361,
      "step": 2689
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010968180486316731,
      "loss": 0.8393,
      "step": 2690
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010962366201416443,
      "loss": 0.7898,
      "step": 2691
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010956551588119673,
      "loss": 0.8482,
      "step": 2692
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010950736648410589,
      "loss": 0.8138,
      "step": 2693
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001094492138427347,
      "loss": 0.7989,
      "step": 2694
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010939105797692713,
      "loss": 0.8527,
      "step": 2695
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010933289890652821,
      "loss": 0.7893,
      "step": 2696
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010927473665138401,
      "loss": 0.7953,
      "step": 2697
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010921657123134178,
      "loss": 0.8269,
      "step": 2698
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010915840266624975,
      "loss": 0.7748,
      "step": 2699
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010910023097595736,
      "loss": 0.7902,
      "step": 2700
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010904205618031497,
      "loss": 0.7681,
      "step": 2701
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010898387829917411,
      "loss": 0.8299,
      "step": 2702
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010892569735238731,
      "loss": 0.8854,
      "step": 2703
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010886751335980811,
      "loss": 0.7559,
      "step": 2704
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010880932634129117,
      "loss": 0.8792,
      "step": 2705
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010875113631669216,
      "loss": 0.8127,
      "step": 2706
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010869294330586773,
      "loss": 0.7956,
      "step": 2707
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010863474732867558,
      "loss": 0.8359,
      "step": 2708
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010857654840497444,
      "loss": 0.8886,
      "step": 2709
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010851834655462405,
      "loss": 0.3374,
      "step": 2710
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010846014179748506,
      "loss": 0.8,
      "step": 2711
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010840193415341926,
      "loss": 0.8294,
      "step": 2712
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010834372364228931,
      "loss": 0.8246,
      "step": 2713
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010828551028395886,
      "loss": 0.8328,
      "step": 2714
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010822729409829257,
      "loss": 0.8387,
      "step": 2715
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010816907510515608,
      "loss": 0.8452,
      "step": 2716
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010811085332441591,
      "loss": 0.8057,
      "step": 2717
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010805262877593957,
      "loss": 0.8744,
      "step": 2718
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010799440147959559,
      "loss": 0.8109,
      "step": 2719
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010793617145525332,
      "loss": 0.8584,
      "step": 2720
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010787793872278308,
      "loss": 0.7744,
      "step": 2721
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010781970330205618,
      "loss": 0.7612,
      "step": 2722
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010776146521294475,
      "loss": 0.8539,
      "step": 2723
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010770322447532188,
      "loss": 0.8036,
      "step": 2724
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010764498110906156,
      "loss": 0.847,
      "step": 2725
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001075867351340387,
      "loss": 0.8263,
      "step": 2726
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010752848657012901,
      "loss": 0.7923,
      "step": 2727
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001074702354372092,
      "loss": 0.8334,
      "step": 2728
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001074119817551568,
      "loss": 0.8126,
      "step": 2729
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010735372554385019,
      "loss": 0.7507,
      "step": 2730
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010729546682316867,
      "loss": 0.3325,
      "step": 2731
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010723720561299233,
      "loss": 0.8118,
      "step": 2732
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010717894193320217,
      "loss": 0.8385,
      "step": 2733
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010712067580367999,
      "loss": 0.8951,
      "step": 2734
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010706240724430845,
      "loss": 0.8077,
      "step": 2735
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010700413627497105,
      "loss": 0.7991,
      "step": 2736
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010694586291555204,
      "loss": 0.821,
      "step": 2737
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010688758718593659,
      "loss": 0.8602,
      "step": 2738
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010682930910601061,
      "loss": 0.8378,
      "step": 2739
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001067710286956608,
      "loss": 0.81,
      "step": 2740
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010671274597477474,
      "loss": 0.8695,
      "step": 2741
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001066544609632407,
      "loss": 0.822,
      "step": 2742
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001065961736809478,
      "loss": 0.8089,
      "step": 2743
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010653788414778592,
      "loss": 0.3364,
      "step": 2744
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010647959238364564,
      "loss": 0.7334,
      "step": 2745
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010642129840841843,
      "loss": 0.8814,
      "step": 2746
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001063630022419964,
      "loss": 0.8704,
      "step": 2747
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010630470390427248,
      "loss": 0.8334,
      "step": 2748
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001062464034151403,
      "loss": 0.3449,
      "step": 2749
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010618810079449419,
      "loss": 0.7873,
      "step": 2750
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010612979606222932,
      "loss": 0.8098,
      "step": 2751
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010607148923824151,
      "loss": 0.8103,
      "step": 2752
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010601318034242726,
      "loss": 0.8301,
      "step": 2753
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010595486939468384,
      "loss": 0.8655,
      "step": 2754
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010589655641490919,
      "loss": 0.8342,
      "step": 2755
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010583824142300198,
      "loss": 0.8183,
      "step": 2756
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010577992443886147,
      "loss": 0.8804,
      "step": 2757
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001057216054823877,
      "loss": 0.85,
      "step": 2758
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010566328457348138,
      "loss": 0.8125,
      "step": 2759
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010560496173204379,
      "loss": 0.3474,
      "step": 2760
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010554663697797698,
      "loss": 0.8334,
      "step": 2761
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010548831033118363,
      "loss": 0.8501,
      "step": 2762
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010542998181156695,
      "loss": 0.8055,
      "step": 2763
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010537165143903096,
      "loss": 0.7968,
      "step": 2764
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010531331923348023,
      "loss": 0.8336,
      "step": 2765
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010525498521481993,
      "loss": 0.7907,
      "step": 2766
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001051966494029559,
      "loss": 0.8388,
      "step": 2767
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010513831181779455,
      "loss": 0.7568,
      "step": 2768
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010507997247924293,
      "loss": 0.7421,
      "step": 2769
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010502163140720866,
      "loss": 0.8412,
      "step": 2770
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010496328862159999,
      "loss": 0.7874,
      "step": 2771
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010490494414232572,
      "loss": 0.8407,
      "step": 2772
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010484659798929521,
      "loss": 0.8248,
      "step": 2773
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010478825018241847,
      "loss": 0.8253,
      "step": 2774
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010472990074160596,
      "loss": 0.7896,
      "step": 2775
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010467154968676884,
      "loss": 0.8034,
      "step": 2776
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010461319703781866,
      "loss": 0.8058,
      "step": 2777
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010455484281466763,
      "loss": 0.7833,
      "step": 2778
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010449648703722852,
      "loss": 0.3756,
      "step": 2779
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010443812972541447,
      "loss": 0.8058,
      "step": 2780
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010437977089913928,
      "loss": 0.7926,
      "step": 2781
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010432141057831731,
      "loss": 0.8125,
      "step": 2782
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010426304878286323,
      "loss": 0.8214,
      "step": 2783
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010420468553269243,
      "loss": 0.8692,
      "step": 2784
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010414632084772067,
      "loss": 0.7992,
      "step": 2785
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010408795474786423,
      "loss": 0.7885,
      "step": 2786
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010402958725303989,
      "loss": 0.8127,
      "step": 2787
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010397121838316488,
      "loss": 0.8509,
      "step": 2788
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010391284815815691,
      "loss": 0.7501,
      "step": 2789
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010385447659793417,
      "loss": 0.8398,
      "step": 2790
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010379610372241524,
      "loss": 0.8454,
      "step": 2791
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010373772955151928,
      "loss": 0.824,
      "step": 2792
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010367935410516572,
      "loss": 0.8429,
      "step": 2793
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010362097740327457,
      "loss": 0.8275,
      "step": 2794
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010356259946576623,
      "loss": 0.8227,
      "step": 2795
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010350422031256144,
      "loss": 0.833,
      "step": 2796
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010344583996358146,
      "loss": 0.8087,
      "step": 2797
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001033874584387479,
      "loss": 0.8499,
      "step": 2798
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010332907575798285,
      "loss": 0.8316,
      "step": 2799
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010327069194120867,
      "loss": 0.8213,
      "step": 2800
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010321230700834816,
      "loss": 0.8608,
      "step": 2801
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010315392097932458,
      "loss": 0.8292,
      "step": 2802
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010309553387406142,
      "loss": 0.8237,
      "step": 2803
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010303714571248264,
      "loss": 0.8321,
      "step": 2804
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010297875651451259,
      "loss": 0.8041,
      "step": 2805
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010292036630007584,
      "loss": 0.8077,
      "step": 2806
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001028619750890974,
      "loss": 0.3428,
      "step": 2807
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010280358290150264,
      "loss": 0.8381,
      "step": 2808
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010274518975721723,
      "loss": 0.7848,
      "step": 2809
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001026867956761671,
      "loss": 0.7997,
      "step": 2810
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010262840067827861,
      "loss": 0.353,
      "step": 2811
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010257000478347843,
      "loss": 0.8205,
      "step": 2812
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001025116080116934,
      "loss": 0.8774,
      "step": 2813
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010245321038285078,
      "loss": 0.3227,
      "step": 2814
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010239481191687817,
      "loss": 0.8114,
      "step": 2815
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010233641263370328,
      "loss": 0.3477,
      "step": 2816
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010227801255325424,
      "loss": 0.8292,
      "step": 2817
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010221961169545941,
      "loss": 0.861,
      "step": 2818
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010216121008024744,
      "loss": 0.8422,
      "step": 2819
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010210280772754717,
      "loss": 0.8909,
      "step": 2820
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010204440465728777,
      "loss": 0.7633,
      "step": 2821
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010198600088939863,
      "loss": 0.8637,
      "step": 2822
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010192759644380932,
      "loss": 0.8348,
      "step": 2823
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010186919134044969,
      "loss": 0.8409,
      "step": 2824
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010181078559924991,
      "loss": 0.8255,
      "step": 2825
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010175237924014016,
      "loss": 0.8436,
      "step": 2826
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010169397228305102,
      "loss": 0.8314,
      "step": 2827
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010163556474791316,
      "loss": 0.8178,
      "step": 2828
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010157715665465747,
      "loss": 0.8284,
      "step": 2829
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001015187480232151,
      "loss": 0.7916,
      "step": 2830
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001014603388735173,
      "loss": 0.8594,
      "step": 2831
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010140192922549551,
      "loss": 0.8503,
      "step": 2832
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010134351909908136,
      "loss": 0.8136,
      "step": 2833
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010128510851420664,
      "loss": 0.8204,
      "step": 2834
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010122669749080335,
      "loss": 0.8541,
      "step": 2835
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001011682860488035,
      "loss": 0.8274,
      "step": 2836
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010110987420813937,
      "loss": 0.3291,
      "step": 2837
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010105146198874338,
      "loss": 0.7476,
      "step": 2838
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010099304941054792,
      "loss": 0.8271,
      "step": 2839
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010093463649348572,
      "loss": 0.8023,
      "step": 2840
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010087622325748944,
      "loss": 0.815,
      "step": 2841
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010081780972249201,
      "loss": 0.806,
      "step": 2842
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010075939590842634,
      "loss": 0.8547,
      "step": 2843
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010070098183522548,
      "loss": 0.843,
      "step": 2844
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010064256752282258,
      "loss": 0.8194,
      "step": 2845
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010058415299115082,
      "loss": 0.3238,
      "step": 2846
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001005257382601435,
      "loss": 0.7643,
      "step": 2847
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010046732334973406,
      "loss": 0.808,
      "step": 2848
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010040890827985583,
      "loss": 0.8071,
      "step": 2849
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010035049307044229,
      "loss": 0.8562,
      "step": 2850
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010029207774142702,
      "loss": 0.8222,
      "step": 2851
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010023366231274354,
      "loss": 0.8048,
      "step": 2852
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010017524680432547,
      "loss": 0.8754,
      "step": 2853
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010011683123610642,
      "loss": 0.8346,
      "step": 2854
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010005841562802005,
      "loss": 0.8192,
      "step": 2855
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001,
      "loss": 0.772,
      "step": 2856
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.994158437197999e-05,
      "loss": 0.8517,
      "step": 2857
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.988316876389359e-05,
      "loss": 0.8676,
      "step": 2858
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.982475319567456e-05,
      "loss": 0.8385,
      "step": 2859
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.97663376872565e-05,
      "loss": 0.8486,
      "step": 2860
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.970792225857299e-05,
      "loss": 0.3634,
      "step": 2861
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.964950692955772e-05,
      "loss": 0.8277,
      "step": 2862
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.959109172014423e-05,
      "loss": 0.8409,
      "step": 2863
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.953267665026596e-05,
      "loss": 0.3673,
      "step": 2864
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.94742617398565e-05,
      "loss": 0.8003,
      "step": 2865
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.941584700884923e-05,
      "loss": 0.8103,
      "step": 2866
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.935743247717745e-05,
      "loss": 0.8434,
      "step": 2867
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.929901816477455e-05,
      "loss": 0.7695,
      "step": 2868
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.924060409157366e-05,
      "loss": 0.8383,
      "step": 2869
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.9182190277508e-05,
      "loss": 0.834,
      "step": 2870
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.912377674251058e-05,
      "loss": 0.8675,
      "step": 2871
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.906536350651429e-05,
      "loss": 0.7961,
      "step": 2872
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.900695058945209e-05,
      "loss": 0.8588,
      "step": 2873
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.894853801125667e-05,
      "loss": 0.8285,
      "step": 2874
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.889012579186062e-05,
      "loss": 0.8292,
      "step": 2875
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.883171395119652e-05,
      "loss": 0.8345,
      "step": 2876
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.87733025091967e-05,
      "loss": 0.8006,
      "step": 2877
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.871489148579335e-05,
      "loss": 0.8378,
      "step": 2878
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.865648090091865e-05,
      "loss": 0.8917,
      "step": 2879
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.859807077450454e-05,
      "loss": 0.3631,
      "step": 2880
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.853966112648271e-05,
      "loss": 0.8578,
      "step": 2881
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.848125197678492e-05,
      "loss": 0.8424,
      "step": 2882
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.842284334534255e-05,
      "loss": 0.8615,
      "step": 2883
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.836443525208686e-05,
      "loss": 0.7833,
      "step": 2884
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.8306027716949e-05,
      "loss": 0.7965,
      "step": 2885
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.824762075985987e-05,
      "loss": 0.8041,
      "step": 2886
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.818921440075012e-05,
      "loss": 0.8267,
      "step": 2887
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.813080865955032e-05,
      "loss": 0.7847,
      "step": 2888
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.807240355619074e-05,
      "loss": 0.82,
      "step": 2889
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.80139991106014e-05,
      "loss": 0.7751,
      "step": 2890
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.795559534271226e-05,
      "loss": 0.82,
      "step": 2891
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.789719227245283e-05,
      "loss": 0.7863,
      "step": 2892
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.783878991975258e-05,
      "loss": 0.8909,
      "step": 2893
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.778038830454061e-05,
      "loss": 0.8,
      "step": 2894
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.772198744674577e-05,
      "loss": 0.3008,
      "step": 2895
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.766358736629674e-05,
      "loss": 0.8776,
      "step": 2896
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.760518808312188e-05,
      "loss": 0.3529,
      "step": 2897
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.75467896171492e-05,
      "loss": 0.8575,
      "step": 2898
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.748839198830663e-05,
      "loss": 0.3605,
      "step": 2899
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.742999521652163e-05,
      "loss": 0.8092,
      "step": 2900
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.737159932172137e-05,
      "loss": 0.7832,
      "step": 2901
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.731320432383291e-05,
      "loss": 0.8156,
      "step": 2902
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.725481024278282e-05,
      "loss": 0.8066,
      "step": 2903
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.719641709849737e-05,
      "loss": 0.3289,
      "step": 2904
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.713802491090261e-05,
      "loss": 0.8089,
      "step": 2905
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.707963369992421e-05,
      "loss": 0.8798,
      "step": 2906
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.702124348548743e-05,
      "loss": 0.8192,
      "step": 2907
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.696285428751738e-05,
      "loss": 0.7832,
      "step": 2908
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.690446612593863e-05,
      "loss": 0.8241,
      "step": 2909
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.684607902067545e-05,
      "loss": 0.835,
      "step": 2910
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.678769299165186e-05,
      "loss": 0.8197,
      "step": 2911
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.672930805879134e-05,
      "loss": 0.8136,
      "step": 2912
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.667092424201717e-05,
      "loss": 0.8134,
      "step": 2913
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.661254156125212e-05,
      "loss": 0.8316,
      "step": 2914
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.655416003641854e-05,
      "loss": 0.7912,
      "step": 2915
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.649577968743859e-05,
      "loss": 0.8649,
      "step": 2916
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.643740053423383e-05,
      "loss": 0.8401,
      "step": 2917
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.637902259672541e-05,
      "loss": 0.7625,
      "step": 2918
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.63206458948343e-05,
      "loss": 0.3699,
      "step": 2919
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.626227044848077e-05,
      "loss": 0.8401,
      "step": 2920
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.620389627758476e-05,
      "loss": 0.8036,
      "step": 2921
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.614552340206586e-05,
      "loss": 0.3444,
      "step": 2922
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.608715184184313e-05,
      "loss": 0.8209,
      "step": 2923
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.602878161683513e-05,
      "loss": 0.8201,
      "step": 2924
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.597041274696012e-05,
      "loss": 0.8162,
      "step": 2925
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.591204525213579e-05,
      "loss": 0.8026,
      "step": 2926
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.585367915227934e-05,
      "loss": 0.3062,
      "step": 2927
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.57953144673076e-05,
      "loss": 0.7371,
      "step": 2928
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.57369512171368e-05,
      "loss": 0.8792,
      "step": 2929
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.567858942168271e-05,
      "loss": 0.8035,
      "step": 2930
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.562022910086074e-05,
      "loss": 0.7769,
      "step": 2931
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.556187027458558e-05,
      "loss": 0.7954,
      "step": 2932
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.550351296277151e-05,
      "loss": 0.8283,
      "step": 2933
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.544515718533238e-05,
      "loss": 0.8241,
      "step": 2934
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.538680296218135e-05,
      "loss": 0.8413,
      "step": 2935
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.532845031323118e-05,
      "loss": 0.8275,
      "step": 2936
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.527009925839405e-05,
      "loss": 0.8091,
      "step": 2937
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.521174981758154e-05,
      "loss": 0.8047,
      "step": 2938
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.51534020107048e-05,
      "loss": 0.8124,
      "step": 2939
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.509505585767432e-05,
      "loss": 0.7897,
      "step": 2940
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.503671137840001e-05,
      "loss": 0.7955,
      "step": 2941
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.497836859279135e-05,
      "loss": 0.8212,
      "step": 2942
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.492002752075711e-05,
      "loss": 0.7664,
      "step": 2943
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.486168818220546e-05,
      "loss": 0.8513,
      "step": 2944
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.480335059704413e-05,
      "loss": 0.831,
      "step": 2945
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.474501478518011e-05,
      "loss": 0.7097,
      "step": 2946
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.468668076651977e-05,
      "loss": 0.8631,
      "step": 2947
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.462834856096904e-05,
      "loss": 0.3569,
      "step": 2948
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.457001818843308e-05,
      "loss": 0.8124,
      "step": 2949
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.45116896688164e-05,
      "loss": 0.7815,
      "step": 2950
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.445336302202303e-05,
      "loss": 0.8097,
      "step": 2951
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.439503826795625e-05,
      "loss": 0.7967,
      "step": 2952
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.433671542651864e-05,
      "loss": 0.7818,
      "step": 2953
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.427839451761233e-05,
      "loss": 0.7993,
      "step": 2954
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.422007556113858e-05,
      "loss": 0.8054,
      "step": 2955
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.416175857699805e-05,
      "loss": 0.8024,
      "step": 2956
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.410344358509082e-05,
      "loss": 0.8405,
      "step": 2957
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.404513060531615e-05,
      "loss": 0.8532,
      "step": 2958
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.398681965757275e-05,
      "loss": 0.8368,
      "step": 2959
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.392851076175853e-05,
      "loss": 0.7784,
      "step": 2960
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.387020393777068e-05,
      "loss": 0.7236,
      "step": 2961
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.381189920550583e-05,
      "loss": 0.8202,
      "step": 2962
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.375359658485976e-05,
      "loss": 0.7824,
      "step": 2963
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.369529609572753e-05,
      "loss": 0.8111,
      "step": 2964
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.363699775800361e-05,
      "loss": 0.8328,
      "step": 2965
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.35787015915816e-05,
      "loss": 0.7851,
      "step": 2966
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.352040761635435e-05,
      "loss": 0.8233,
      "step": 2967
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.346211585221411e-05,
      "loss": 0.3288,
      "step": 2968
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.340382631905223e-05,
      "loss": 0.8319,
      "step": 2969
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.334553903675931e-05,
      "loss": 0.8261,
      "step": 2970
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.32872540252253e-05,
      "loss": 0.8323,
      "step": 2971
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.322897130433924e-05,
      "loss": 0.8269,
      "step": 2972
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.317069089398942e-05,
      "loss": 0.8534,
      "step": 2973
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.311241281406345e-05,
      "loss": 0.7655,
      "step": 2974
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.305413708444801e-05,
      "loss": 0.7812,
      "step": 2975
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.299586372502897e-05,
      "loss": 0.773,
      "step": 2976
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.293759275569157e-05,
      "loss": 0.8046,
      "step": 2977
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.287932419632001e-05,
      "loss": 0.7979,
      "step": 2978
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.282105806679784e-05,
      "loss": 0.7731,
      "step": 2979
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.27627943870077e-05,
      "loss": 0.8462,
      "step": 2980
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.270453317683134e-05,
      "loss": 0.836,
      "step": 2981
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.264627445614983e-05,
      "loss": 0.8126,
      "step": 2982
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.258801824484325e-05,
      "loss": 0.8112,
      "step": 2983
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.252976456279081e-05,
      "loss": 0.7952,
      "step": 2984
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.247151342987101e-05,
      "loss": 0.7401,
      "step": 2985
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.241326486596136e-05,
      "loss": 0.8096,
      "step": 2986
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.235501889093845e-05,
      "loss": 0.8448,
      "step": 2987
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.229677552467815e-05,
      "loss": 0.8177,
      "step": 2988
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.223853478705529e-05,
      "loss": 0.8247,
      "step": 2989
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.218029669794383e-05,
      "loss": 0.8138,
      "step": 2990
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.212206127721693e-05,
      "loss": 0.8149,
      "step": 2991
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.206382854474672e-05,
      "loss": 0.7995,
      "step": 2992
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.200559852040442e-05,
      "loss": 0.7848,
      "step": 2993
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.194737122406045e-05,
      "loss": 0.8441,
      "step": 2994
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.188914667558415e-05,
      "loss": 0.8315,
      "step": 2995
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.183092489484395e-05,
      "loss": 0.7642,
      "step": 2996
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.177270590170745e-05,
      "loss": 0.7958,
      "step": 2997
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.171448971604118e-05,
      "loss": 0.8314,
      "step": 2998
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.165627635771071e-05,
      "loss": 0.7934,
      "step": 2999
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.159806584658076e-05,
      "loss": 0.7792,
      "step": 3000
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.153985820251493e-05,
      "loss": 0.8165,
      "step": 3001
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.148165344537597e-05,
      "loss": 0.8319,
      "step": 3002
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.142345159502558e-05,
      "loss": 0.8393,
      "step": 3003
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.136525267132443e-05,
      "loss": 0.7898,
      "step": 3004
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.130705669413231e-05,
      "loss": 0.8564,
      "step": 3005
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.124886368330789e-05,
      "loss": 0.818,
      "step": 3006
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.119067365870884e-05,
      "loss": 0.8538,
      "step": 3007
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.113248664019191e-05,
      "loss": 0.803,
      "step": 3008
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.107430264761274e-05,
      "loss": 0.7927,
      "step": 3009
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.10161217008259e-05,
      "loss": 0.817,
      "step": 3010
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.095794381968503e-05,
      "loss": 0.7387,
      "step": 3011
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.089976902404269e-05,
      "loss": 0.815,
      "step": 3012
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.084159733375025e-05,
      "loss": 0.7818,
      "step": 3013
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.078342876865826e-05,
      "loss": 0.3655,
      "step": 3014
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.072526334861604e-05,
      "loss": 0.8598,
      "step": 3015
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.066710109347181e-05,
      "loss": 0.3431,
      "step": 3016
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.060894202307288e-05,
      "loss": 0.7894,
      "step": 3017
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.055078615726534e-05,
      "loss": 0.8235,
      "step": 3018
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.049263351589414e-05,
      "loss": 0.8073,
      "step": 3019
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.04344841188033e-05,
      "loss": 0.8383,
      "step": 3020
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.03763379858356e-05,
      "loss": 0.7806,
      "step": 3021
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.031819513683272e-05,
      "loss": 0.8336,
      "step": 3022
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.026005559163529e-05,
      "loss": 0.8477,
      "step": 3023
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.020191937008271e-05,
      "loss": 0.766,
      "step": 3024
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.014378649201338e-05,
      "loss": 0.8235,
      "step": 3025
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.008565697726445e-05,
      "loss": 0.8069,
      "step": 3026
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.002753084567189e-05,
      "loss": 0.8072,
      "step": 3027
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.996940811707066e-05,
      "loss": 0.853,
      "step": 3028
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.991128881129447e-05,
      "loss": 0.8029,
      "step": 3029
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.98531729481758e-05,
      "loss": 0.9085,
      "step": 3030
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.979506054754612e-05,
      "loss": 0.3452,
      "step": 3031
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.973695162923556e-05,
      "loss": 0.79,
      "step": 3032
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.967884621307312e-05,
      "loss": 0.84,
      "step": 3033
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.962074431888668e-05,
      "loss": 0.8111,
      "step": 3034
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.956264596650278e-05,
      "loss": 0.8708,
      "step": 3035
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.950455117574681e-05,
      "loss": 0.8628,
      "step": 3036
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.944645996644304e-05,
      "loss": 0.8512,
      "step": 3037
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.938837235841436e-05,
      "loss": 0.3436,
      "step": 3038
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.933028837148247e-05,
      "loss": 0.7556,
      "step": 3039
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.927220802546798e-05,
      "loss": 0.8023,
      "step": 3040
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.921413134019008e-05,
      "loss": 0.8327,
      "step": 3041
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.915605833546675e-05,
      "loss": 0.774,
      "step": 3042
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.90979890311148e-05,
      "loss": 0.7262,
      "step": 3043
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.903992344694963e-05,
      "loss": 0.8292,
      "step": 3044
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.898186160278559e-05,
      "loss": 0.8534,
      "step": 3045
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.892380351843552e-05,
      "loss": 0.8147,
      "step": 3046
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.886574921371107e-05,
      "loss": 0.8526,
      "step": 3047
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.880769870842271e-05,
      "loss": 0.8298,
      "step": 3048
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.874965202237943e-05,
      "loss": 0.7531,
      "step": 3049
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.869160917538899e-05,
      "loss": 0.8344,
      "step": 3050
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.863357018725791e-05,
      "loss": 0.7832,
      "step": 3051
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.857553507779129e-05,
      "loss": 0.7537,
      "step": 3052
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.851750386679293e-05,
      "loss": 0.8093,
      "step": 3053
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.845947657406538e-05,
      "loss": 0.7927,
      "step": 3054
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.840145321940975e-05,
      "loss": 0.7803,
      "step": 3055
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.834343382262583e-05,
      "loss": 0.7983,
      "step": 3056
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.828541840351214e-05,
      "loss": 0.8386,
      "step": 3057
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.822740698186575e-05,
      "loss": 0.7878,
      "step": 3058
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.816939957748232e-05,
      "loss": 0.8178,
      "step": 3059
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.811139621015636e-05,
      "loss": 0.8437,
      "step": 3060
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.805339689968076e-05,
      "loss": 0.8133,
      "step": 3061
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.79954016658471e-05,
      "loss": 0.7836,
      "step": 3062
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.793741052844568e-05,
      "loss": 0.8604,
      "step": 3063
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.787942350726524e-05,
      "loss": 0.764,
      "step": 3064
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.782144062209323e-05,
      "loss": 0.8056,
      "step": 3065
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.776346189271565e-05,
      "loss": 0.7636,
      "step": 3066
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.770548733891703e-05,
      "loss": 0.8125,
      "step": 3067
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.76475169804806e-05,
      "loss": 0.7752,
      "step": 3068
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.758955083718804e-05,
      "loss": 0.8324,
      "step": 3069
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.75315889288196e-05,
      "loss": 0.8292,
      "step": 3070
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.74736312751542e-05,
      "loss": 0.8331,
      "step": 3071
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.741567789596918e-05,
      "loss": 0.7534,
      "step": 3072
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.735772881104045e-05,
      "loss": 0.8163,
      "step": 3073
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.729978404014252e-05,
      "loss": 0.8149,
      "step": 3074
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.724184360304838e-05,
      "loss": 0.8746,
      "step": 3075
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.718390751952948e-05,
      "loss": 0.8101,
      "step": 3076
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.712597580935592e-05,
      "loss": 0.8278,
      "step": 3077
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.706804849229621e-05,
      "loss": 0.822,
      "step": 3078
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.701012558811736e-05,
      "loss": 0.3602,
      "step": 3079
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.695220711658495e-05,
      "loss": 0.8358,
      "step": 3080
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.689429309746295e-05,
      "loss": 0.8306,
      "step": 3081
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.683638355051389e-05,
      "loss": 0.8559,
      "step": 3082
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.677847849549873e-05,
      "loss": 0.8594,
      "step": 3083
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.672057795217692e-05,
      "loss": 0.8416,
      "step": 3084
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.666268194030634e-05,
      "loss": 0.8107,
      "step": 3085
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.660479047964336e-05,
      "loss": 0.8113,
      "step": 3086
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.654690358994275e-05,
      "loss": 0.8387,
      "step": 3087
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.64890212909578e-05,
      "loss": 0.809,
      "step": 3088
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.643114360244016e-05,
      "loss": 0.776,
      "step": 3089
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.637327054413989e-05,
      "loss": 0.8408,
      "step": 3090
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.631540213580557e-05,
      "loss": 0.806,
      "step": 3091
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.625753839718415e-05,
      "loss": 0.8314,
      "step": 3092
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.619967934802086e-05,
      "loss": 0.3308,
      "step": 3093
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.614182500805959e-05,
      "loss": 0.8543,
      "step": 3094
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.608397539704239e-05,
      "loss": 0.8375,
      "step": 3095
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.602613053470973e-05,
      "loss": 0.8149,
      "step": 3096
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.596829044080062e-05,
      "loss": 0.868,
      "step": 3097
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.591045513505224e-05,
      "loss": 0.8913,
      "step": 3098
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.585262463720027e-05,
      "loss": 0.8038,
      "step": 3099
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.579479896697874e-05,
      "loss": 0.8598,
      "step": 3100
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.573697814411993e-05,
      "loss": 0.7935,
      "step": 3101
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.567916218835459e-05,
      "loss": 0.7799,
      "step": 3102
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.562135111941174e-05,
      "loss": 0.7932,
      "step": 3103
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.556354495701873e-05,
      "loss": 0.8222,
      "step": 3104
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.550574372090126e-05,
      "loss": 0.8267,
      "step": 3105
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.544794743078337e-05,
      "loss": 0.8306,
      "step": 3106
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.539015610638735e-05,
      "loss": 0.7909,
      "step": 3107
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.533236976743382e-05,
      "loss": 0.784,
      "step": 3108
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.527458843364178e-05,
      "loss": 0.7808,
      "step": 3109
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.521681212472835e-05,
      "loss": 0.8173,
      "step": 3110
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.515904086040913e-05,
      "loss": 0.7571,
      "step": 3111
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.510127466039787e-05,
      "loss": 0.7799,
      "step": 3112
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.504351354440656e-05,
      "loss": 0.7777,
      "step": 3113
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.498575753214564e-05,
      "loss": 0.8054,
      "step": 3114
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.492800664332364e-05,
      "loss": 0.8471,
      "step": 3115
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.487026089764731e-05,
      "loss": 0.8012,
      "step": 3116
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.481252031482188e-05,
      "loss": 0.8016,
      "step": 3117
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.475478491455053e-05,
      "loss": 0.8855,
      "step": 3118
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.469705471653488e-05,
      "loss": 0.8142,
      "step": 3119
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.46393297404747e-05,
      "loss": 0.7938,
      "step": 3120
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.458161000606795e-05,
      "loss": 0.9042,
      "step": 3121
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.452389553301083e-05,
      "loss": 0.8207,
      "step": 3122
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.44661863409978e-05,
      "loss": 0.8821,
      "step": 3123
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.440848244972138e-05,
      "loss": 0.8457,
      "step": 3124
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.43507838788724e-05,
      "loss": 0.7817,
      "step": 3125
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.429309064813989e-05,
      "loss": 0.8408,
      "step": 3126
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.423540277721093e-05,
      "loss": 0.8569,
      "step": 3127
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.417772028577088e-05,
      "loss": 0.7796,
      "step": 3128
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.412004319350325e-05,
      "loss": 0.8127,
      "step": 3129
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.406237152008966e-05,
      "loss": 0.8062,
      "step": 3130
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.40047052852099e-05,
      "loss": 0.8244,
      "step": 3131
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.394704450854197e-05,
      "loss": 0.842,
      "step": 3132
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.388938920976188e-05,
      "loss": 0.7891,
      "step": 3133
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.383173940854392e-05,
      "loss": 0.8756,
      "step": 3134
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.377409512456036e-05,
      "loss": 0.8006,
      "step": 3135
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.371645637748169e-05,
      "loss": 0.7882,
      "step": 3136
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.365882318697649e-05,
      "loss": 0.8314,
      "step": 3137
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.360119557271137e-05,
      "loss": 0.8302,
      "step": 3138
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.354357355435114e-05,
      "loss": 0.8151,
      "step": 3139
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.348595715155869e-05,
      "loss": 0.82,
      "step": 3140
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.342834638399487e-05,
      "loss": 0.862,
      "step": 3141
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.337074127131874e-05,
      "loss": 0.8423,
      "step": 3142
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.331314183318743e-05,
      "loss": 0.8174,
      "step": 3143
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.325554808925602e-05,
      "loss": 0.8222,
      "step": 3144
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.319796005917774e-05,
      "loss": 0.8267,
      "step": 3145
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.31403777626039e-05,
      "loss": 0.8003,
      "step": 3146
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.308280121918372e-05,
      "loss": 0.3494,
      "step": 3147
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.302523044856461e-05,
      "loss": 0.8036,
      "step": 3148
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.296766547039189e-05,
      "loss": 0.7923,
      "step": 3149
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.291010630430897e-05,
      "loss": 0.7227,
      "step": 3150
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.285255296995727e-05,
      "loss": 0.8373,
      "step": 3151
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.279500548697617e-05,
      "loss": 0.8026,
      "step": 3152
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.273746387500315e-05,
      "loss": 0.7855,
      "step": 3153
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.267992815367359e-05,
      "loss": 0.8218,
      "step": 3154
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.262239834262087e-05,
      "loss": 0.8011,
      "step": 3155
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.256487446147643e-05,
      "loss": 0.8624,
      "step": 3156
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.250735652986966e-05,
      "loss": 0.7967,
      "step": 3157
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.244984456742783e-05,
      "loss": 0.3762,
      "step": 3158
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.239233859377629e-05,
      "loss": 0.7831,
      "step": 3159
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.233483862853828e-05,
      "loss": 0.7614,
      "step": 3160
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.227734469133499e-05,
      "loss": 0.8369,
      "step": 3161
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.221985680178559e-05,
      "loss": 0.8308,
      "step": 3162
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.216237497950719e-05,
      "loss": 0.7889,
      "step": 3163
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.210489924411476e-05,
      "loss": 0.7853,
      "step": 3164
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.204742961522126e-05,
      "loss": 0.8194,
      "step": 3165
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.198996611243755e-05,
      "loss": 0.8402,
      "step": 3166
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.193250875537237e-05,
      "loss": 0.8763,
      "step": 3167
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.187505756363241e-05,
      "loss": 0.8307,
      "step": 3168
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.181761255682225e-05,
      "loss": 0.8541,
      "step": 3169
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.176017375454431e-05,
      "loss": 0.7953,
      "step": 3170
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.170274117639894e-05,
      "loss": 0.3316,
      "step": 3171
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.164531484198432e-05,
      "loss": 0.7779,
      "step": 3172
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.158789477089661e-05,
      "loss": 0.8657,
      "step": 3173
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.153048098272968e-05,
      "loss": 0.8205,
      "step": 3174
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.147307349707535e-05,
      "loss": 0.3533,
      "step": 3175
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.141567233352328e-05,
      "loss": 0.8331,
      "step": 3176
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.135827751166098e-05,
      "loss": 0.8508,
      "step": 3177
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.130088905107372e-05,
      "loss": 0.8117,
      "step": 3178
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.12435069713447e-05,
      "loss": 0.8716,
      "step": 3179
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.118613129205492e-05,
      "loss": 0.7466,
      "step": 3180
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.112876203278312e-05,
      "loss": 0.8288,
      "step": 3181
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.107139921310594e-05,
      "loss": 0.8041,
      "step": 3182
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.10140428525978e-05,
      "loss": 0.7648,
      "step": 3183
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.095669297083085e-05,
      "loss": 0.8381,
      "step": 3184
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.089934958737513e-05,
      "loss": 0.8589,
      "step": 3185
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.084201272179841e-05,
      "loss": 0.8485,
      "step": 3186
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.078468239366624e-05,
      "loss": 0.8305,
      "step": 3187
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.07273586225419e-05,
      "loss": 0.7954,
      "step": 3188
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.067004142798651e-05,
      "loss": 0.7885,
      "step": 3189
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.061273082955895e-05,
      "loss": 0.7894,
      "step": 3190
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.055542684681572e-05,
      "loss": 0.8281,
      "step": 3191
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.049812949931122e-05,
      "loss": 0.8227,
      "step": 3192
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.044083880659749e-05,
      "loss": 0.771,
      "step": 3193
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.03835547882243e-05,
      "loss": 0.8326,
      "step": 3194
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.03262774637392e-05,
      "loss": 0.8042,
      "step": 3195
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.026900685268742e-05,
      "loss": 0.8303,
      "step": 3196
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.02117429746119e-05,
      "loss": 0.8353,
      "step": 3197
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.015448584905327e-05,
      "loss": 0.7918,
      "step": 3198
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.009723549554988e-05,
      "loss": 0.7591,
      "step": 3199
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.003999193363778e-05,
      "loss": 0.783,
      "step": 3200
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.998275518285066e-05,
      "loss": 0.7776,
      "step": 3201
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.992552526271991e-05,
      "loss": 0.7994,
      "step": 3202
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.986830219277459e-05,
      "loss": 0.8219,
      "step": 3203
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.981108599254141e-05,
      "loss": 0.7735,
      "step": 3204
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.975387668154478e-05,
      "loss": 0.3318,
      "step": 3205
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.969667427930667e-05,
      "loss": 0.871,
      "step": 3206
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.963947880534681e-05,
      "loss": 0.8097,
      "step": 3207
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.958229027918244e-05,
      "loss": 0.34,
      "step": 3208
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.952510872032854e-05,
      "loss": 0.8695,
      "step": 3209
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.946793414829765e-05,
      "loss": 0.7857,
      "step": 3210
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.941076658259993e-05,
      "loss": 0.7664,
      "step": 3211
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.935360604274315e-05,
      "loss": 0.7779,
      "step": 3212
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.929645254823274e-05,
      "loss": 0.8141,
      "step": 3213
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.923930611857162e-05,
      "loss": 0.8273,
      "step": 3214
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.918216677326038e-05,
      "loss": 0.8007,
      "step": 3215
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.91250345317972e-05,
      "loss": 0.754,
      "step": 3216
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.906790941367775e-05,
      "loss": 0.789,
      "step": 3217
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.901079143839536e-05,
      "loss": 0.8163,
      "step": 3218
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.895368062544091e-05,
      "loss": 0.8055,
      "step": 3219
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.889657699430274e-05,
      "loss": 0.8618,
      "step": 3220
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.883948056446688e-05,
      "loss": 0.7935,
      "step": 3221
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.878239135541679e-05,
      "loss": 0.7877,
      "step": 3222
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.872530938663356e-05,
      "loss": 0.8135,
      "step": 3223
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.86682346775957e-05,
      "loss": 0.7963,
      "step": 3224
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.861116724777934e-05,
      "loss": 0.8147,
      "step": 3225
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.85541071166581e-05,
      "loss": 0.8321,
      "step": 3226
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.849705430370309e-05,
      "loss": 0.8138,
      "step": 3227
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.844000882838292e-05,
      "loss": 0.8024,
      "step": 3228
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.83829707101637e-05,
      "loss": 0.808,
      "step": 3229
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.832593996850907e-05,
      "loss": 0.8204,
      "step": 3230
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.826891662288007e-05,
      "loss": 0.7913,
      "step": 3231
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.821190069273531e-05,
      "loss": 0.8431,
      "step": 3232
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.815489219753082e-05,
      "loss": 0.8484,
      "step": 3233
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.809789115672007e-05,
      "loss": 0.8414,
      "step": 3234
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.804089758975403e-05,
      "loss": 0.8209,
      "step": 3235
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.798391151608115e-05,
      "loss": 0.8293,
      "step": 3236
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.792693295514719e-05,
      "loss": 0.752,
      "step": 3237
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.786996192639548e-05,
      "loss": 0.8077,
      "step": 3238
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.781299844926677e-05,
      "loss": 0.335,
      "step": 3239
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.775604254319913e-05,
      "loss": 0.8383,
      "step": 3240
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.769909422762814e-05,
      "loss": 0.7493,
      "step": 3241
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.764215352198678e-05,
      "loss": 0.836,
      "step": 3242
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.758522044570537e-05,
      "loss": 0.8495,
      "step": 3243
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.752829501821174e-05,
      "loss": 0.8597,
      "step": 3244
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.7471377258931e-05,
      "loss": 0.8196,
      "step": 3245
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.741446718728567e-05,
      "loss": 0.7351,
      "step": 3246
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.735756482269572e-05,
      "loss": 0.8512,
      "step": 3247
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.730067018457837e-05,
      "loss": 0.7949,
      "step": 3248
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.724378329234829e-05,
      "loss": 0.7859,
      "step": 3249
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.718690416541752e-05,
      "loss": 0.8294,
      "step": 3250
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.713003282319535e-05,
      "loss": 0.8272,
      "step": 3251
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.70731692850885e-05,
      "loss": 0.3092,
      "step": 3252
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.701631357050104e-05,
      "loss": 0.8077,
      "step": 3253
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.695946569883428e-05,
      "loss": 0.8175,
      "step": 3254
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.690262568948692e-05,
      "loss": 0.796,
      "step": 3255
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.684579356185501e-05,
      "loss": 0.8472,
      "step": 3256
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.67889693353318e-05,
      "loss": 0.835,
      "step": 3257
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.673215302930794e-05,
      "loss": 0.7968,
      "step": 3258
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.667534466317137e-05,
      "loss": 0.7801,
      "step": 3259
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.661854425630721e-05,
      "loss": 0.8297,
      "step": 3260
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.656175182809808e-05,
      "loss": 0.7731,
      "step": 3261
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.650496739792366e-05,
      "loss": 0.8043,
      "step": 3262
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.644819098516098e-05,
      "loss": 0.7409,
      "step": 3263
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.639142260918442e-05,
      "loss": 0.8212,
      "step": 3264
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.633466228936546e-05,
      "loss": 0.8895,
      "step": 3265
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.627791004507294e-05,
      "loss": 0.8575,
      "step": 3266
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.622116589567296e-05,
      "loss": 0.8143,
      "step": 3267
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.616442986052873e-05,
      "loss": 0.8033,
      "step": 3268
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.610770195900084e-05,
      "loss": 0.7794,
      "step": 3269
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.605098221044701e-05,
      "loss": 0.7473,
      "step": 3270
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.599427063422222e-05,
      "loss": 0.7794,
      "step": 3271
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.593756724967863e-05,
      "loss": 0.333,
      "step": 3272
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.588087207616566e-05,
      "loss": 0.8276,
      "step": 3273
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.582418513302985e-05,
      "loss": 0.7975,
      "step": 3274
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.576750643961497e-05,
      "loss": 0.866,
      "step": 3275
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.571083601526203e-05,
      "loss": 0.8686,
      "step": 3276
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.565417387930911e-05,
      "loss": 0.8325,
      "step": 3277
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.559752005109153e-05,
      "loss": 0.8318,
      "step": 3278
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.554087454994181e-05,
      "loss": 0.8129,
      "step": 3279
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.548423739518948e-05,
      "loss": 0.8255,
      "step": 3280
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.542760860616143e-05,
      "loss": 0.7406,
      "step": 3281
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.537098820218156e-05,
      "loss": 0.7939,
      "step": 3282
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.531437620257086e-05,
      "loss": 0.8329,
      "step": 3283
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.525777262664764e-05,
      "loss": 0.7994,
      "step": 3284
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.520117749372717e-05,
      "loss": 0.7748,
      "step": 3285
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.514459082312186e-05,
      "loss": 0.852,
      "step": 3286
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.508801263414134e-05,
      "loss": 0.8379,
      "step": 3287
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.50314429460922e-05,
      "loss": 0.7587,
      "step": 3288
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.497488177827825e-05,
      "loss": 0.8447,
      "step": 3289
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.491832915000035e-05,
      "loss": 0.8504,
      "step": 3290
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.486178508055636e-05,
      "loss": 0.7393,
      "step": 3291
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.48052495892414e-05,
      "loss": 0.7515,
      "step": 3292
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.474872269534752e-05,
      "loss": 0.7508,
      "step": 3293
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.469220441816384e-05,
      "loss": 0.8287,
      "step": 3294
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.463569477697661e-05,
      "loss": 0.7687,
      "step": 3295
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.457919379106915e-05,
      "loss": 0.8551,
      "step": 3296
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.452270147972166e-05,
      "loss": 0.8059,
      "step": 3297
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.44662178622116e-05,
      "loss": 0.8056,
      "step": 3298
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.440974295781334e-05,
      "loss": 0.8436,
      "step": 3299
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.435327678579824e-05,
      "loss": 0.3321,
      "step": 3300
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.429681936543482e-05,
      "loss": 0.7795,
      "step": 3301
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.424037071598846e-05,
      "loss": 0.7612,
      "step": 3302
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.418393085672164e-05,
      "loss": 0.8113,
      "step": 3303
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.412749980689385e-05,
      "loss": 0.8719,
      "step": 3304
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.407107758576151e-05,
      "loss": 0.8153,
      "step": 3305
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.4014664212578e-05,
      "loss": 0.8052,
      "step": 3306
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.395825970659389e-05,
      "loss": 0.8202,
      "step": 3307
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.390186408705645e-05,
      "loss": 0.7901,
      "step": 3308
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.384547737321008e-05,
      "loss": 0.8623,
      "step": 3309
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.378909958429611e-05,
      "loss": 0.7985,
      "step": 3310
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.37327307395528e-05,
      "loss": 0.8376,
      "step": 3311
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.36763708582154e-05,
      "loss": 0.819,
      "step": 3312
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.362001995951604e-05,
      "loss": 0.836,
      "step": 3313
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.356367806268383e-05,
      "loss": 0.8418,
      "step": 3314
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.350734518694485e-05,
      "loss": 0.8104,
      "step": 3315
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.345102135152202e-05,
      "loss": 0.8197,
      "step": 3316
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.339470657563514e-05,
      "loss": 0.7922,
      "step": 3317
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.333840087850108e-05,
      "loss": 0.7943,
      "step": 3318
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.328210427933351e-05,
      "loss": 0.7776,
      "step": 3319
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.32258167973429e-05,
      "loss": 0.7767,
      "step": 3320
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.316953845173684e-05,
      "loss": 0.7986,
      "step": 3321
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.311326926171961e-05,
      "loss": 0.7706,
      "step": 3322
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.30570092464924e-05,
      "loss": 0.7839,
      "step": 3323
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.300075842525336e-05,
      "loss": 0.7997,
      "step": 3324
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.294451681719743e-05,
      "loss": 0.8074,
      "step": 3325
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.288828444151634e-05,
      "loss": 0.8365,
      "step": 3326
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.283206131739885e-05,
      "loss": 0.8487,
      "step": 3327
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.277584746403044e-05,
      "loss": 0.3191,
      "step": 3328
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.271964290059336e-05,
      "loss": 0.3755,
      "step": 3329
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.266344764626689e-05,
      "loss": 0.836,
      "step": 3330
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.260726172022696e-05,
      "loss": 0.7741,
      "step": 3331
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.255108514164637e-05,
      "loss": 0.7994,
      "step": 3332
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.249491792969477e-05,
      "loss": 0.7498,
      "step": 3333
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.243876010353851e-05,
      "loss": 0.8078,
      "step": 3334
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.238261168234091e-05,
      "loss": 0.7987,
      "step": 3335
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.232647268526194e-05,
      "loss": 0.899,
      "step": 3336
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.227034313145832e-05,
      "loss": 0.8414,
      "step": 3337
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.221422304008371e-05,
      "loss": 0.3356,
      "step": 3338
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.215811243028845e-05,
      "loss": 0.7729,
      "step": 3339
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.210201132121954e-05,
      "loss": 0.769,
      "step": 3340
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.204591973202098e-05,
      "loss": 0.7903,
      "step": 3341
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.198983768183331e-05,
      "loss": 0.8134,
      "step": 3342
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.193376518979387e-05,
      "loss": 0.8203,
      "step": 3343
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.187770227503683e-05,
      "loss": 0.8376,
      "step": 3344
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.182164895669296e-05,
      "loss": 0.8371,
      "step": 3345
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.176560525388983e-05,
      "loss": 0.8438,
      "step": 3346
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.170957118575173e-05,
      "loss": 0.8388,
      "step": 3347
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.165354677139965e-05,
      "loss": 0.8026,
      "step": 3348
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.159753202995122e-05,
      "loss": 0.7735,
      "step": 3349
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.154152698052092e-05,
      "loss": 0.827,
      "step": 3350
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.14855316422198e-05,
      "loss": 0.8924,
      "step": 3351
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.142954603415556e-05,
      "loss": 0.8093,
      "step": 3352
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.137357017543278e-05,
      "loss": 0.8247,
      "step": 3353
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.131760408515247e-05,
      "loss": 0.8025,
      "step": 3354
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.126164778241252e-05,
      "loss": 0.7206,
      "step": 3355
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.120570128630728e-05,
      "loss": 0.7791,
      "step": 3356
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.11497646159279e-05,
      "loss": 0.7258,
      "step": 3357
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.109383779036216e-05,
      "loss": 0.3434,
      "step": 3358
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.10379208286944e-05,
      "loss": 0.7894,
      "step": 3359
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.098201375000564e-05,
      "loss": 0.8764,
      "step": 3360
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.09261165733736e-05,
      "loss": 0.7715,
      "step": 3361
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.087022931787252e-05,
      "loss": 0.7894,
      "step": 3362
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.081435200257322e-05,
      "loss": 0.7546,
      "step": 3363
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.07584846465433e-05,
      "loss": 0.7968,
      "step": 3364
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.070262726884684e-05,
      "loss": 0.7799,
      "step": 3365
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.064677988854445e-05,
      "loss": 0.8156,
      "step": 3366
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.059094252469349e-05,
      "loss": 0.8354,
      "step": 3367
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.053511519634782e-05,
      "loss": 0.7681,
      "step": 3368
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.047929792255781e-05,
      "loss": 0.3577,
      "step": 3369
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.042349072237058e-05,
      "loss": 0.768,
      "step": 3370
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.036769361482962e-05,
      "loss": 0.8239,
      "step": 3371
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.031190661897501e-05,
      "loss": 0.852,
      "step": 3372
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.025612975384357e-05,
      "loss": 0.8477,
      "step": 3373
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.020036303846839e-05,
      "loss": 0.3203,
      "step": 3374
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.01446064918793e-05,
      "loss": 0.8017,
      "step": 3375
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.008886013310256e-05,
      "loss": 0.8506,
      "step": 3376
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.003312398116095e-05,
      "loss": 0.7938,
      "step": 3377
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.997739805507386e-05,
      "loss": 0.7699,
      "step": 3378
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.992168237385708e-05,
      "loss": 0.7998,
      "step": 3379
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.986597695652293e-05,
      "loss": 0.7993,
      "step": 3380
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.981028182208033e-05,
      "loss": 0.817,
      "step": 3381
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.975459698953453e-05,
      "loss": 0.8671,
      "step": 3382
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.969892247788735e-05,
      "loss": 0.8336,
      "step": 3383
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.964325830613714e-05,
      "loss": 0.801,
      "step": 3384
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.958760449327863e-05,
      "loss": 0.7904,
      "step": 3385
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.9531961058303e-05,
      "loss": 0.7534,
      "step": 3386
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.947632802019801e-05,
      "loss": 0.7928,
      "step": 3387
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.94207053979478e-05,
      "loss": 0.7871,
      "step": 3388
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.936509321053286e-05,
      "loss": 0.776,
      "step": 3389
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.930949147693033e-05,
      "loss": 0.837,
      "step": 3390
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.925390021611362e-05,
      "loss": 0.8314,
      "step": 3391
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.919831944705257e-05,
      "loss": 0.8197,
      "step": 3392
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.914274918871357e-05,
      "loss": 0.8172,
      "step": 3393
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.908718946005929e-05,
      "loss": 0.8131,
      "step": 3394
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.903164028004881e-05,
      "loss": 0.8197,
      "step": 3395
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.897610166763775e-05,
      "loss": 0.823,
      "step": 3396
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.892057364177792e-05,
      "loss": 0.7668,
      "step": 3397
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.886505622141772e-05,
      "loss": 0.7769,
      "step": 3398
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.880954942550182e-05,
      "loss": 0.7816,
      "step": 3399
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.87540532729712e-05,
      "loss": 0.8848,
      "step": 3400
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.869856778276339e-05,
      "loss": 0.7735,
      "step": 3401
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.864309297381213e-05,
      "loss": 0.8381,
      "step": 3402
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.858762886504755e-05,
      "loss": 0.8113,
      "step": 3403
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.85321754753962e-05,
      "loss": 0.8203,
      "step": 3404
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.847673282378089e-05,
      "loss": 0.7874,
      "step": 3405
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.842130092912073e-05,
      "loss": 0.8231,
      "step": 3406
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.836587981033133e-05,
      "loss": 0.7256,
      "step": 3407
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.831046948632447e-05,
      "loss": 0.8271,
      "step": 3408
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.825506997600821e-05,
      "loss": 0.8023,
      "step": 3409
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.819968129828714e-05,
      "loss": 0.8284,
      "step": 3410
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.814430347206195e-05,
      "loss": 0.7473,
      "step": 3411
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.808893651622963e-05,
      "loss": 0.8279,
      "step": 3412
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.803358044968364e-05,
      "loss": 0.7647,
      "step": 3413
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.797823529131352e-05,
      "loss": 0.7813,
      "step": 3414
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.792290106000518e-05,
      "loss": 0.8281,
      "step": 3415
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.786757777464085e-05,
      "loss": 0.3794,
      "step": 3416
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.781226545409895e-05,
      "loss": 0.3367,
      "step": 3417
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.775696411725412e-05,
      "loss": 0.8453,
      "step": 3418
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.770167378297738e-05,
      "loss": 0.8305,
      "step": 3419
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.764639447013586e-05,
      "loss": 0.8036,
      "step": 3420
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.759112619759308e-05,
      "loss": 0.3485,
      "step": 3421
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.753586898420866e-05,
      "loss": 0.8094,
      "step": 3422
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.748062284883845e-05,
      "loss": 0.8788,
      "step": 3423
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.742538781033465e-05,
      "loss": 0.8627,
      "step": 3424
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.737016388754556e-05,
      "loss": 0.7927,
      "step": 3425
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.731495109931564e-05,
      "loss": 0.8653,
      "step": 3426
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.725974946448573e-05,
      "loss": 0.7916,
      "step": 3427
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.72045590018927e-05,
      "loss": 0.7863,
      "step": 3428
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.714937973036963e-05,
      "loss": 0.8271,
      "step": 3429
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.709421166874592e-05,
      "loss": 0.8242,
      "step": 3430
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.703905483584698e-05,
      "loss": 0.7752,
      "step": 3431
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.698390925049441e-05,
      "loss": 0.8155,
      "step": 3432
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.692877493150611e-05,
      "loss": 0.7967,
      "step": 3433
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.687365189769596e-05,
      "loss": 0.8261,
      "step": 3434
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.681854016787405e-05,
      "loss": 0.7959,
      "step": 3435
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.676343976084673e-05,
      "loss": 0.3472,
      "step": 3436
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.67083506954163e-05,
      "loss": 0.794,
      "step": 3437
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.665327299038126e-05,
      "loss": 0.7612,
      "step": 3438
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.659820666453634e-05,
      "loss": 0.7705,
      "step": 3439
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.654315173667225e-05,
      "loss": 0.7865,
      "step": 3440
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.648810822557579e-05,
      "loss": 0.8154,
      "step": 3441
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.643307615003003e-05,
      "loss": 0.8019,
      "step": 3442
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.637805552881395e-05,
      "loss": 0.7911,
      "step": 3443
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.63230463807028e-05,
      "loss": 0.7927,
      "step": 3444
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.626804872446776e-05,
      "loss": 0.3643,
      "step": 3445
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.621306257887616e-05,
      "loss": 0.7795,
      "step": 3446
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.615808796269142e-05,
      "loss": 0.7943,
      "step": 3447
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.610312489467301e-05,
      "loss": 0.8309,
      "step": 3448
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.604817339357638e-05,
      "loss": 0.8155,
      "step": 3449
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.599323347815318e-05,
      "loss": 0.3235,
      "step": 3450
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.593830516715101e-05,
      "loss": 0.7967,
      "step": 3451
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.588338847931346e-05,
      "loss": 0.8379,
      "step": 3452
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.582848343338034e-05,
      "loss": 0.3498,
      "step": 3453
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.57735900480873e-05,
      "loss": 0.8231,
      "step": 3454
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.571870834216605e-05,
      "loss": 0.8309,
      "step": 3455
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.566383833434443e-05,
      "loss": 0.7941,
      "step": 3456
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.560898004334617e-05,
      "loss": 0.3413,
      "step": 3457
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.555413348789096e-05,
      "loss": 0.8159,
      "step": 3458
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.549929868669471e-05,
      "loss": 0.7677,
      "step": 3459
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.544447565846907e-05,
      "loss": 0.8299,
      "step": 3460
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.538966442192173e-05,
      "loss": 0.8296,
      "step": 3461
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.533486499575652e-05,
      "loss": 0.8021,
      "step": 3462
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.528007739867304e-05,
      "loss": 0.7964,
      "step": 3463
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.522530164936697e-05,
      "loss": 0.834,
      "step": 3464
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.517053776652988e-05,
      "loss": 0.8276,
      "step": 3465
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.511578576884931e-05,
      "loss": 0.7965,
      "step": 3466
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.506104567500881e-05,
      "loss": 0.8577,
      "step": 3467
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.500631750368779e-05,
      "loss": 0.326,
      "step": 3468
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.495160127356155e-05,
      "loss": 0.3506,
      "step": 3469
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.489689700330147e-05,
      "loss": 0.8074,
      "step": 3470
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.484220471157472e-05,
      "loss": 0.3382,
      "step": 3471
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.478752441704439e-05,
      "loss": 0.7683,
      "step": 3472
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.473285613836957e-05,
      "loss": 0.8114,
      "step": 3473
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.467819989420514e-05,
      "loss": 0.8548,
      "step": 3474
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.462355570320191e-05,
      "loss": 0.8442,
      "step": 3475
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.456892358400663e-05,
      "loss": 0.845,
      "step": 3476
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.451430355526189e-05,
      "loss": 0.8097,
      "step": 3477
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.445969563560607e-05,
      "loss": 0.7977,
      "step": 3478
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.440509984367362e-05,
      "loss": 0.8669,
      "step": 3479
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.435051619809465e-05,
      "loss": 0.8307,
      "step": 3480
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.429594471749518e-05,
      "loss": 0.7911,
      "step": 3481
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.424138542049719e-05,
      "loss": 0.7951,
      "step": 3482
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.418683832571839e-05,
      "loss": 0.3637,
      "step": 3483
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.413230345177228e-05,
      "loss": 0.822,
      "step": 3484
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.407778081726839e-05,
      "loss": 0.718,
      "step": 3485
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.402327044081183e-05,
      "loss": 0.7959,
      "step": 3486
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.396877234100373e-05,
      "loss": 0.7783,
      "step": 3487
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.391428653644091e-05,
      "loss": 0.7443,
      "step": 3488
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.385981304571598e-05,
      "loss": 0.7784,
      "step": 3489
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.38053518874175e-05,
      "loss": 0.8647,
      "step": 3490
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.375090308012967e-05,
      "loss": 0.7853,
      "step": 3491
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.369646664243247e-05,
      "loss": 0.8372,
      "step": 3492
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.364204259290178e-05,
      "loss": 0.7877,
      "step": 3493
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.35876309501092e-05,
      "loss": 0.819,
      "step": 3494
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.3533231732622e-05,
      "loss": 0.8089,
      "step": 3495
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.347884495900338e-05,
      "loss": 0.8067,
      "step": 3496
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.342447064781217e-05,
      "loss": 0.8165,
      "step": 3497
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.337010881760293e-05,
      "loss": 0.8682,
      "step": 3498
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.331575948692612e-05,
      "loss": 0.8541,
      "step": 3499
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.326142267432776e-05,
      "loss": 0.8681,
      "step": 3500
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.320709839834966e-05,
      "loss": 0.7522,
      "step": 3501
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.31527866775294e-05,
      "loss": 0.7902,
      "step": 3502
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.309848753040023e-05,
      "loss": 0.8636,
      "step": 3503
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.304420097549105e-05,
      "loss": 0.7446,
      "step": 3504
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.298992703132662e-05,
      "loss": 0.6889,
      "step": 3505
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.293566571642728e-05,
      "loss": 0.7793,
      "step": 3506
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.2881417049309e-05,
      "loss": 0.8537,
      "step": 3507
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.282718104848365e-05,
      "loss": 0.843,
      "step": 3508
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.277295773245851e-05,
      "loss": 0.836,
      "step": 3509
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.27187471197368e-05,
      "loss": 0.812,
      "step": 3510
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.26645492288172e-05,
      "loss": 0.7543,
      "step": 3511
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.26103640781941e-05,
      "loss": 0.7545,
      "step": 3512
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.255619168635764e-05,
      "loss": 0.7688,
      "step": 3513
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.25020320717935e-05,
      "loss": 0.862,
      "step": 3514
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.244788525298296e-05,
      "loss": 0.828,
      "step": 3515
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.239375124840311e-05,
      "loss": 0.7973,
      "step": 3516
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.233963007652652e-05,
      "loss": 0.7701,
      "step": 3517
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.228552175582136e-05,
      "loss": 0.3244,
      "step": 3518
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.223142630475158e-05,
      "loss": 0.8226,
      "step": 3519
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.217734374177659e-05,
      "loss": 0.7855,
      "step": 3520
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.212327408535142e-05,
      "loss": 0.3452,
      "step": 3521
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.206921735392679e-05,
      "loss": 0.7965,
      "step": 3522
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.201517356594888e-05,
      "loss": 0.817,
      "step": 3523
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.196114273985951e-05,
      "loss": 0.8337,
      "step": 3524
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.190712489409615e-05,
      "loss": 0.791,
      "step": 3525
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.185312004709174e-05,
      "loss": 0.8368,
      "step": 3526
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.179912821727477e-05,
      "loss": 0.794,
      "step": 3527
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.174514942306942e-05,
      "loss": 0.8726,
      "step": 3528
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.169118368289526e-05,
      "loss": 0.3345,
      "step": 3529
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.163723101516755e-05,
      "loss": 0.8153,
      "step": 3530
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.1583291438297e-05,
      "loss": 0.7038,
      "step": 3531
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.152936497068982e-05,
      "loss": 0.8177,
      "step": 3532
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.147545163074789e-05,
      "loss": 0.7404,
      "step": 3533
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.142155143686848e-05,
      "loss": 0.3331,
      "step": 3534
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.136766440744437e-05,
      "loss": 0.8419,
      "step": 3535
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.131379056086397e-05,
      "loss": 0.3666,
      "step": 3536
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.125992991551111e-05,
      "loss": 0.8099,
      "step": 3537
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.120608248976504e-05,
      "loss": 0.7712,
      "step": 3538
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.115224830200068e-05,
      "loss": 0.7936,
      "step": 3539
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.109842737058829e-05,
      "loss": 0.8348,
      "step": 3540
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.104461971389361e-05,
      "loss": 0.7505,
      "step": 3541
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.099082535027796e-05,
      "loss": 0.8438,
      "step": 3542
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.093704429809801e-05,
      "loss": 0.8049,
      "step": 3543
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0883276575705914e-05,
      "loss": 0.7807,
      "step": 3544
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.082952220144935e-05,
      "loss": 0.8789,
      "step": 3545
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0775781193671356e-05,
      "loss": 0.8467,
      "step": 3546
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0722053570710394e-05,
      "loss": 0.8092,
      "step": 3547
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0668339350900485e-05,
      "loss": 0.7851,
      "step": 3548
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.061463855257095e-05,
      "loss": 0.8145,
      "step": 3549
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0560951194046525e-05,
      "loss": 0.7909,
      "step": 3550
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.050727729364751e-05,
      "loss": 0.8041,
      "step": 3551
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.045361686968942e-05,
      "loss": 0.7992,
      "step": 3552
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0399969940483334e-05,
      "loss": 0.7796,
      "step": 3553
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0346336524335634e-05,
      "loss": 0.8008,
      "step": 3554
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.029271663954804e-05,
      "loss": 0.8672,
      "step": 3555
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0239110304417845e-05,
      "loss": 0.7795,
      "step": 3556
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.018551753723753e-05,
      "loss": 0.8133,
      "step": 3557
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.013193835629498e-05,
      "loss": 0.7701,
      "step": 3558
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.007837277987358e-05,
      "loss": 0.8328,
      "step": 3559
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.002482082625189e-05,
      "loss": 0.7673,
      "step": 3560
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.997128251370392e-05,
      "loss": 0.844,
      "step": 3561
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9917757860499036e-05,
      "loss": 0.8508,
      "step": 3562
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.986424688490191e-05,
      "loss": 0.8029,
      "step": 3563
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.981074960517248e-05,
      "loss": 0.3484,
      "step": 3564
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9757266039566195e-05,
      "loss": 0.6768,
      "step": 3565
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.970379620633364e-05,
      "loss": 0.9029,
      "step": 3566
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.965034012372077e-05,
      "loss": 0.8655,
      "step": 3567
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9596897809968944e-05,
      "loss": 0.822,
      "step": 3568
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.954346928331469e-05,
      "loss": 0.8212,
      "step": 3569
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9490054561989836e-05,
      "loss": 0.784,
      "step": 3570
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.943665366422164e-05,
      "loss": 0.7876,
      "step": 3571
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.938326660823251e-05,
      "loss": 0.8467,
      "step": 3572
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9329893412240124e-05,
      "loss": 0.7916,
      "step": 3573
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.927653409445755e-05,
      "loss": 0.8538,
      "step": 3574
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9223188673092976e-05,
      "loss": 0.7254,
      "step": 3575
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.916985716634999e-05,
      "loss": 0.8231,
      "step": 3576
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.911653959242735e-05,
      "loss": 0.8201,
      "step": 3577
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.906323596951898e-05,
      "loss": 0.7772,
      "step": 3578
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.900994631581425e-05,
      "loss": 0.7941,
      "step": 3579
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8956670649497595e-05,
      "loss": 0.7934,
      "step": 3580
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8903408988748665e-05,
      "loss": 0.869,
      "step": 3581
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.885016135174253e-05,
      "loss": 0.7793,
      "step": 3582
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8796927756649246e-05,
      "loss": 0.8063,
      "step": 3583
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.874370822163415e-05,
      "loss": 0.8295,
      "step": 3584
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.869050276485788e-05,
      "loss": 0.8164,
      "step": 3585
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.863731140447615e-05,
      "loss": 0.7569,
      "step": 3586
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.858413415863987e-05,
      "loss": 0.7865,
      "step": 3587
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8530971045495244e-05,
      "loss": 0.83,
      "step": 3588
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.847782208318353e-05,
      "loss": 0.378,
      "step": 3589
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8424687289841184e-05,
      "loss": 0.7963,
      "step": 3590
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.837156668359989e-05,
      "loss": 0.7546,
      "step": 3591
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.831846028258645e-05,
      "loss": 0.8051,
      "step": 3592
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8265368104922776e-05,
      "loss": 0.7842,
      "step": 3593
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8212290168726027e-05,
      "loss": 0.3677,
      "step": 3594
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.815922649210841e-05,
      "loss": 0.7852,
      "step": 3595
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.810617709317725e-05,
      "loss": 0.7767,
      "step": 3596
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8053141990035155e-05,
      "loss": 0.3114,
      "step": 3597
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.800012120077969e-05,
      "loss": 0.8044,
      "step": 3598
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.794711474350354e-05,
      "loss": 0.7945,
      "step": 3599
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7894122636294666e-05,
      "loss": 0.8523,
      "step": 3600
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7841144897235964e-05,
      "loss": 0.8234,
      "step": 3601
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.778818154440544e-05,
      "loss": 0.799,
      "step": 3602
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.773523259587632e-05,
      "loss": 0.7672,
      "step": 3603
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.768229806971678e-05,
      "loss": 0.8032,
      "step": 3604
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.762937798399008e-05,
      "loss": 0.7909,
      "step": 3605
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7576472356754696e-05,
      "loss": 0.7839,
      "step": 3606
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.752358120606403e-05,
      "loss": 0.8023,
      "step": 3607
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.74707045499665e-05,
      "loss": 0.816,
      "step": 3608
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7417842406505795e-05,
      "loss": 0.8644,
      "step": 3609
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.73649947937204e-05,
      "loss": 0.828,
      "step": 3610
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7312161729644064e-05,
      "loss": 0.8124,
      "step": 3611
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.725934323230543e-05,
      "loss": 0.8396,
      "step": 3612
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7206539319728146e-05,
      "loss": 0.7564,
      "step": 3613
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7153750009931026e-05,
      "loss": 0.8354,
      "step": 3614
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.71009753209278e-05,
      "loss": 0.7069,
      "step": 3615
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.704821527072717e-05,
      "loss": 0.7742,
      "step": 3616
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.6995469877333e-05,
      "loss": 0.7036,
      "step": 3617
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.694273915874398e-05,
      "loss": 0.7974,
      "step": 3618
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.6890023132953865e-05,
      "loss": 0.8591,
      "step": 3619
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.683732181795145e-05,
      "loss": 0.832,
      "step": 3620
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.678463523172044e-05,
      "loss": 0.7827,
      "step": 3621
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.6731963392239474e-05,
      "loss": 0.7791,
      "step": 3622
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.667930631748228e-05,
      "loss": 0.7428,
      "step": 3623
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.66266640254175e-05,
      "loss": 0.7769,
      "step": 3624
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.657403653400861e-05,
      "loss": 0.8132,
      "step": 3625
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.6521423861214264e-05,
      "loss": 0.768,
      "step": 3626
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.646882602498784e-05,
      "loss": 0.8135,
      "step": 3627
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.641624304327784e-05,
      "loss": 0.8525,
      "step": 3628
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.636367493402755e-05,
      "loss": 0.8034,
      "step": 3629
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.631112171517521e-05,
      "loss": 0.8119,
      "step": 3630
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.6258583404654063e-05,
      "loss": 0.8033,
      "step": 3631
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.620606002039222e-05,
      "loss": 0.8171,
      "step": 3632
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.615355158031259e-05,
      "loss": 0.8057,
      "step": 3633
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.610105810233319e-05,
      "loss": 0.8048,
      "step": 3634
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.604857960436678e-05,
      "loss": 0.772,
      "step": 3635
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.599611610432099e-05,
      "loss": 0.8076,
      "step": 3636
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5943667620098496e-05,
      "loss": 0.8248,
      "step": 3637
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.589123416959668e-05,
      "loss": 0.789,
      "step": 3638
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5838815770707844e-05,
      "loss": 0.8,
      "step": 3639
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.578641244131923e-05,
      "loss": 0.8344,
      "step": 3640
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.573402419931283e-05,
      "loss": 0.755,
      "step": 3641
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.568165106256551e-05,
      "loss": 0.7624,
      "step": 3642
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5629293048949086e-05,
      "loss": 0.7654,
      "step": 3643
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.557695017633008e-05,
      "loss": 0.7837,
      "step": 3644
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.552462246256986e-05,
      "loss": 0.8409,
      "step": 3645
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5472309925524765e-05,
      "loss": 0.8114,
      "step": 3646
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5420012583045745e-05,
      "loss": 0.7839,
      "step": 3647
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5367730452978764e-05,
      "loss": 0.7523,
      "step": 3648
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5315463553164436e-05,
      "loss": 0.8295,
      "step": 3649
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.526321190143825e-05,
      "loss": 0.7899,
      "step": 3650
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.521097551563055e-05,
      "loss": 0.7752,
      "step": 3651
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5158754413566336e-05,
      "loss": 0.8674,
      "step": 3652
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5106548613065455e-05,
      "loss": 0.8042,
      "step": 3653
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.505435813194262e-05,
      "loss": 0.7718,
      "step": 3654
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.500218298800718e-05,
      "loss": 0.3501,
      "step": 3655
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4950023199063283e-05,
      "loss": 0.8086,
      "step": 3656
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4897878782909925e-05,
      "loss": 0.8099,
      "step": 3657
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.484574975734077e-05,
      "loss": 0.7843,
      "step": 3658
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.479363614014421e-05,
      "loss": 0.8091,
      "step": 3659
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.474153794910349e-05,
      "loss": 0.8956,
      "step": 3660
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.468945520199651e-05,
      "loss": 0.7822,
      "step": 3661
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.463738791659583e-05,
      "loss": 0.8,
      "step": 3662
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.458533611066894e-05,
      "loss": 0.7543,
      "step": 3663
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.453329980197782e-05,
      "loss": 0.7822,
      "step": 3664
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.448127900827935e-05,
      "loss": 0.8645,
      "step": 3665
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4429273747325e-05,
      "loss": 0.7865,
      "step": 3666
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.437728403686093e-05,
      "loss": 0.8343,
      "step": 3667
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.432530989462812e-05,
      "loss": 0.8624,
      "step": 3668
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.427335133836209e-05,
      "loss": 0.8359,
      "step": 3669
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.422140838579308e-05,
      "loss": 0.786,
      "step": 3670
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.416948105464612e-05,
      "loss": 0.3071,
      "step": 3671
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4117569362640765e-05,
      "loss": 0.7906,
      "step": 3672
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.406567332749126e-05,
      "loss": 0.8228,
      "step": 3673
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4013792966906594e-05,
      "loss": 0.7645,
      "step": 3674
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.396192829859033e-05,
      "loss": 0.8506,
      "step": 3675
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.391007934024064e-05,
      "loss": 0.7978,
      "step": 3676
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.385824610955047e-05,
      "loss": 0.7858,
      "step": 3677
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.380642862420729e-05,
      "loss": 0.3297,
      "step": 3678
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.375462690189317e-05,
      "loss": 0.8112,
      "step": 3679
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.370284096028495e-05,
      "loss": 0.7925,
      "step": 3680
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.365107081705396e-05,
      "loss": 0.8487,
      "step": 3681
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.359931648986609e-05,
      "loss": 0.8128,
      "step": 3682
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.354757799638204e-05,
      "loss": 0.7875,
      "step": 3683
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.3495855354256874e-05,
      "loss": 0.7388,
      "step": 3684
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.344414858114045e-05,
      "loss": 0.3126,
      "step": 3685
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.339245769467705e-05,
      "loss": 0.7997,
      "step": 3686
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.334078271250559e-05,
      "loss": 0.3359,
      "step": 3687
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.328912365225962e-05,
      "loss": 0.7987,
      "step": 3688
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.3237480531567186e-05,
      "loss": 0.8435,
      "step": 3689
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.3185853368050866e-05,
      "loss": 0.8204,
      "step": 3690
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.313424217932791e-05,
      "loss": 0.8107,
      "step": 3691
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.308264698301002e-05,
      "loss": 0.8617,
      "step": 3692
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.303106779670341e-05,
      "loss": 0.7897,
      "step": 3693
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2979504638009e-05,
      "loss": 0.784,
      "step": 3694
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.292795752452208e-05,
      "loss": 0.7885,
      "step": 3695
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.287642647383244e-05,
      "loss": 0.8067,
      "step": 3696
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.28249115035246e-05,
      "loss": 0.753,
      "step": 3697
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2773412631177366e-05,
      "loss": 0.858,
      "step": 3698
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2721929874364126e-05,
      "loss": 0.8512,
      "step": 3699
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.267046325065286e-05,
      "loss": 0.7955,
      "step": 3700
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2619012777605904e-05,
      "loss": 0.8538,
      "step": 3701
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.256757847278014e-05,
      "loss": 0.6938,
      "step": 3702
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2516160353727004e-05,
      "loss": 0.8589,
      "step": 3703
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.24647584379923e-05,
      "loss": 0.33,
      "step": 3704
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.241337274311631e-05,
      "loss": 0.8327,
      "step": 3705
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2362003286633875e-05,
      "loss": 0.78,
      "step": 3706
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2310650086074184e-05,
      "loss": 0.7854,
      "step": 3707
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2259313158961e-05,
      "loss": 0.7639,
      "step": 3708
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.220799252281241e-05,
      "loss": 0.7859,
      "step": 3709
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.215668819514097e-05,
      "loss": 0.8385,
      "step": 3710
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.210540019345378e-05,
      "loss": 0.7762,
      "step": 3711
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2054128535252246e-05,
      "loss": 0.7807,
      "step": 3712
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.200287323803218e-05,
      "loss": 0.829,
      "step": 3713
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.195163431928396e-05,
      "loss": 0.8097,
      "step": 3714
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.190041179649223e-05,
      "loss": 0.7741,
      "step": 3715
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.184920568713608e-05,
      "loss": 0.6905,
      "step": 3716
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.179801600868905e-05,
      "loss": 0.7724,
      "step": 3717
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.174684277861903e-05,
      "loss": 0.8309,
      "step": 3718
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.169568601438825e-05,
      "loss": 0.7399,
      "step": 3719
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.1644545733453445e-05,
      "loss": 0.8242,
      "step": 3720
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.159342195326561e-05,
      "loss": 0.8156,
      "step": 3721
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.154231469127011e-05,
      "loss": 0.7986,
      "step": 3722
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.149122396490682e-05,
      "loss": 0.8897,
      "step": 3723
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.1440149791609796e-05,
      "loss": 0.8116,
      "step": 3724
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.138909218880748e-05,
      "loss": 0.7656,
      "step": 3725
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.133805117392281e-05,
      "loss": 0.7371,
      "step": 3726
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.128702676437283e-05,
      "loss": 0.7917,
      "step": 3727
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.123601897756913e-05,
      "loss": 0.7405,
      "step": 3728
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.118502783091752e-05,
      "loss": 0.8226,
      "step": 3729
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.1134053341818066e-05,
      "loss": 0.8543,
      "step": 3730
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.108309552766534e-05,
      "loss": 0.8129,
      "step": 3731
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.1032154405848054e-05,
      "loss": 0.8104,
      "step": 3732
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.098122999374927e-05,
      "loss": 0.7891,
      "step": 3733
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.093032230874641e-05,
      "loss": 0.8093,
      "step": 3734
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.087943136821114e-05,
      "loss": 0.8121,
      "step": 3735
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.082855718950934e-05,
      "loss": 0.7964,
      "step": 3736
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.077769979000133e-05,
      "loss": 0.7998,
      "step": 3737
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.072685918704158e-05,
      "loss": 0.7644,
      "step": 3738
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.067603539797885e-05,
      "loss": 0.7178,
      "step": 3739
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.062522844015621e-05,
      "loss": 0.8182,
      "step": 3740
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.0574438330910957e-05,
      "loss": 0.7683,
      "step": 3741
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.052366508757457e-05,
      "loss": 0.7619,
      "step": 3742
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.047290872747292e-05,
      "loss": 0.797,
      "step": 3743
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.042216926792602e-05,
      "loss": 0.7598,
      "step": 3744
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.037144672624807e-05,
      "loss": 0.7416,
      "step": 3745
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.032074111974765e-05,
      "loss": 0.3664,
      "step": 3746
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.02700524657274e-05,
      "loss": 0.8258,
      "step": 3747
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.0219380781484246e-05,
      "loss": 0.7831,
      "step": 3748
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.016872608430938e-05,
      "loss": 0.7439,
      "step": 3749
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.011808839148807e-05,
      "loss": 0.828,
      "step": 3750
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.006746772029992e-05,
      "loss": 0.7782,
      "step": 3751
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.001686408801862e-05,
      "loss": 0.7763,
      "step": 3752
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.996627751191204e-05,
      "loss": 0.8153,
      "step": 3753
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.991570800924236e-05,
      "loss": 0.8359,
      "step": 3754
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.98651555972658e-05,
      "loss": 0.8172,
      "step": 3755
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.981462029323275e-05,
      "loss": 0.3005,
      "step": 3756
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.976410211438789e-05,
      "loss": 0.7636,
      "step": 3757
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.971360107796994e-05,
      "loss": 0.8407,
      "step": 3758
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.966311720121175e-05,
      "loss": 0.8249,
      "step": 3759
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.961265050134044e-05,
      "loss": 0.8297,
      "step": 3760
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.956220099557718e-05,
      "loss": 0.8804,
      "step": 3761
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.951176870113723e-05,
      "loss": 0.7907,
      "step": 3762
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.946135363523012e-05,
      "loss": 0.7804,
      "step": 3763
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.941095581505938e-05,
      "loss": 0.7563,
      "step": 3764
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.936057525782265e-05,
      "loss": 0.8272,
      "step": 3765
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.9310211980711786e-05,
      "loss": 0.8364,
      "step": 3766
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.925986600091266e-05,
      "loss": 0.848,
      "step": 3767
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.920953733560521e-05,
      "loss": 0.8314,
      "step": 3768
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.915922600196361e-05,
      "loss": 0.8064,
      "step": 3769
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.910893201715598e-05,
      "loss": 0.8109,
      "step": 3770
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.9058655398344533e-05,
      "loss": 0.7825,
      "step": 3771
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.9008396162685667e-05,
      "loss": 0.8482,
      "step": 3772
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.89581543273297e-05,
      "loss": 0.8015,
      "step": 3773
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.890792990942116e-05,
      "loss": 0.8159,
      "step": 3774
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.885772292609852e-05,
      "loss": 0.7786,
      "step": 3775
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.8807533394494286e-05,
      "loss": 0.798,
      "step": 3776
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.875736133173516e-05,
      "loss": 0.8137,
      "step": 3777
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.870720675494174e-05,
      "loss": 0.8077,
      "step": 3778
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.865706968122866e-05,
      "loss": 0.7697,
      "step": 3779
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.860695012770472e-05,
      "loss": 0.7832,
      "step": 3780
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.855684811147257e-05,
      "loss": 0.8397,
      "step": 3781
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.8506763649628964e-05,
      "loss": 0.7983,
      "step": 3782
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.845669675926468e-05,
      "loss": 0.8091,
      "step": 3783
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.840664745746447e-05,
      "loss": 0.3109,
      "step": 3784
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.8356615761307024e-05,
      "loss": 0.7656,
      "step": 3785
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.830660168786518e-05,
      "loss": 0.7868,
      "step": 3786
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.825660525420562e-05,
      "loss": 0.85,
      "step": 3787
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.820662647738904e-05,
      "loss": 0.8057,
      "step": 3788
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.815666537447018e-05,
      "loss": 0.7719,
      "step": 3789
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.810672196249767e-05,
      "loss": 0.8112,
      "step": 3790
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.805679625851408e-05,
      "loss": 0.7983,
      "step": 3791
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.800688827955607e-05,
      "loss": 0.8419,
      "step": 3792
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.795699804265411e-05,
      "loss": 0.712,
      "step": 3793
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.790712556483272e-05,
      "loss": 0.765,
      "step": 3794
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.7857270863110295e-05,
      "loss": 0.7193,
      "step": 3795
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.780743395449914e-05,
      "loss": 0.8207,
      "step": 3796
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.77576148560056e-05,
      "loss": 0.7944,
      "step": 3797
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.770781358462984e-05,
      "loss": 0.8126,
      "step": 3798
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.765803015736594e-05,
      "loss": 0.8464,
      "step": 3799
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.760826459120201e-05,
      "loss": 0.8159,
      "step": 3800
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.755851690311993e-05,
      "loss": 0.802,
      "step": 3801
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.750878711009551e-05,
      "loss": 0.8188,
      "step": 3802
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.7459075229098525e-05,
      "loss": 0.7645,
      "step": 3803
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.740938127709258e-05,
      "loss": 0.7369,
      "step": 3804
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.735970527103511e-05,
      "loss": 0.7513,
      "step": 3805
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.731004722787755e-05,
      "loss": 0.8263,
      "step": 3806
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.726040716456514e-05,
      "loss": 0.6958,
      "step": 3807
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.721078509803691e-05,
      "loss": 0.8352,
      "step": 3808
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.716118104522592e-05,
      "loss": 0.3136,
      "step": 3809
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.711159502305894e-05,
      "loss": 0.726,
      "step": 3810
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.706202704845661e-05,
      "loss": 0.8119,
      "step": 3811
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.70124771383335e-05,
      "loss": 0.7706,
      "step": 3812
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6962945309597915e-05,
      "loss": 0.7974,
      "step": 3813
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.691343157915199e-05,
      "loss": 0.8165,
      "step": 3814
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6863935963891794e-05,
      "loss": 0.7479,
      "step": 3815
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6814458480707056e-05,
      "loss": 0.7936,
      "step": 3816
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.67649991464815e-05,
      "loss": 0.7722,
      "step": 3817
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.671555797809252e-05,
      "loss": 0.811,
      "step": 3818
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.666613499241128e-05,
      "loss": 0.8116,
      "step": 3819
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.661673020630293e-05,
      "loss": 0.8117,
      "step": 3820
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.656734363662622e-05,
      "loss": 0.785,
      "step": 3821
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.651797530023373e-05,
      "loss": 0.8128,
      "step": 3822
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6468625213971906e-05,
      "loss": 0.8143,
      "step": 3823
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6419293394680876e-05,
      "loss": 0.8307,
      "step": 3824
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6369979859194525e-05,
      "loss": 0.7939,
      "step": 3825
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6320684624340594e-05,
      "loss": 0.7895,
      "step": 3826
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6271407706940486e-05,
      "loss": 0.7768,
      "step": 3827
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6222149123809356e-05,
      "loss": 0.8186,
      "step": 3828
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.617290889175621e-05,
      "loss": 0.7825,
      "step": 3829
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.612368702758367e-05,
      "loss": 0.8352,
      "step": 3830
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.60744835480881e-05,
      "loss": 0.8732,
      "step": 3831
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.602529847005971e-05,
      "loss": 0.816,
      "step": 3832
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5976131810282295e-05,
      "loss": 0.801,
      "step": 3833
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.592698358553338e-05,
      "loss": 0.8035,
      "step": 3834
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.587785381258433e-05,
      "loss": 0.8161,
      "step": 3835
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.582874250820006e-05,
      "loss": 0.7587,
      "step": 3836
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5779649689139215e-05,
      "loss": 0.8011,
      "step": 3837
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.573057537215423e-05,
      "loss": 0.7773,
      "step": 3838
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.568151957399108e-05,
      "loss": 0.7724,
      "step": 3839
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.563248231138958e-05,
      "loss": 0.781,
      "step": 3840
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.55834636010831e-05,
      "loss": 0.8201,
      "step": 3841
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.553446345979866e-05,
      "loss": 0.8073,
      "step": 3842
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.548548190425709e-05,
      "loss": 0.8427,
      "step": 3843
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5436518951172735e-05,
      "loss": 0.8541,
      "step": 3844
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5387574617253625e-05,
      "loss": 0.7696,
      "step": 3845
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.533864891920152e-05,
      "loss": 0.7712,
      "step": 3846
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.528974187371172e-05,
      "loss": 0.8123,
      "step": 3847
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.524085349747316e-05,
      "loss": 0.8032,
      "step": 3848
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.519198380716851e-05,
      "loss": 0.7347,
      "step": 3849
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5143132819473964e-05,
      "loss": 0.8047,
      "step": 3850
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.509430055105933e-05,
      "loss": 0.3493,
      "step": 3851
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.504548701858813e-05,
      "loss": 0.8111,
      "step": 3852
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.499669223871738e-05,
      "loss": 0.7334,
      "step": 3853
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.494791622809772e-05,
      "loss": 0.7633,
      "step": 3854
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.489915900337347e-05,
      "loss": 0.8084,
      "step": 3855
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.485042058118244e-05,
      "loss": 0.8484,
      "step": 3856
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4801700978156034e-05,
      "loss": 0.783,
      "step": 3857
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4753000210919325e-05,
      "loss": 0.7962,
      "step": 3858
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.470431829609083e-05,
      "loss": 0.7822,
      "step": 3859
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4655655250282756e-05,
      "loss": 0.3481,
      "step": 3860
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4607011090100784e-05,
      "loss": 0.8052,
      "step": 3861
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4558385832144143e-05,
      "loss": 0.7697,
      "step": 3862
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.450977949300571e-05,
      "loss": 0.8025,
      "step": 3863
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.446119208927183e-05,
      "loss": 0.7693,
      "step": 3864
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.441262363752234e-05,
      "loss": 0.8086,
      "step": 3865
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.436407415433075e-05,
      "loss": 0.8088,
      "step": 3866
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4315543656263994e-05,
      "loss": 0.7674,
      "step": 3867
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.426703215988249e-05,
      "loss": 0.7984,
      "step": 3868
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4218539681740326e-05,
      "loss": 0.8287,
      "step": 3869
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4170066238384954e-05,
      "loss": 0.8145,
      "step": 3870
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.412161184635737e-05,
      "loss": 0.7854,
      "step": 3871
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.407317652219213e-05,
      "loss": 0.8289,
      "step": 3872
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.402476028241721e-05,
      "loss": 0.7871,
      "step": 3873
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3976363143554067e-05,
      "loss": 0.7373,
      "step": 3874
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3927985122117745e-05,
      "loss": 0.7978,
      "step": 3875
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3879626234616644e-05,
      "loss": 0.7475,
      "step": 3876
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3831286497552645e-05,
      "loss": 0.78,
      "step": 3877
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.378296592742124e-05,
      "loss": 0.8283,
      "step": 3878
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.373466454071119e-05,
      "loss": 0.3363,
      "step": 3879
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.368638235390478e-05,
      "loss": 0.8062,
      "step": 3880
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.363811938347781e-05,
      "loss": 0.7316,
      "step": 3881
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.358987564589942e-05,
      "loss": 0.7643,
      "step": 3882
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.35416511576323e-05,
      "loss": 0.8135,
      "step": 3883
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.349344593513246e-05,
      "loss": 0.779,
      "step": 3884
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.344525999484934e-05,
      "loss": 0.7932,
      "step": 3885
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.339709335322594e-05,
      "loss": 0.7901,
      "step": 3886
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.334894602669852e-05,
      "loss": 0.7997,
      "step": 3887
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3300818031696786e-05,
      "loss": 0.307,
      "step": 3888
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.325270938464392e-05,
      "loss": 0.8205,
      "step": 3889
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3204620101956425e-05,
      "loss": 0.7799,
      "step": 3890
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.315655020004419e-05,
      "loss": 0.7652,
      "step": 3891
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.310849969531057e-05,
      "loss": 0.779,
      "step": 3892
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.306046860415224e-05,
      "loss": 0.7748,
      "step": 3893
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.30124569429592e-05,
      "loss": 0.8367,
      "step": 3894
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.296446472811498e-05,
      "loss": 0.7834,
      "step": 3895
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.291649197599633e-05,
      "loss": 0.8583,
      "step": 3896
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.286853870297336e-05,
      "loss": 0.7926,
      "step": 3897
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.282060492540966e-05,
      "loss": 0.7788,
      "step": 3898
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.277269065966204e-05,
      "loss": 0.8291,
      "step": 3899
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2724795922080664e-05,
      "loss": 0.7871,
      "step": 3900
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.267692072900914e-05,
      "loss": 0.8275,
      "step": 3901
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2629065096784294e-05,
      "loss": 0.767,
      "step": 3902
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.258122904173627e-05,
      "loss": 0.7916,
      "step": 3903
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.253341258018867e-05,
      "loss": 0.835,
      "step": 3904
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2485615728458214e-05,
      "loss": 0.8584,
      "step": 3905
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.243783850285512e-05,
      "loss": 0.7382,
      "step": 3906
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.23900809196828e-05,
      "loss": 0.7944,
      "step": 3907
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2342342995237913e-05,
      "loss": 0.8071,
      "step": 3908
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.229462474581057e-05,
      "loss": 0.3268,
      "step": 3909
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.2246926187684064e-05,
      "loss": 0.8308,
      "step": 3910
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.219924733713492e-05,
      "loss": 0.7912,
      "step": 3911
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.215158821043308e-05,
      "loss": 0.7659,
      "step": 3912
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.210394882384165e-05,
      "loss": 0.756,
      "step": 3913
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.2056329193616985e-05,
      "loss": 0.8033,
      "step": 3914
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.20087293360088e-05,
      "loss": 0.3282,
      "step": 3915
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.196114926725999e-05,
      "loss": 0.8232,
      "step": 3916
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.191358900360667e-05,
      "loss": 0.8353,
      "step": 3917
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.186604856127829e-05,
      "loss": 0.794,
      "step": 3918
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.181852795649747e-05,
      "loss": 0.827,
      "step": 3919
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1771027205480026e-05,
      "loss": 0.8217,
      "step": 3920
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.172354632443513e-05,
      "loss": 0.8075,
      "step": 3921
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.167608532956503e-05,
      "loss": 0.7453,
      "step": 3922
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.162864423706524e-05,
      "loss": 0.7914,
      "step": 3923
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.158122306312455e-05,
      "loss": 0.8456,
      "step": 3924
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1533821823924834e-05,
      "loss": 0.7772,
      "step": 3925
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.148644053564128e-05,
      "loss": 0.8091,
      "step": 3926
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.143907921444219e-05,
      "loss": 0.7641,
      "step": 3927
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1391737876489026e-05,
      "loss": 0.8454,
      "step": 3928
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.134441653793656e-05,
      "loss": 0.783,
      "step": 3929
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1297115214932614e-05,
      "loss": 0.8136,
      "step": 3930
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.124983392361821e-05,
      "loss": 0.747,
      "step": 3931
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.120257268012759e-05,
      "loss": 0.7708,
      "step": 3932
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1155331500588103e-05,
      "loss": 0.8231,
      "step": 3933
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.110811040112021e-05,
      "loss": 0.7712,
      "step": 3934
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.106090939783766e-05,
      "loss": 0.8326,
      "step": 3935
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1013728506847216e-05,
      "loss": 0.819,
      "step": 3936
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.096656774424878e-05,
      "loss": 0.7423,
      "step": 3937
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.091942712613548e-05,
      "loss": 0.7695,
      "step": 3938
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0872306668593505e-05,
      "loss": 0.8274,
      "step": 3939
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.082520638770213e-05,
      "loss": 0.8053,
      "step": 3940
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.077812629953385e-05,
      "loss": 0.7643,
      "step": 3941
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0731066420154174e-05,
      "loss": 0.7939,
      "step": 3942
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.068402676562172e-05,
      "loss": 0.8119,
      "step": 3943
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.063700735198831e-05,
      "loss": 0.7698,
      "step": 3944
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0590008195298724e-05,
      "loss": 0.7735,
      "step": 3945
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0543029311590876e-05,
      "loss": 0.8615,
      "step": 3946
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.049607071689584e-05,
      "loss": 0.7628,
      "step": 3947
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0449132427237624e-05,
      "loss": 0.7959,
      "step": 3948
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0402214458633446e-05,
      "loss": 0.7378,
      "step": 3949
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0355316827093514e-05,
      "loss": 0.7263,
      "step": 3950
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.030843954862106e-05,
      "loss": 0.824,
      "step": 3951
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.026158263921249e-05,
      "loss": 0.7752,
      "step": 3952
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.021474611485716e-05,
      "loss": 0.8154,
      "step": 3953
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.016792999153746e-05,
      "loss": 0.7728,
      "step": 3954
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.012113428522894e-05,
      "loss": 0.7766,
      "step": 3955
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0074359011900045e-05,
      "loss": 0.7783,
      "step": 3956
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.002760418751228e-05,
      "loss": 0.7487,
      "step": 3957
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.998086982802026e-05,
      "loss": 0.7065,
      "step": 3958
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.9934155949371534e-05,
      "loss": 0.7395,
      "step": 3959
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.9887462567506616e-05,
      "loss": 0.7754,
      "step": 3960
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.984078969835919e-05,
      "loss": 0.7381,
      "step": 3961
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.979413735785579e-05,
      "loss": 0.76,
      "step": 3962
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.974750556191596e-05,
      "loss": 0.8244,
      "step": 3963
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.970089432645234e-05,
      "loss": 0.3379,
      "step": 3964
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.965430366737044e-05,
      "loss": 0.8078,
      "step": 3965
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.960773360056876e-05,
      "loss": 0.7609,
      "step": 3966
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.956118414193889e-05,
      "loss": 0.7435,
      "step": 3967
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.9514655307365244e-05,
      "loss": 0.3428,
      "step": 3968
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.946814711272523e-05,
      "loss": 0.7925,
      "step": 3969
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.9421659573889315e-05,
      "loss": 0.7742,
      "step": 3970
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.9375192706720754e-05,
      "loss": 0.7877,
      "step": 3971
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.932874652707592e-05,
      "loss": 0.8531,
      "step": 3972
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.9282321050804005e-05,
      "loss": 0.3102,
      "step": 3973
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.9235916293747135e-05,
      "loss": 0.3538,
      "step": 3974
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.918953227174047e-05,
      "loss": 0.7348,
      "step": 3975
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.9143169000612e-05,
      "loss": 0.7692,
      "step": 3976
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.909682649618262e-05,
      "loss": 0.7997,
      "step": 3977
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.9050504774266264e-05,
      "loss": 0.8369,
      "step": 3978
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.900420385066964e-05,
      "loss": 0.775,
      "step": 3979
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8957923741192384e-05,
      "loss": 0.7979,
      "step": 3980
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8911664461627126e-05,
      "loss": 0.8173,
      "step": 3981
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8865426027759276e-05,
      "loss": 0.7916,
      "step": 3982
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.881920845536714e-05,
      "loss": 0.82,
      "step": 3983
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.877301176022201e-05,
      "loss": 0.817,
      "step": 3984
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.872683595808795e-05,
      "loss": 0.7759,
      "step": 3985
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.868068106472188e-05,
      "loss": 0.3297,
      "step": 3986
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8634547095873694e-05,
      "loss": 0.7846,
      "step": 3987
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.858843406728607e-05,
      "loss": 0.6892,
      "step": 3988
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.854234199469451e-05,
      "loss": 0.8347,
      "step": 3989
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.849627089382747e-05,
      "loss": 0.8264,
      "step": 3990
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.845022078040616e-05,
      "loss": 0.7883,
      "step": 3991
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.840419167014461e-05,
      "loss": 0.7901,
      "step": 3992
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8358183578749805e-05,
      "loss": 0.7967,
      "step": 3993
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.831219652192142e-05,
      "loss": 0.7766,
      "step": 3994
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8266230515352076e-05,
      "loss": 0.7297,
      "step": 3995
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.822028557472711e-05,
      "loss": 0.7695,
      "step": 3996
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.817436171572467e-05,
      "loss": 0.8375,
      "step": 3997
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8128458954015835e-05,
      "loss": 0.8433,
      "step": 3998
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.808257730526434e-05,
      "loss": 0.8182,
      "step": 3999
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.803671678512677e-05,
      "loss": 0.3506,
      "step": 4000
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7990877409252553e-05,
      "loss": 0.7677,
      "step": 4001
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7945059193283816e-05,
      "loss": 0.7111,
      "step": 4002
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.789926215285547e-05,
      "loss": 0.7014,
      "step": 4003
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.785348630359532e-05,
      "loss": 0.7967,
      "step": 4004
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.78077316611238e-05,
      "loss": 0.7567,
      "step": 4005
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.776199824105412e-05,
      "loss": 0.7302,
      "step": 4006
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.771628605899237e-05,
      "loss": 0.777,
      "step": 4007
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.767059513053727e-05,
      "loss": 0.7793,
      "step": 4008
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.762492547128029e-05,
      "loss": 0.7564,
      "step": 4009
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.757927709680577e-05,
      "loss": 0.7605,
      "step": 4010
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.753365002269062e-05,
      "loss": 0.8018,
      "step": 4011
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.748804426450456e-05,
      "loss": 0.8014,
      "step": 4012
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.744245983781011e-05,
      "loss": 0.3456,
      "step": 4013
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.739689675816233e-05,
      "loss": 0.7959,
      "step": 4014
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.735135504110919e-05,
      "loss": 0.7952,
      "step": 4015
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.730583470219125e-05,
      "loss": 0.8128,
      "step": 4016
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.726033575694178e-05,
      "loss": 0.3565,
      "step": 4017
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7214858220886826e-05,
      "loss": 0.8029,
      "step": 4018
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7169402109545046e-05,
      "loss": 0.8427,
      "step": 4019
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.71239674384278e-05,
      "loss": 0.7816,
      "step": 4020
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.70785542230392e-05,
      "loss": 0.7641,
      "step": 4021
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.703316247887596e-05,
      "loss": 0.8197,
      "step": 4022
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.698779222142747e-05,
      "loss": 0.8298,
      "step": 4023
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.694244346617587e-05,
      "loss": 0.8234,
      "step": 4024
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6897116228595876e-05,
      "loss": 0.7936,
      "step": 4025
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.685181052415486e-05,
      "loss": 0.7408,
      "step": 4026
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.680652636831293e-05,
      "loss": 0.7896,
      "step": 4027
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.676126377652278e-05,
      "loss": 0.8582,
      "step": 4028
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6716022764229695e-05,
      "loss": 0.8031,
      "step": 4029
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.667080334687173e-05,
      "loss": 0.7982,
      "step": 4030
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.662560553987946e-05,
      "loss": 0.8141,
      "step": 4031
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6580429358676094e-05,
      "loss": 0.7235,
      "step": 4032
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.653527481867755e-05,
      "loss": 0.8072,
      "step": 4033
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.649014193529225e-05,
      "loss": 0.8112,
      "step": 4034
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.644503072392127e-05,
      "loss": 0.8133,
      "step": 4035
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.639994119995835e-05,
      "loss": 0.7346,
      "step": 4036
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.635487337878971e-05,
      "loss": 0.7864,
      "step": 4037
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6309827275794303e-05,
      "loss": 0.3315,
      "step": 4038
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6264802906343554e-05,
      "loss": 0.7623,
      "step": 4039
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6219800285801475e-05,
      "loss": 0.3539,
      "step": 4040
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6174819429524787e-05,
      "loss": 0.7601,
      "step": 4041
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.612986035286264e-05,
      "loss": 0.7366,
      "step": 4042
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6084923071156785e-05,
      "loss": 0.7941,
      "step": 4043
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.604000759974161e-05,
      "loss": 0.7034,
      "step": 4044
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.599511395394398e-05,
      "loss": 0.8578,
      "step": 4045
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.595024214908332e-05,
      "loss": 0.7821,
      "step": 4046
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.590539220047167e-05,
      "loss": 0.8126,
      "step": 4047
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.586056412341353e-05,
      "loss": 0.7656,
      "step": 4048
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.581575793320595e-05,
      "loss": 0.7543,
      "step": 4049
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.577097364513859e-05,
      "loss": 0.8183,
      "step": 4050
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5726211274493536e-05,
      "loss": 0.4007,
      "step": 4051
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.568147083654543e-05,
      "loss": 0.8369,
      "step": 4052
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.563675234656146e-05,
      "loss": 0.8173,
      "step": 4053
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.55920558198013e-05,
      "loss": 0.8,
      "step": 4054
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.554738127151709e-05,
      "loss": 0.799,
      "step": 4055
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5502728716953574e-05,
      "loss": 0.8505,
      "step": 4056
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.545809817134789e-05,
      "loss": 0.8032,
      "step": 4057
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.541348964992967e-05,
      "loss": 0.8076,
      "step": 4058
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.536890316792113e-05,
      "loss": 0.7341,
      "step": 4059
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.532433874053683e-05,
      "loss": 0.8425,
      "step": 4060
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5279796382983944e-05,
      "loss": 0.7341,
      "step": 4061
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.523527611046201e-05,
      "loss": 0.807,
      "step": 4062
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.519077793816303e-05,
      "loss": 0.7576,
      "step": 4063
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.514630188127156e-05,
      "loss": 0.8652,
      "step": 4064
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.510184795496452e-05,
      "loss": 0.8186,
      "step": 4065
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.505741617441125e-05,
      "loss": 0.3146,
      "step": 4066
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.501300655477368e-05,
      "loss": 0.7712,
      "step": 4067
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.496861911120605e-05,
      "loss": 0.8284,
      "step": 4068
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4924253858855025e-05,
      "loss": 0.7335,
      "step": 4069
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4879910812859815e-05,
      "loss": 0.7323,
      "step": 4070
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4835589988351943e-05,
      "loss": 0.7661,
      "step": 4071
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.479129140045535e-05,
      "loss": 0.8045,
      "step": 4072
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.474701506428648e-05,
      "loss": 0.8585,
      "step": 4073
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.470276099495412e-05,
      "loss": 0.7816,
      "step": 4074
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4658529207559406e-05,
      "loss": 0.7881,
      "step": 4075
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.461431971719602e-05,
      "loss": 0.7522,
      "step": 4076
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.45701325389499e-05,
      "loss": 0.8067,
      "step": 4077
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4525967687899374e-05,
      "loss": 0.8105,
      "step": 4078
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.448182517911528e-05,
      "loss": 0.8221,
      "step": 4079
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.443770502766067e-05,
      "loss": 0.7595,
      "step": 4080
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.439360724859111e-05,
      "loss": 0.7877,
      "step": 4081
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.434953185695442e-05,
      "loss": 0.738,
      "step": 4082
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.430547886779081e-05,
      "loss": 0.8388,
      "step": 4083
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.426144829613291e-05,
      "loss": 0.8646,
      "step": 4084
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.421744015700563e-05,
      "loss": 0.81,
      "step": 4085
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.41734544654262e-05,
      "loss": 0.7959,
      "step": 4086
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.412949123640432e-05,
      "loss": 0.8033,
      "step": 4087
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.408555048494189e-05,
      "loss": 0.7743,
      "step": 4088
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.404163222603315e-05,
      "loss": 0.7544,
      "step": 4089
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.399773647466479e-05,
      "loss": 0.7748,
      "step": 4090
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3953863245815696e-05,
      "loss": 0.8356,
      "step": 4091
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.391001255445707e-05,
      "loss": 0.7948,
      "step": 4092
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.386618441555253e-05,
      "loss": 0.7913,
      "step": 4093
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.382237884405788e-05,
      "loss": 0.7829,
      "step": 4094
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.377859585492126e-05,
      "loss": 0.8514,
      "step": 4095
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.373483546308317e-05,
      "loss": 0.8124,
      "step": 4096
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.369109768347631e-05,
      "loss": 0.8364,
      "step": 4097
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3647382531025686e-05,
      "loss": 0.7624,
      "step": 4098
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.360369002064863e-05,
      "loss": 0.7945,
      "step": 4099
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.356002016725469e-05,
      "loss": 0.8467,
      "step": 4100
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.351637298574568e-05,
      "loss": 0.767,
      "step": 4101
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.347274849101577e-05,
      "loss": 0.7971,
      "step": 4102
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.342914669795124e-05,
      "loss": 0.7325,
      "step": 4103
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.338556762143079e-05,
      "loss": 0.7738,
      "step": 4104
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.334201127632524e-05,
      "loss": 0.8397,
      "step": 4105
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.329847767749764e-05,
      "loss": 0.7538,
      "step": 4106
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3254966839803446e-05,
      "loss": 0.7582,
      "step": 4107
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.321147877809016e-05,
      "loss": 0.8246,
      "step": 4108
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.316801350719758e-05,
      "loss": 0.7343,
      "step": 4109
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3124571041957796e-05,
      "loss": 0.8572,
      "step": 4110
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.308115139719501e-05,
      "loss": 0.7833,
      "step": 4111
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3037754587725654e-05,
      "loss": 0.8242,
      "step": 4112
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.299438062835846e-05,
      "loss": 0.8189,
      "step": 4113
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.295102953389426e-05,
      "loss": 0.7539,
      "step": 4114
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.290770131912609e-05,
      "loss": 0.8,
      "step": 4115
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.286439599883928e-05,
      "loss": 0.7463,
      "step": 4116
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2821113587811225e-05,
      "loss": 0.8429,
      "step": 4117
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.277785410081154e-05,
      "loss": 0.7948,
      "step": 4118
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2734617552602096e-05,
      "loss": 0.7723,
      "step": 4119
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.269140395793684e-05,
      "loss": 0.8197,
      "step": 4120
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.264821333156187e-05,
      "loss": 0.7874,
      "step": 4121
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.260504568821557e-05,
      "loss": 0.8129,
      "step": 4122
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2561901042628377e-05,
      "loss": 0.7986,
      "step": 4123
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.251877940952286e-05,
      "loss": 0.7798,
      "step": 4124
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.247568080361387e-05,
      "loss": 0.83,
      "step": 4125
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.243260523960824e-05,
      "loss": 0.8042,
      "step": 4126
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2389552732205076e-05,
      "loss": 0.8087,
      "step": 4127
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2346523296095534e-05,
      "loss": 0.8129,
      "step": 4128
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2303516945962876e-05,
      "loss": 0.8263,
      "step": 4129
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.226053369648259e-05,
      "loss": 0.8186,
      "step": 4130
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.221757356232219e-05,
      "loss": 0.8798,
      "step": 4131
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.21746365581413e-05,
      "loss": 0.7994,
      "step": 4132
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.213172269859176e-05,
      "loss": 0.7546,
      "step": 4133
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.2088831998317384e-05,
      "loss": 0.8283,
      "step": 4134
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.204596447195411e-05,
      "loss": 0.7336,
      "step": 4135
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.200312013413006e-05,
      "loss": 0.8059,
      "step": 4136
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.196029899946533e-05,
      "loss": 0.7899,
      "step": 4137
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.191750108257213e-05,
      "loss": 0.8036,
      "step": 4138
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.187472639805481e-05,
      "loss": 0.8049,
      "step": 4139
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.183197496050971e-05,
      "loss": 0.7697,
      "step": 4140
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.178924678452525e-05,
      "loss": 0.7473,
      "step": 4141
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1746541884681994e-05,
      "loss": 0.7794,
      "step": 4142
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1703860275552466e-05,
      "loss": 0.7973,
      "step": 4143
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1661201971701246e-05,
      "loss": 0.8069,
      "step": 4144
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.161856698768505e-05,
      "loss": 0.772,
      "step": 4145
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.157595533805253e-05,
      "loss": 0.809,
      "step": 4146
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1533367037344485e-05,
      "loss": 0.7897,
      "step": 4147
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.149080210009365e-05,
      "loss": 0.8614,
      "step": 4148
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.14482605408248e-05,
      "loss": 0.7268,
      "step": 4149
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1405742374054816e-05,
      "loss": 0.7322,
      "step": 4150
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.136324761429249e-05,
      "loss": 0.8336,
      "step": 4151
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.132077627603867e-05,
      "loss": 0.8091,
      "step": 4152
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.127832837378626e-05,
      "loss": 0.7483,
      "step": 4153
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.12359039220201e-05,
      "loss": 0.7486,
      "step": 4154
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.119350293521701e-05,
      "loss": 0.78,
      "step": 4155
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.115112542784592e-05,
      "loss": 0.738,
      "step": 4156
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.110877141436762e-05,
      "loss": 0.7738,
      "step": 4157
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1066440909234914e-05,
      "loss": 0.3411,
      "step": 4158
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1024133926892674e-05,
      "loss": 0.8034,
      "step": 4159
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.098185048177763e-05,
      "loss": 0.8171,
      "step": 4160
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.093959058831849e-05,
      "loss": 0.8756,
      "step": 4161
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.089735426093604e-05,
      "loss": 0.8774,
      "step": 4162
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0855141514042904e-05,
      "loss": 0.7771,
      "step": 4163
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.081295236204367e-05,
      "loss": 0.7683,
      "step": 4164
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0770786819334965e-05,
      "loss": 0.7913,
      "step": 4165
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.072864490030527e-05,
      "loss": 0.784,
      "step": 4166
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.068652661933501e-05,
      "loss": 0.7619,
      "step": 4167
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0644431990796616e-05,
      "loss": 0.7689,
      "step": 4168
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.060236102905434e-05,
      "loss": 0.8137,
      "step": 4169
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.05603137484645e-05,
      "loss": 0.7224,
      "step": 4170
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.051829016337519e-05,
      "loss": 0.7692,
      "step": 4171
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0476290288126464e-05,
      "loss": 0.3371,
      "step": 4172
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0434314137050357e-05,
      "loss": 0.3181,
      "step": 4173
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.039236172447073e-05,
      "loss": 0.7488,
      "step": 4174
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0350433064703322e-05,
      "loss": 0.7246,
      "step": 4175
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0308528172055884e-05,
      "loss": 0.7697,
      "step": 4176
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0266647060827934e-05,
      "loss": 0.8142,
      "step": 4177
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.022478974531091e-05,
      "loss": 0.8005,
      "step": 4178
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0182956239788195e-05,
      "loss": 0.7655,
      "step": 4179
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0141146558534973e-05,
      "loss": 0.7685,
      "step": 4180
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.009936071581828e-05,
      "loss": 0.7892,
      "step": 4181
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.005759872589713e-05,
      "loss": 0.7848,
      "step": 4182
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0015860603022283e-05,
      "loss": 0.7634,
      "step": 4183
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.9974146361436373e-05,
      "loss": 0.7751,
      "step": 4184
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.9932456015373968e-05,
      "loss": 0.7363,
      "step": 4185
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.9890789579061394e-05,
      "loss": 0.8266,
      "step": 4186
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9849147066716808e-05,
      "loss": 0.8432,
      "step": 4187
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.980752849255033e-05,
      "loss": 0.8039,
      "step": 4188
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9765933870763764e-05,
      "loss": 0.8209,
      "step": 4189
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9724363215550778e-05,
      "loss": 0.8003,
      "step": 4190
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9682816541096946e-05,
      "loss": 0.8021,
      "step": 4191
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9641293861579533e-05,
      "loss": 0.6725,
      "step": 4192
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9599795191167757e-05,
      "loss": 0.8304,
      "step": 4193
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9558320544022512e-05,
      "loss": 0.722,
      "step": 4194
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.951686993429653e-05,
      "loss": 0.8711,
      "step": 4195
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9475443376134416e-05,
      "loss": 0.3295,
      "step": 4196
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9434040883672475e-05,
      "loss": 0.7974,
      "step": 4197
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9392662471038823e-05,
      "loss": 0.8092,
      "step": 4198
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9351308152353407e-05,
      "loss": 0.3235,
      "step": 4199
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9309977941727906e-05,
      "loss": 0.8177,
      "step": 4200
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9268671853265738e-05,
      "loss": 0.7639,
      "step": 4201
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9227389901062208e-05,
      "loss": 0.3254,
      "step": 4202
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9186132099204268e-05,
      "loss": 0.748,
      "step": 4203
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9144898461770654e-05,
      "loss": 0.7985,
      "step": 4204
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9103689002831923e-05,
      "loss": 0.796,
      "step": 4205
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.906250373645032e-05,
      "loss": 0.3583,
      "step": 4206
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9021342676679796e-05,
      "loss": 0.3677,
      "step": 4207
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8980205837566166e-05,
      "loss": 0.7632,
      "step": 4208
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8939093233146875e-05,
      "loss": 0.8038,
      "step": 4209
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.889800487745109e-05,
      "loss": 0.7485,
      "step": 4210
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.885694078449982e-05,
      "loss": 0.7863,
      "step": 4211
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8815900968305652e-05,
      "loss": 0.7675,
      "step": 4212
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8774885442873013e-05,
      "loss": 0.7683,
      "step": 4213
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8733894222197945e-05,
      "loss": 0.799,
      "step": 4214
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8692927320268214e-05,
      "loss": 0.8087,
      "step": 4215
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8651984751063365e-05,
      "loss": 0.786,
      "step": 4216
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8611066528554564e-05,
      "loss": 0.7023,
      "step": 4217
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8570172666704644e-05,
      "loss": 0.7498,
      "step": 4218
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8529303179468226e-05,
      "loss": 0.8059,
      "step": 4219
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8488458080791536e-05,
      "loss": 0.8025,
      "step": 4220
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.844763738461247e-05,
      "loss": 0.7602,
      "step": 4221
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8406841104860672e-05,
      "loss": 0.8253,
      "step": 4222
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8366069255457383e-05,
      "loss": 0.7416,
      "step": 4223
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.83253218503155e-05,
      "loss": 0.803,
      "step": 4224
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8284598903339677e-05,
      "loss": 0.795,
      "step": 4225
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8243900428426116e-05,
      "loss": 0.7161,
      "step": 4226
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.820322643946268e-05,
      "loss": 0.8246,
      "step": 4227
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.816257695032898e-05,
      "loss": 0.8023,
      "step": 4228
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8121951974896134e-05,
      "loss": 0.739,
      "step": 4229
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8081351527026923e-05,
      "loss": 0.6718,
      "step": 4230
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8040775620575866e-05,
      "loss": 0.8059,
      "step": 4231
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8000224269388985e-05,
      "loss": 0.8092,
      "step": 4232
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7959697487303948e-05,
      "loss": 0.8562,
      "step": 4233
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.79191952881501e-05,
      "loss": 0.3823,
      "step": 4234
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.78787176857483e-05,
      "loss": 0.7985,
      "step": 4235
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.783826469391113e-05,
      "loss": 0.7756,
      "step": 4236
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7797836326442685e-05,
      "loss": 0.8086,
      "step": 4237
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7757432597138655e-05,
      "loss": 0.7246,
      "step": 4238
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7717053519786397e-05,
      "loss": 0.7686,
      "step": 4239
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7676699108164796e-05,
      "loss": 0.761,
      "step": 4240
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7636369376044302e-05,
      "loss": 0.7456,
      "step": 4241
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7596064337187035e-05,
      "loss": 0.3391,
      "step": 4242
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7555784005346595e-05,
      "loss": 0.7588,
      "step": 4243
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7515528394268153e-05,
      "loss": 0.7922,
      "step": 4244
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7475297517688546e-05,
      "loss": 0.7881,
      "step": 4245
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7435091389336065e-05,
      "loss": 0.8024,
      "step": 4246
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7394910022930576e-05,
      "loss": 0.863,
      "step": 4247
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7354753432183554e-05,
      "loss": 0.8066,
      "step": 4248
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7314621630797953e-05,
      "loss": 0.8008,
      "step": 4249
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7274514632468263e-05,
      "loss": 0.78,
      "step": 4250
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7234432450880597e-05,
      "loss": 0.8576,
      "step": 4251
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7194375099712522e-05,
      "loss": 0.7795,
      "step": 4252
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.715434259263312e-05,
      "loss": 0.7952,
      "step": 4253
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.711433494330308e-05,
      "loss": 0.7515,
      "step": 4254
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7074352165374518e-05,
      "loss": 0.8604,
      "step": 4255
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7034394272491092e-05,
      "loss": 0.3238,
      "step": 4256
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.699446127828802e-05,
      "loss": 0.8104,
      "step": 4257
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.695455319639193e-05,
      "loss": 0.8323,
      "step": 4258
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6914670040421052e-05,
      "loss": 0.813,
      "step": 4259
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6874811823985034e-05,
      "loss": 0.8332,
      "step": 4260
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.683497856068501e-05,
      "loss": 0.7984,
      "step": 4261
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6795170264113667e-05,
      "loss": 0.7862,
      "step": 4262
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6755386947855133e-05,
      "loss": 0.8071,
      "step": 4263
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6715628625484956e-05,
      "loss": 0.7832,
      "step": 4264
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6675895310570288e-05,
      "loss": 0.8418,
      "step": 4265
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6636187016669624e-05,
      "loss": 0.7934,
      "step": 4266
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6596503757332934e-05,
      "loss": 0.7702,
      "step": 4267
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.655684554610176e-05,
      "loss": 0.7773,
      "step": 4268
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.651721239650896e-05,
      "loss": 0.7618,
      "step": 4269
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6477604322078874e-05,
      "loss": 0.7841,
      "step": 4270
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.643802133632738e-05,
      "loss": 0.3571,
      "step": 4271
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.639846345276166e-05,
      "loss": 0.7058,
      "step": 4272
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.635893068488038e-05,
      "loss": 0.773,
      "step": 4273
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6319423046173698e-05,
      "loss": 0.7608,
      "step": 4274
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6279940550123117e-05,
      "loss": 0.8042,
      "step": 4275
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.624048321020156e-05,
      "loss": 0.7374,
      "step": 4276
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6201051039873458e-05,
      "loss": 0.7843,
      "step": 4277
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6161644052594526e-05,
      "loss": 0.7917,
      "step": 4278
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6122262261812004e-05,
      "loss": 0.7605,
      "step": 4279
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6082905680964455e-05,
      "loss": 0.8449,
      "step": 4280
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6043574323481823e-05,
      "loss": 0.826,
      "step": 4281
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6004268202785553e-05,
      "loss": 0.8151,
      "step": 4282
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5964987332288383e-05,
      "loss": 0.784,
      "step": 4283
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.592573172539443e-05,
      "loss": 0.8289,
      "step": 4284
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5886501395499286e-05,
      "loss": 0.746,
      "step": 4285
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5847296355989802e-05,
      "loss": 0.8233,
      "step": 4286
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5808116620244248e-05,
      "loss": 0.799,
      "step": 4287
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5768962201632306e-05,
      "loss": 0.8405,
      "step": 4288
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5729833113514947e-05,
      "loss": 0.7977,
      "step": 4289
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.569072936924449e-05,
      "loss": 0.3597,
      "step": 4290
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.565165098216472e-05,
      "loss": 0.7377,
      "step": 4291
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.561259796561064e-05,
      "loss": 0.3362,
      "step": 4292
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5573570332908637e-05,
      "loss": 0.7884,
      "step": 4293
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5534568097376487e-05,
      "loss": 0.7483,
      "step": 4294
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.549559127232324e-05,
      "loss": 0.7361,
      "step": 4295
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5456639871049248e-05,
      "loss": 0.7541,
      "step": 4296
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5417713906846295e-05,
      "loss": 0.8034,
      "step": 4297
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5378813392997413e-05,
      "loss": 0.7379,
      "step": 4298
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5339938342776915e-05,
      "loss": 0.7583,
      "step": 4299
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.530108876945052e-05,
      "loss": 0.8535,
      "step": 4300
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.526226468627515e-05,
      "loss": 0.7479,
      "step": 4301
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.522346610649914e-05,
      "loss": 0.7755,
      "step": 4302
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.518469304336203e-05,
      "loss": 0.7917,
      "step": 4303
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5145945510094658e-05,
      "loss": 0.7879,
      "step": 4304
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5107223519919244e-05,
      "loss": 0.8005,
      "step": 4305
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5068527086049175e-05,
      "loss": 0.7879,
      "step": 4306
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.502985622168916e-05,
      "loss": 0.807,
      "step": 4307
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4991210940035237e-05,
      "loss": 0.8347,
      "step": 4308
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4952591254274636e-05,
      "loss": 0.7599,
      "step": 4309
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.491399717758587e-05,
      "loss": 0.7786,
      "step": 4310
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4875428723138772e-05,
      "loss": 0.8362,
      "step": 4311
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4836885904094353e-05,
      "loss": 0.797,
      "step": 4312
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4798368733604904e-05,
      "loss": 0.8098,
      "step": 4313
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.475987722481401e-05,
      "loss": 0.7475,
      "step": 4314
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4721411390856418e-05,
      "loss": 0.8367,
      "step": 4315
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4682971244858155e-05,
      "loss": 0.7625,
      "step": 4316
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4644556799936524e-05,
      "loss": 0.7692,
      "step": 4317
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4606168069199997e-05,
      "loss": 0.7638,
      "step": 4318
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.456780506574825e-05,
      "loss": 0.7911,
      "step": 4319
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.452946780267228e-05,
      "loss": 0.7945,
      "step": 4320
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4491156293054206e-05,
      "loss": 0.8028,
      "step": 4321
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.445287054996739e-05,
      "loss": 0.783,
      "step": 4322
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4414610586476426e-05,
      "loss": 0.7982,
      "step": 4323
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.437637641563706e-05,
      "loss": 0.7735,
      "step": 4324
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4338168050496312e-05,
      "loss": 0.7502,
      "step": 4325
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4299985504092336e-05,
      "loss": 0.7843,
      "step": 4326
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4261828789454445e-05,
      "loss": 0.787,
      "step": 4327
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.422369791960324e-05,
      "loss": 0.7222,
      "step": 4328
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4185592907550425e-05,
      "loss": 0.7855,
      "step": 4329
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4147513766298867e-05,
      "loss": 0.8213,
      "step": 4330
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4109460508842697e-05,
      "loss": 0.8536,
      "step": 4331
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.407143314816713e-05,
      "loss": 0.8398,
      "step": 4332
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4033431697248544e-05,
      "loss": 0.8059,
      "step": 4333
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3995456169054543e-05,
      "loss": 0.8334,
      "step": 4334
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.395750657654382e-05,
      "loss": 0.8015,
      "step": 4335
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3919582932666218e-05,
      "loss": 0.7594,
      "step": 4336
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3881685250362794e-05,
      "loss": 0.7613,
      "step": 4337
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3843813542565686e-05,
      "loss": 0.7193,
      "step": 4338
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.380596782219814e-05,
      "loss": 0.8139,
      "step": 4339
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3768148102174635e-05,
      "loss": 0.764,
      "step": 4340
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.37303543954007e-05,
      "loss": 0.7505,
      "step": 4341
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.369258671477297e-05,
      "loss": 0.767,
      "step": 4342
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.36548450731793e-05,
      "loss": 0.8163,
      "step": 4343
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3617129483498523e-05,
      "loss": 0.8066,
      "step": 4344
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3579439958600725e-05,
      "loss": 0.7338,
      "step": 4345
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.354177651134698e-05,
      "loss": 0.8315,
      "step": 4346
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.35041391545895e-05,
      "loss": 0.7393,
      "step": 4347
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3466527901171655e-05,
      "loss": 0.8259,
      "step": 4348
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.342894276392783e-05,
      "loss": 0.8339,
      "step": 4349
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.339138375568348e-05,
      "loss": 0.8135,
      "step": 4350
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3353850889255256e-05,
      "loss": 0.7945,
      "step": 4351
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.33163441774508e-05,
      "loss": 0.7813,
      "step": 4352
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3278863633068813e-05,
      "loss": 0.76,
      "step": 4353
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3241409268899163e-05,
      "loss": 0.7196,
      "step": 4354
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3203981097722684e-05,
      "loss": 0.8133,
      "step": 4355
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3166579132311295e-05,
      "loss": 0.7528,
      "step": 4356
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3129203385428045e-05,
      "loss": 0.813,
      "step": 4357
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3091853869826953e-05,
      "loss": 0.8182,
      "step": 4358
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.305453059825309e-05,
      "loss": 0.7605,
      "step": 4359
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3017233583442644e-05,
      "loss": 0.7839,
      "step": 4360
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.297996283812278e-05,
      "loss": 0.8064,
      "step": 4361
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2942718375011673e-05,
      "loss": 0.8263,
      "step": 4362
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.290550020681863e-05,
      "loss": 0.8048,
      "step": 4363
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2868308346243905e-05,
      "loss": 0.7688,
      "step": 4364
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2831142805978765e-05,
      "loss": 0.7643,
      "step": 4365
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.279400359870558e-05,
      "loss": 0.8022,
      "step": 4366
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2756890737097623e-05,
      "loss": 0.7872,
      "step": 4367
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2719804233819298e-05,
      "loss": 0.3684,
      "step": 4368
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.268274410152592e-05,
      "loss": 0.8029,
      "step": 4369
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2645710352863815e-05,
      "loss": 0.7228,
      "step": 4370
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2608703000470365e-05,
      "loss": 0.7915,
      "step": 4371
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.257172205697389e-05,
      "loss": 0.3347,
      "step": 4372
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2534767534993683e-05,
      "loss": 0.3354,
      "step": 4373
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2497839447140124e-05,
      "loss": 0.7651,
      "step": 4374
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2460937806014447e-05,
      "loss": 0.7715,
      "step": 4375
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.242406262420892e-05,
      "loss": 0.8497,
      "step": 4376
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2387213914306805e-05,
      "loss": 0.798,
      "step": 4377
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2350391688882288e-05,
      "loss": 0.3516,
      "step": 4378
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.23135959605005e-05,
      "loss": 0.8487,
      "step": 4379
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.227682674171764e-05,
      "loss": 0.7964,
      "step": 4380
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2240084045080723e-05,
      "loss": 0.8491,
      "step": 4381
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2203367883127767e-05,
      "loss": 0.8202,
      "step": 4382
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.216667826838781e-05,
      "loss": 0.7673,
      "step": 4383
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2130015213380706e-05,
      "loss": 0.8006,
      "step": 4384
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.209337873061731e-05,
      "loss": 0.7784,
      "step": 4385
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.205676883259943e-05,
      "loss": 0.7929,
      "step": 4386
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2020185531819758e-05,
      "loss": 0.7562,
      "step": 4387
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1983628840761916e-05,
      "loss": 0.7973,
      "step": 4388
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1947098771900498e-05,
      "loss": 0.78,
      "step": 4389
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1910595337700912e-05,
      "loss": 0.8764,
      "step": 4390
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1874118550619606e-05,
      "loss": 0.7881,
      "step": 4391
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1837668423103828e-05,
      "loss": 0.7621,
      "step": 4392
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.180124496759176e-05,
      "loss": 0.8328,
      "step": 4393
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.176484819651252e-05,
      "loss": 0.3035,
      "step": 4394
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.172847812228609e-05,
      "loss": 0.7647,
      "step": 4395
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.169213475732329e-05,
      "loss": 0.7619,
      "step": 4396
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1655818114025962e-05,
      "loss": 0.8104,
      "step": 4397
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1619528204786687e-05,
      "loss": 0.7837,
      "step": 4398
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1583265041988976e-05,
      "loss": 0.7053,
      "step": 4399
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.154702863800726e-05,
      "loss": 0.7933,
      "step": 4400
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1510819005206796e-05,
      "loss": 0.8218,
      "step": 4401
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.147463615594365e-05,
      "loss": 0.7806,
      "step": 4402
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1438480102564885e-05,
      "loss": 0.7864,
      "step": 4403
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1402350857408295e-05,
      "loss": 0.8155,
      "step": 4404
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.136624843280256e-05,
      "loss": 0.8034,
      "step": 4405
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1330172841067265e-05,
      "loss": 0.3466,
      "step": 4406
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.129412409451278e-05,
      "loss": 0.7471,
      "step": 4407
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1258102205440277e-05,
      "loss": 0.7844,
      "step": 4408
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1222107186141904e-05,
      "loss": 0.7839,
      "step": 4409
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1186139048900466e-05,
      "loss": 0.7963,
      "step": 4410
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1150197805989747e-05,
      "loss": 0.7798,
      "step": 4411
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1114283469674256e-05,
      "loss": 0.7612,
      "step": 4412
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1078396052209326e-05,
      "loss": 0.7361,
      "step": 4413
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1042535565841183e-05,
      "loss": 0.7581,
      "step": 4414
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1006702022806767e-05,
      "loss": 0.8235,
      "step": 4415
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.097089543533386e-05,
      "loss": 0.8225,
      "step": 4416
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0935115815641103e-05,
      "loss": 0.7379,
      "step": 4417
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0899363175937838e-05,
      "loss": 0.7917,
      "step": 4418
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.086363752842424e-05,
      "loss": 0.8083,
      "step": 4419
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.082793888529131e-05,
      "loss": 0.7879,
      "step": 4420
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.07922672587208e-05,
      "loss": 0.7034,
      "step": 4421
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0756622660885193e-05,
      "loss": 0.745,
      "step": 4422
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0721005103947877e-05,
      "loss": 0.8119,
      "step": 4423
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0685414600062902e-05,
      "loss": 0.8156,
      "step": 4424
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.06498511613751e-05,
      "loss": 0.7743,
      "step": 4425
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.061431480002013e-05,
      "loss": 0.7759,
      "step": 4426
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0578805528124345e-05,
      "loss": 0.7623,
      "step": 4427
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0543323357804857e-05,
      "loss": 0.8036,
      "step": 4428
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0507868301169608e-05,
      "loss": 0.7117,
      "step": 4429
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0472440370317204e-05,
      "loss": 0.8041,
      "step": 4430
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.043703957733698e-05,
      "loss": 0.7933,
      "step": 4431
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0401665934309134e-05,
      "loss": 0.7386,
      "step": 4432
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0366319453304437e-05,
      "loss": 0.773,
      "step": 4433
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0331000146384548e-05,
      "loss": 0.3627,
      "step": 4434
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0295708025601735e-05,
      "loss": 0.8184,
      "step": 4435
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0260443102999018e-05,
      "loss": 0.772,
      "step": 4436
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.022520539061019e-05,
      "loss": 0.7638,
      "step": 4437
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0189994900459698e-05,
      "loss": 0.755,
      "step": 4438
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0154811644562687e-05,
      "loss": 0.7799,
      "step": 4439
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0119655634925106e-05,
      "loss": 0.838,
      "step": 4440
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0084526883543508e-05,
      "loss": 0.7731,
      "step": 4441
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0049425402405153e-05,
      "loss": 0.7868,
      "step": 4442
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.001435120348808e-05,
      "loss": 0.7891,
      "step": 4443
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9979304298760916e-05,
      "loss": 0.7805,
      "step": 4444
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9944284700183012e-05,
      "loss": 0.8361,
      "step": 4445
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.990929241970445e-05,
      "loss": 0.8735,
      "step": 4446
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9874327469265907e-05,
      "loss": 0.8432,
      "step": 4447
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9839389860798773e-05,
      "loss": 0.7652,
      "step": 4448
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9804479606225145e-05,
      "loss": 0.835,
      "step": 4449
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9769596717457717e-05,
      "loss": 0.8099,
      "step": 4450
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9734741206399864e-05,
      "loss": 0.7973,
      "step": 4451
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9699913084945675e-05,
      "loss": 0.719,
      "step": 4452
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.966511236497983e-05,
      "loss": 0.7564,
      "step": 4453
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9630339058377657e-05,
      "loss": 0.819,
      "step": 4454
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.959559317700518e-05,
      "loss": 0.8135,
      "step": 4455
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9560874732718993e-05,
      "loss": 0.8151,
      "step": 4456
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9526183737366432e-05,
      "loss": 0.8227,
      "step": 4457
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9491520202785362e-05,
      "loss": 0.8291,
      "step": 4458
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.94568841408043e-05,
      "loss": 0.7688,
      "step": 4459
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9422275563242463e-05,
      "loss": 0.7779,
      "step": 4460
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.938769448190959e-05,
      "loss": 0.7847,
      "step": 4461
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9353140908606082e-05,
      "loss": 0.3507,
      "step": 4462
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9318614855122986e-05,
      "loss": 0.8366,
      "step": 4463
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9284116333241885e-05,
      "loss": 0.7769,
      "step": 4464
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9249645354735012e-05,
      "loss": 0.7885,
      "step": 4465
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9215201931365223e-05,
      "loss": 0.7981,
      "step": 4466
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9180786074885925e-05,
      "loss": 0.7774,
      "step": 4467
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9146397797041105e-05,
      "loss": 0.836,
      "step": 4468
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9112037109565427e-05,
      "loss": 0.7949,
      "step": 4469
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.907770402418406e-05,
      "loss": 0.7916,
      "step": 4470
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9043398552612758e-05,
      "loss": 0.3507,
      "step": 4471
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9009120706557905e-05,
      "loss": 0.84,
      "step": 4472
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8974870497716412e-05,
      "loss": 0.7867,
      "step": 4473
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8940647937775734e-05,
      "loss": 0.759,
      "step": 4474
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8906453038413995e-05,
      "loss": 0.823,
      "step": 4475
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8872285811299783e-05,
      "loss": 0.8169,
      "step": 4476
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.883814626809225e-05,
      "loss": 0.7904,
      "step": 4477
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8804034420441176e-05,
      "loss": 0.867,
      "step": 4478
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8769950279986793e-05,
      "loss": 0.764,
      "step": 4479
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8735893858359955e-05,
      "loss": 0.8335,
      "step": 4480
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8701865167182032e-05,
      "loss": 0.786,
      "step": 4481
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8667864218064878e-05,
      "loss": 0.7711,
      "step": 4482
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8633891022610996e-05,
      "loss": 0.8477,
      "step": 4483
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8599945592413303e-05,
      "loss": 0.7942,
      "step": 4484
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8566027939055296e-05,
      "loss": 0.7329,
      "step": 4485
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8532138074111018e-05,
      "loss": 0.76,
      "step": 4486
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8498276009144967e-05,
      "loss": 0.7619,
      "step": 4487
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.846444175571218e-05,
      "loss": 0.8303,
      "step": 4488
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.843063532535825e-05,
      "loss": 0.7936,
      "step": 4489
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8396856729619215e-05,
      "loss": 0.7354,
      "step": 4490
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.83631059800216e-05,
      "loss": 0.7897,
      "step": 4491
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.832938308808252e-05,
      "loss": 0.7856,
      "step": 4492
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8295688065309512e-05,
      "loss": 0.7952,
      "step": 4493
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8262020923200573e-05,
      "loss": 0.7898,
      "step": 4494
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8228381673244288e-05,
      "loss": 0.802,
      "step": 4495
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8194770326919652e-05,
      "loss": 0.8325,
      "step": 4496
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8161186895696125e-05,
      "loss": 0.7957,
      "step": 4497
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8127631391033716e-05,
      "loss": 0.3149,
      "step": 4498
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8094103824382803e-05,
      "loss": 0.7499,
      "step": 4499
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.806060420718434e-05,
      "loss": 0.762,
      "step": 4500
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8027132550869673e-05,
      "loss": 0.7926,
      "step": 4501
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.799368886686058e-05,
      "loss": 0.7954,
      "step": 4502
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.796027316656941e-05,
      "loss": 0.7311,
      "step": 4503
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7926885461398845e-05,
      "loss": 0.7952,
      "step": 4504
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7893525762742035e-05,
      "loss": 0.7237,
      "step": 4505
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.786019408198265e-05,
      "loss": 0.8822,
      "step": 4506
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7826890430494724e-05,
      "loss": 0.7607,
      "step": 4507
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7793614819642724e-05,
      "loss": 0.8483,
      "step": 4508
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7760367260781617e-05,
      "loss": 0.7396,
      "step": 4509
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7727147765256724e-05,
      "loss": 0.7706,
      "step": 4510
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7693956344403806e-05,
      "loss": 0.8227,
      "step": 4511
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7660793009549103e-05,
      "loss": 0.7978,
      "step": 4512
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.76276577720092e-05,
      "loss": 0.8119,
      "step": 4513
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.759455064309108e-05,
      "loss": 0.8127,
      "step": 4514
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.756147163409223e-05,
      "loss": 0.8323,
      "step": 4515
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.752842075630048e-05,
      "loss": 0.7846,
      "step": 4516
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7495398020994013e-05,
      "loss": 0.7985,
      "step": 4517
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7462403439441522e-05,
      "loss": 0.7632,
      "step": 4518
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7429437022901996e-05,
      "loss": 0.8106,
      "step": 4519
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.739649878262485e-05,
      "loss": 0.8201,
      "step": 4520
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.73635887298499e-05,
      "loss": 0.7987,
      "step": 4521
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.73307068758073e-05,
      "loss": 0.3251,
      "step": 4522
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7297853231717644e-05,
      "loss": 0.7016,
      "step": 4523
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7265027808791844e-05,
      "loss": 0.3526,
      "step": 4524
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.723223061823117e-05,
      "loss": 0.7826,
      "step": 4525
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7199461671227356e-05,
      "loss": 0.7655,
      "step": 4526
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7166720978962382e-05,
      "loss": 0.8372,
      "step": 4527
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7134008552608626e-05,
      "loss": 0.7813,
      "step": 4528
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7101324403328877e-05,
      "loss": 0.7519,
      "step": 4529
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.70686685422762e-05,
      "loss": 0.8034,
      "step": 4530
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7036040980594024e-05,
      "loss": 0.7937,
      "step": 4531
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7003441729416157e-05,
      "loss": 0.7201,
      "step": 4532
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6970870799866722e-05,
      "loss": 0.7658,
      "step": 4533
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6938328203060127e-05,
      "loss": 0.7796,
      "step": 4534
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6905813950101233e-05,
      "loss": 0.8237,
      "step": 4535
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6873328052085123e-05,
      "loss": 0.8114,
      "step": 4536
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6840870520097218e-05,
      "loss": 0.8159,
      "step": 4537
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6808441365213324e-05,
      "loss": 0.8241,
      "step": 4538
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.677604059849951e-05,
      "loss": 0.7792,
      "step": 4539
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6743668231012122e-05,
      "loss": 0.8125,
      "step": 4540
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6711324273797924e-05,
      "loss": 0.7955,
      "step": 4541
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6679008737893896e-05,
      "loss": 0.7747,
      "step": 4542
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6646721634327313e-05,
      "loss": 0.7959,
      "step": 4543
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6614462974115842e-05,
      "loss": 0.794,
      "step": 4544
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6582232768267326e-05,
      "loss": 0.7533,
      "step": 4545
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.655003102778001e-05,
      "loss": 0.7949,
      "step": 4546
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6517857763642342e-05,
      "loss": 0.802,
      "step": 4547
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.648571298683306e-05,
      "loss": 0.7551,
      "step": 4548
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.645359670832124e-05,
      "loss": 0.7322,
      "step": 4549
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6421508939066178e-05,
      "loss": 0.7777,
      "step": 4550
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6389449690017444e-05,
      "loss": 0.8171,
      "step": 4551
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6357418972114926e-05,
      "loss": 0.8014,
      "step": 4552
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.632541679628873e-05,
      "loss": 0.7846,
      "step": 4553
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6293443173459188e-05,
      "loss": 0.795,
      "step": 4554
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.626149811453701e-05,
      "loss": 0.6918,
      "step": 4555
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.622958163042304e-05,
      "loss": 0.7699,
      "step": 4556
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.619769373200839e-05,
      "loss": 0.7833,
      "step": 4557
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6165834430174498e-05,
      "loss": 0.7177,
      "step": 4558
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6134003735792957e-05,
      "loss": 0.7561,
      "step": 4559
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6102201659725603e-05,
      "loss": 0.8154,
      "step": 4560
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6070428212824585e-05,
      "loss": 0.746,
      "step": 4561
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6038683405932208e-05,
      "loss": 0.823,
      "step": 4562
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6006967249880988e-05,
      "loss": 0.7599,
      "step": 4563
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5975279755493767e-05,
      "loss": 0.7967,
      "step": 4564
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.594362093358347e-05,
      "loss": 0.7657,
      "step": 4565
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.591199079495338e-05,
      "loss": 0.7513,
      "step": 4566
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5880389350396884e-05,
      "loss": 0.8224,
      "step": 4567
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5848816610697592e-05,
      "loss": 0.8345,
      "step": 4568
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5817272586629393e-05,
      "loss": 0.8116,
      "step": 4569
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5785757288956304e-05,
      "loss": 0.813,
      "step": 4570
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.575427072843252e-05,
      "loss": 0.7799,
      "step": 4571
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5722812915802533e-05,
      "loss": 0.7895,
      "step": 4572
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5691383861800933e-05,
      "loss": 0.3238,
      "step": 4573
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5659983577152515e-05,
      "loss": 0.7865,
      "step": 4574
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5628612072572303e-05,
      "loss": 0.7947,
      "step": 4575
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5597269358765432e-05,
      "loss": 0.7449,
      "step": 4576
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5565955446427238e-05,
      "loss": 0.812,
      "step": 4577
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5534670346243284e-05,
      "loss": 0.8363,
      "step": 4578
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5503414068889223e-05,
      "loss": 0.3337,
      "step": 4579
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5472186625030873e-05,
      "loss": 0.3231,
      "step": 4580
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.54409880253243e-05,
      "loss": 0.8536,
      "step": 4581
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5409818280415657e-05,
      "loss": 0.7945,
      "step": 4582
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5378677400941212e-05,
      "loss": 0.8298,
      "step": 4583
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5347565397527507e-05,
      "loss": 0.8162,
      "step": 4584
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5316482280791113e-05,
      "loss": 0.3287,
      "step": 4585
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.528542806133879e-05,
      "loss": 0.738,
      "step": 4586
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.525440274976746e-05,
      "loss": 0.7694,
      "step": 4587
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5223406356664128e-05,
      "loss": 0.7643,
      "step": 4588
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5192438892605998e-05,
      "loss": 0.7964,
      "step": 4589
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5161500368160342e-05,
      "loss": 0.8451,
      "step": 4590
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5130590793884547e-05,
      "loss": 0.7848,
      "step": 4591
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5099710180326209e-05,
      "loss": 0.8042,
      "step": 4592
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5068858538022957e-05,
      "loss": 0.7386,
      "step": 4593
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5038035877502543e-05,
      "loss": 0.8307,
      "step": 4594
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.500724220928289e-05,
      "loss": 0.7854,
      "step": 4595
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4976477543871958e-05,
      "loss": 0.782,
      "step": 4596
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4945741891767829e-05,
      "loss": 0.8234,
      "step": 4597
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4915035263458732e-05,
      "loss": 0.8203,
      "step": 4598
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4884357669422932e-05,
      "loss": 0.3091,
      "step": 4599
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4853709120128778e-05,
      "loss": 0.7864,
      "step": 4600
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4823089626034793e-05,
      "loss": 0.7666,
      "step": 4601
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4792499197589505e-05,
      "loss": 0.748,
      "step": 4602
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4761937845231521e-05,
      "loss": 0.7773,
      "step": 4603
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.47314055793896e-05,
      "loss": 0.8139,
      "step": 4604
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4700902410482509e-05,
      "loss": 0.747,
      "step": 4605
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4670428348919086e-05,
      "loss": 0.776,
      "step": 4606
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.46399834050983e-05,
      "loss": 0.718,
      "step": 4607
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4609567589409123e-05,
      "loss": 0.7857,
      "step": 4608
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4579180912230572e-05,
      "loss": 0.7789,
      "step": 4609
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4548823383931798e-05,
      "loss": 0.814,
      "step": 4610
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4518495014871925e-05,
      "loss": 0.7514,
      "step": 4611
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4488195815400208e-05,
      "loss": 0.8172,
      "step": 4612
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4457925795855875e-05,
      "loss": 0.8315,
      "step": 4613
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.44276849665682e-05,
      "loss": 0.8057,
      "step": 4614
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4397473337856582e-05,
      "loss": 0.7688,
      "step": 4615
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4367290920030351e-05,
      "loss": 0.7975,
      "step": 4616
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4337137723388905e-05,
      "loss": 0.8154,
      "step": 4617
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4307013758221722e-05,
      "loss": 0.8051,
      "step": 4618
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4276919034808223e-05,
      "loss": 0.7341,
      "step": 4619
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4246853563417884e-05,
      "loss": 0.7673,
      "step": 4620
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4216817354310242e-05,
      "loss": 0.7817,
      "step": 4621
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4186810417734798e-05,
      "loss": 0.8039,
      "step": 4622
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4156832763931038e-05,
      "loss": 0.7916,
      "step": 4623
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4126884403128548e-05,
      "loss": 0.7849,
      "step": 4624
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4096965345546842e-05,
      "loss": 0.8084,
      "step": 4625
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.406707560139543e-05,
      "loss": 0.8231,
      "step": 4626
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4037215180873897e-05,
      "loss": 0.7738,
      "step": 4627
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.400738409417175e-05,
      "loss": 0.8091,
      "step": 4628
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3977582351468477e-05,
      "loss": 0.8298,
      "step": 4629
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3947809962933623e-05,
      "loss": 0.7813,
      "step": 4630
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3918066938726647e-05,
      "loss": 0.7867,
      "step": 4631
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3888353288997047e-05,
      "loss": 0.8121,
      "step": 4632
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3858669023884263e-05,
      "loss": 0.7921,
      "step": 4633
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3829014153517662e-05,
      "loss": 0.7755,
      "step": 4634
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3799388688016702e-05,
      "loss": 0.7127,
      "step": 4635
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3769792637490697e-05,
      "loss": 0.7673,
      "step": 4636
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3740226012038937e-05,
      "loss": 0.799,
      "step": 4637
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.371068882175075e-05,
      "loss": 0.7598,
      "step": 4638
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.368118107670534e-05,
      "loss": 0.7994,
      "step": 4639
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3651702786971864e-05,
      "loss": 0.7962,
      "step": 4640
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.36222539626095e-05,
      "loss": 0.8144,
      "step": 4641
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3592834613667294e-05,
      "loss": 0.8193,
      "step": 4642
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.356344475018424e-05,
      "loss": 0.7515,
      "step": 4643
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.353408438218935e-05,
      "loss": 0.7948,
      "step": 4644
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3504753519701486e-05,
      "loss": 0.8003,
      "step": 4645
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3475452172729453e-05,
      "loss": 0.846,
      "step": 4646
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3446180351272031e-05,
      "loss": 0.8532,
      "step": 4647
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3416938065317886e-05,
      "loss": 0.8165,
      "step": 4648
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3387725324845579e-05,
      "loss": 0.7808,
      "step": 4649
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3358542139823694e-05,
      "loss": 0.7166,
      "step": 4650
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3329388520210606e-05,
      "loss": 0.7816,
      "step": 4651
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.330026447595466e-05,
      "loss": 0.7803,
      "step": 4652
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3271170016994128e-05,
      "loss": 0.7842,
      "step": 4653
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3242105153257134e-05,
      "loss": 0.7796,
      "step": 4654
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3213069894661756e-05,
      "loss": 0.7547,
      "step": 4655
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3184064251115946e-05,
      "loss": 0.7709,
      "step": 4656
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.315508823251751e-05,
      "loss": 0.343,
      "step": 4657
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3126141848754237e-05,
      "loss": 0.7991,
      "step": 4658
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3097225109703715e-05,
      "loss": 0.7747,
      "step": 4659
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3068338025233439e-05,
      "loss": 0.7931,
      "step": 4660
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.303948060520085e-05,
      "loss": 0.7789,
      "step": 4661
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3010652859453177e-05,
      "loss": 0.7848,
      "step": 4662
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2981854797827542e-05,
      "loss": 0.739,
      "step": 4663
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2953086430150996e-05,
      "loss": 0.806,
      "step": 4664
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2924347766240396e-05,
      "loss": 0.7572,
      "step": 4665
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2895638815902455e-05,
      "loss": 0.8342,
      "step": 4666
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2866959588933836e-05,
      "loss": 0.7833,
      "step": 4667
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2838310095120964e-05,
      "loss": 0.8156,
      "step": 4668
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2809690344240132e-05,
      "loss": 0.8334,
      "step": 4669
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.278110034605755e-05,
      "loss": 0.7805,
      "step": 4670
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2752540110329192e-05,
      "loss": 0.6638,
      "step": 4671
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2724009646800916e-05,
      "loss": 0.7702,
      "step": 4672
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2695508965208435e-05,
      "loss": 0.7759,
      "step": 4673
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2667038075277271e-05,
      "loss": 0.8268,
      "step": 4674
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2638596986722784e-05,
      "loss": 0.7789,
      "step": 4675
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2610185709250188e-05,
      "loss": 0.8228,
      "step": 4676
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2581804252554474e-05,
      "loss": 0.7445,
      "step": 4677
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2553452626320539e-05,
      "loss": 0.8367,
      "step": 4678
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2525130840223021e-05,
      "loss": 0.8019,
      "step": 4679
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.249683890392639e-05,
      "loss": 0.8009,
      "step": 4680
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2468576827084988e-05,
      "loss": 0.7588,
      "step": 4681
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2440344619342914e-05,
      "loss": 0.8214,
      "step": 4682
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.241214229033405e-05,
      "loss": 0.7537,
      "step": 4683
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2383969849682175e-05,
      "loss": 0.7906,
      "step": 4684
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2355827307000778e-05,
      "loss": 0.789,
      "step": 4685
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2327714671893175e-05,
      "loss": 0.7824,
      "step": 4686
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2299631953952517e-05,
      "loss": 0.794,
      "step": 4687
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.227157916276168e-05,
      "loss": 0.7664,
      "step": 4688
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2243556307893356e-05,
      "loss": 0.8487,
      "step": 4689
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2215563398910045e-05,
      "loss": 0.7665,
      "step": 4690
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2187600445364011e-05,
      "loss": 0.8668,
      "step": 4691
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2159667456797253e-05,
      "loss": 0.7851,
      "step": 4692
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2131764442741632e-05,
      "loss": 0.8186,
      "step": 4693
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2103891412718726e-05,
      "loss": 0.7822,
      "step": 4694
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2076048376239845e-05,
      "loss": 0.752,
      "step": 4695
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2048235342806158e-05,
      "loss": 0.3349,
      "step": 4696
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2020452321908515e-05,
      "loss": 0.82,
      "step": 4697
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1992699323027578e-05,
      "loss": 0.7233,
      "step": 4698
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1964976355633728e-05,
      "loss": 0.7628,
      "step": 4699
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1937283429187097e-05,
      "loss": 0.825,
      "step": 4700
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1909620553137612e-05,
      "loss": 0.8018,
      "step": 4701
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1881987736924893e-05,
      "loss": 0.7695,
      "step": 4702
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1854384989978318e-05,
      "loss": 0.7765,
      "step": 4703
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.182681232171703e-05,
      "loss": 0.7362,
      "step": 4704
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1799269741549879e-05,
      "loss": 0.3412,
      "step": 4705
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1771757258875426e-05,
      "loss": 0.8082,
      "step": 4706
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1744274883082052e-05,
      "loss": 0.7931,
      "step": 4707
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1716822623547762e-05,
      "loss": 0.7987,
      "step": 4708
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1689400489640323e-05,
      "loss": 0.7341,
      "step": 4709
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1662008490717259e-05,
      "loss": 0.3088,
      "step": 4710
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1634646636125746e-05,
      "loss": 0.7644,
      "step": 4711
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1607314935202695e-05,
      "loss": 0.7846,
      "step": 4712
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1580013397274781e-05,
      "loss": 0.7566,
      "step": 4713
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1552742031658314e-05,
      "loss": 0.7537,
      "step": 4714
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1525500847659332e-05,
      "loss": 0.761,
      "step": 4715
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1498289854573597e-05,
      "loss": 0.762,
      "step": 4716
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1471109061686535e-05,
      "loss": 0.7083,
      "step": 4717
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1443958478273275e-05,
      "loss": 0.7843,
      "step": 4718
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.141683811359866e-05,
      "loss": 0.8283,
      "step": 4719
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1389747976917199e-05,
      "loss": 0.7965,
      "step": 4720
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1362688077473094e-05,
      "loss": 0.8295,
      "step": 4721
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1335658424500239e-05,
      "loss": 0.8085,
      "step": 4722
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1308659027222157e-05,
      "loss": 0.7865,
      "step": 4723
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1281689894852122e-05,
      "loss": 0.7802,
      "step": 4724
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1254751036593037e-05,
      "loss": 0.3242,
      "step": 4725
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1227842461637438e-05,
      "loss": 0.7346,
      "step": 4726
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.120096417916764e-05,
      "loss": 0.7717,
      "step": 4727
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.11741161983555e-05,
      "loss": 0.8032,
      "step": 4728
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1147298528362593e-05,
      "loss": 0.7582,
      "step": 4729
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1120511178340154e-05,
      "loss": 0.8003,
      "step": 4730
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1093754157429059e-05,
      "loss": 0.8312,
      "step": 4731
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1067027474759817e-05,
      "loss": 0.8053,
      "step": 4732
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1040331139452631e-05,
      "loss": 0.3471,
      "step": 4733
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1013665160617303e-05,
      "loss": 0.7556,
      "step": 4734
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.098702954735329e-05,
      "loss": 0.7895,
      "step": 4735
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0960424308749718e-05,
      "loss": 0.8124,
      "step": 4736
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0933849453885303e-05,
      "loss": 0.7442,
      "step": 4737
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0907304991828382e-05,
      "loss": 0.7527,
      "step": 4738
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0880790931637009e-05,
      "loss": 0.8202,
      "step": 4739
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0854307282358744e-05,
      "loss": 0.7716,
      "step": 4740
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0827854053030894e-05,
      "loss": 0.7834,
      "step": 4741
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0801431252680283e-05,
      "loss": 0.7995,
      "step": 4742
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0775038890323375e-05,
      "loss": 0.7919,
      "step": 4743
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0748676974966298e-05,
      "loss": 0.7297,
      "step": 4744
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0722345515604737e-05,
      "loss": 0.7798,
      "step": 4745
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0696044521223992e-05,
      "loss": 0.8151,
      "step": 4746
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0669774000799004e-05,
      "loss": 0.7961,
      "step": 4747
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0643533963294283e-05,
      "loss": 0.7566,
      "step": 4748
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.06173244176639e-05,
      "loss": 0.7405,
      "step": 4749
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0591145372851629e-05,
      "loss": 0.7567,
      "step": 4750
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.056499683779073e-05,
      "loss": 0.8011,
      "step": 4751
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0538878821404085e-05,
      "loss": 0.6975,
      "step": 4752
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.05127913326042e-05,
      "loss": 0.7301,
      "step": 4753
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0486734380293118e-05,
      "loss": 0.7731,
      "step": 4754
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.046070797336246e-05,
      "loss": 0.8222,
      "step": 4755
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0434712120693469e-05,
      "loss": 0.776,
      "step": 4756
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0408746831156902e-05,
      "loss": 0.8123,
      "step": 4757
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.038281211361316e-05,
      "loss": 0.7545,
      "step": 4758
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0356907976912144e-05,
      "loss": 0.8034,
      "step": 4759
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0331034429893327e-05,
      "loss": 0.7654,
      "step": 4760
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0305191481385801e-05,
      "loss": 0.7703,
      "step": 4761
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0279379140208157e-05,
      "loss": 0.8039,
      "step": 4762
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0253597415168537e-05,
      "loss": 0.7748,
      "step": 4763
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0227846315064716e-05,
      "loss": 0.7531,
      "step": 4764
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.020212584868392e-05,
      "loss": 0.3313,
      "step": 4765
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0176436024802971e-05,
      "loss": 0.7993,
      "step": 4766
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0150776852188238e-05,
      "loss": 0.8095,
      "step": 4767
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0125148339595636e-05,
      "loss": 0.857,
      "step": 4768
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.009955049577055e-05,
      "loss": 0.8143,
      "step": 4769
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.007398332944801e-05,
      "loss": 0.7904,
      "step": 4770
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0048446849352499e-05,
      "loss": 0.7579,
      "step": 4771
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0022941064198022e-05,
      "loss": 0.7668,
      "step": 4772
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.99746598268818e-06,
      "loss": 0.7263,
      "step": 4773
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.972021613516048e-06,
      "loss": 0.7552,
      "step": 4774
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.946607965364185e-06,
      "loss": 0.8425,
      "step": 4775
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.921225046904759e-06,
      "loss": 0.7624,
      "step": 4776
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.89587286679935e-06,
      "loss": 0.3096,
      "step": 4777
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.87055143369916e-06,
      "loss": 0.7538,
      "step": 4778
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.845260756244811e-06,
      "loss": 0.777,
      "step": 4779
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.820000843066423e-06,
      "loss": 0.7618,
      "step": 4780
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.794771702783723e-06,
      "loss": 0.8125,
      "step": 4781
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.769573344005823e-06,
      "loss": 0.8032,
      "step": 4782
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.744405775331356e-06,
      "loss": 0.7565,
      "step": 4783
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.719269005348519e-06,
      "loss": 0.7654,
      "step": 4784
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.694163042634929e-06,
      "loss": 0.8202,
      "step": 4785
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.669087895757679e-06,
      "loss": 0.7564,
      "step": 4786
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.64404357327342e-06,
      "loss": 0.7816,
      "step": 4787
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.61903008372822e-06,
      "loss": 0.7562,
      "step": 4788
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.594047435657627e-06,
      "loss": 0.7833,
      "step": 4789
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.569095637586734e-06,
      "loss": 0.7953,
      "step": 4790
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.544174698030028e-06,
      "loss": 0.8053,
      "step": 4791
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.519284625491487e-06,
      "loss": 0.82,
      "step": 4792
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.494425428464581e-06,
      "loss": 0.7882,
      "step": 4793
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.469597115432205e-06,
      "loss": 0.7518,
      "step": 4794
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.444799694866779e-06,
      "loss": 0.7658,
      "step": 4795
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.420033175230103e-06,
      "loss": 0.7277,
      "step": 4796
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.395297564973471e-06,
      "loss": 0.7146,
      "step": 4797
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.370592872537654e-06,
      "loss": 0.754,
      "step": 4798
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.34591910635283e-06,
      "loss": 0.7665,
      "step": 4799
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.32127627483862e-06,
      "loss": 0.782,
      "step": 4800
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.296664386404152e-06,
      "loss": 0.7746,
      "step": 4801
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.272083449447922e-06,
      "loss": 0.3397,
      "step": 4802
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.247533472357884e-06,
      "loss": 0.7734,
      "step": 4803
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.223014463511482e-06,
      "loss": 0.8125,
      "step": 4804
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.198526431275512e-06,
      "loss": 0.853,
      "step": 4805
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.174069384006223e-06,
      "loss": 0.7353,
      "step": 4806
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.149643330049351e-06,
      "loss": 0.802,
      "step": 4807
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.125248277739972e-06,
      "loss": 0.794,
      "step": 4808
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.100884235402619e-06,
      "loss": 0.7705,
      "step": 4809
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.07655121135127e-06,
      "loss": 0.7656,
      "step": 4810
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.052249213889263e-06,
      "loss": 0.8339,
      "step": 4811
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.027978251309399e-06,
      "loss": 0.8246,
      "step": 4812
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.003738331893862e-06,
      "loss": 0.796,
      "step": 4813
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.979529463914226e-06,
      "loss": 0.3322,
      "step": 4814
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.955351655631516e-06,
      "loss": 0.6758,
      "step": 4815
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.931204915296143e-06,
      "loss": 0.7449,
      "step": 4816
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.90708925114786e-06,
      "loss": 0.7936,
      "step": 4817
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.883004671415906e-06,
      "loss": 0.7909,
      "step": 4818
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.858951184318864e-06,
      "loss": 0.7298,
      "step": 4819
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.834928798064679e-06,
      "loss": 0.7429,
      "step": 4820
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.81093752085076e-06,
      "loss": 0.8225,
      "step": 4821
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.786977360863835e-06,
      "loss": 0.7536,
      "step": 4822
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.76304832628002e-06,
      "loss": 0.7428,
      "step": 4823
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.739150425264852e-06,
      "loss": 0.7662,
      "step": 4824
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.715283665973217e-06,
      "loss": 0.7383,
      "step": 4825
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.691448056549345e-06,
      "loss": 0.7789,
      "step": 4826
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.667643605126896e-06,
      "loss": 0.7347,
      "step": 4827
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.64387031982884e-06,
      "loss": 0.781,
      "step": 4828
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.620128208767564e-06,
      "loss": 0.9035,
      "step": 4829
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.596417280044799e-06,
      "loss": 0.7257,
      "step": 4830
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.572737541751585e-06,
      "loss": 0.7676,
      "step": 4831
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.549089001968414e-06,
      "loss": 0.7761,
      "step": 4832
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.525471668765062e-06,
      "loss": 0.7467,
      "step": 4833
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.501885550200639e-06,
      "loss": 0.8312,
      "step": 4834
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.478330654323696e-06,
      "loss": 0.7906,
      "step": 4835
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.454806989172037e-06,
      "loss": 0.8799,
      "step": 4836
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.431314562772853e-06,
      "loss": 0.3319,
      "step": 4837
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.407853383142671e-06,
      "loss": 0.7303,
      "step": 4838
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.384423458287349e-06,
      "loss": 0.8144,
      "step": 4839
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.361024796202055e-06,
      "loss": 0.8318,
      "step": 4840
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.337657404871368e-06,
      "loss": 0.7028,
      "step": 4841
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.314321292269112e-06,
      "loss": 0.836,
      "step": 4842
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.291016466358448e-06,
      "loss": 0.8012,
      "step": 4843
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.267742935091926e-06,
      "loss": 0.7586,
      "step": 4844
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.244500706411362e-06,
      "loss": 0.7684,
      "step": 4845
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.221289788247865e-06,
      "loss": 0.8499,
      "step": 4846
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.198110188521946e-06,
      "loss": 0.7811,
      "step": 4847
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.174961915143341e-06,
      "loss": 0.7511,
      "step": 4848
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.151844976011158e-06,
      "loss": 0.8082,
      "step": 4849
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.128759379013785e-06,
      "loss": 0.7996,
      "step": 4850
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.105705132028895e-06,
      "loss": 0.7153,
      "step": 4851
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.082682242923522e-06,
      "loss": 0.7924,
      "step": 4852
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.059690719553958e-06,
      "loss": 0.296,
      "step": 4853
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.036730569765761e-06,
      "loss": 0.797,
      "step": 4854
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.01380180139386e-06,
      "loss": 0.7937,
      "step": 4855
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.990904422262424e-06,
      "loss": 0.7513,
      "step": 4856
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.968038440184911e-06,
      "loss": 0.7472,
      "step": 4857
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.945203862964102e-06,
      "loss": 0.806,
      "step": 4858
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.922400698392008e-06,
      "loss": 0.7313,
      "step": 4859
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.899628954249949e-06,
      "loss": 0.8152,
      "step": 4860
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.876888638308555e-06,
      "loss": 0.7139,
      "step": 4861
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.854179758327684e-06,
      "loss": 0.7776,
      "step": 4862
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.83150232205645e-06,
      "loss": 0.7926,
      "step": 4863
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.808856337233317e-06,
      "loss": 0.8279,
      "step": 4864
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.786241811585925e-06,
      "loss": 0.8033,
      "step": 4865
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.76365875283127e-06,
      "loss": 0.7892,
      "step": 4866
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.741107168675543e-06,
      "loss": 0.7566,
      "step": 4867
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.718587066814186e-06,
      "loss": 0.3498,
      "step": 4868
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.696098454931977e-06,
      "loss": 0.8145,
      "step": 4869
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.673641340702865e-06,
      "loss": 0.8297,
      "step": 4870
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.651215731790073e-06,
      "loss": 0.8655,
      "step": 4871
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.628821635846129e-06,
      "loss": 0.8095,
      "step": 4872
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.606459060512727e-06,
      "loss": 0.796,
      "step": 4873
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.584128013420833e-06,
      "loss": 0.8319,
      "step": 4874
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.561828502190704e-06,
      "loss": 0.7729,
      "step": 4875
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.5395605344317665e-06,
      "loss": 0.7503,
      "step": 4876
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.517324117742696e-06,
      "loss": 0.8059,
      "step": 4877
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.495119259711448e-06,
      "loss": 0.7859,
      "step": 4878
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.472945967915157e-06,
      "loss": 0.7844,
      "step": 4879
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.450804249920196e-06,
      "loss": 0.8274,
      "step": 4880
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.428694113282208e-06,
      "loss": 0.7862,
      "step": 4881
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.406615565545993e-06,
      "loss": 0.7663,
      "step": 4882
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.384568614245624e-06,
      "loss": 0.8433,
      "step": 4883
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.3625532669043704e-06,
      "loss": 0.6901,
      "step": 4884
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.340569531034691e-06,
      "loss": 0.7227,
      "step": 4885
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.318617414138318e-06,
      "loss": 0.7844,
      "step": 4886
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.2966969237061476e-06,
      "loss": 0.709,
      "step": 4887
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.274808067218286e-06,
      "loss": 0.7423,
      "step": 4888
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.25295085214408e-06,
      "loss": 0.7882,
      "step": 4889
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.2311252859420305e-06,
      "loss": 0.7756,
      "step": 4890
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.209331376059869e-06,
      "loss": 0.7296,
      "step": 4891
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.1875691299345464e-06,
      "loss": 0.7749,
      "step": 4892
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.165838554992144e-06,
      "loss": 0.7123,
      "step": 4893
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.144139658647975e-06,
      "loss": 0.7729,
      "step": 4894
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.122472448306572e-06,
      "loss": 0.8207,
      "step": 4895
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.1008369313615985e-06,
      "loss": 0.8218,
      "step": 4896
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.079233115195916e-06,
      "loss": 0.3329,
      "step": 4897
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.057661007181615e-06,
      "loss": 0.8104,
      "step": 4898
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.036120614679909e-06,
      "loss": 0.7435,
      "step": 4899
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.014611945041194e-06,
      "loss": 0.3406,
      "step": 4900
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.993135005605101e-06,
      "loss": 0.7802,
      "step": 4901
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.971689803700354e-06,
      "loss": 0.3263,
      "step": 4902
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.950276346644913e-06,
      "loss": 0.3258,
      "step": 4903
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.928894641745853e-06,
      "loss": 0.7793,
      "step": 4904
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.9075446962994396e-06,
      "loss": 0.7532,
      "step": 4905
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.88622651759111e-06,
      "loss": 0.7688,
      "step": 4906
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.864940112895446e-06,
      "loss": 0.801,
      "step": 4907
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.84368548947617e-06,
      "loss": 0.8418,
      "step": 4908
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.822462654586214e-06,
      "loss": 0.7814,
      "step": 4909
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.801271615467608e-06,
      "loss": 0.7743,
      "step": 4910
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.780112379351544e-06,
      "loss": 0.7798,
      "step": 4911
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.758984953458391e-06,
      "loss": 0.8095,
      "step": 4912
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.7378893449976276e-06,
      "loss": 0.7559,
      "step": 4913
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.716825561167883e-06,
      "loss": 0.8035,
      "step": 4914
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.695793609156953e-06,
      "loss": 0.8244,
      "step": 4915
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.674793496141751e-06,
      "loss": 0.2902,
      "step": 4916
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.653825229288302e-06,
      "loss": 0.808,
      "step": 4917
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.632888815751814e-06,
      "loss": 0.7585,
      "step": 4918
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.611984262676563e-06,
      "loss": 0.3367,
      "step": 4919
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.591111577196041e-06,
      "loss": 0.8202,
      "step": 4920
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.570270766432784e-06,
      "loss": 0.7319,
      "step": 4921
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.549461837498471e-06,
      "loss": 0.7765,
      "step": 4922
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.528684797493933e-06,
      "loss": 0.8252,
      "step": 4923
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.507939653509088e-06,
      "loss": 0.842,
      "step": 4924
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.487226412622971e-06,
      "loss": 0.8086,
      "step": 4925
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.466545081903763e-06,
      "loss": 0.8038,
      "step": 4926
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.445895668408708e-06,
      "loss": 0.7779,
      "step": 4927
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.4252781791841906e-06,
      "loss": 0.7925,
      "step": 4928
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.404692621265707e-06,
      "loss": 0.7741,
      "step": 4929
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.3841390016778245e-06,
      "loss": 0.8471,
      "step": 4930
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.363617327434235e-06,
      "loss": 0.7319,
      "step": 4931
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.343127605537735e-06,
      "loss": 0.799,
      "step": 4932
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.322669842980211e-06,
      "loss": 0.8234,
      "step": 4933
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.302244046742611e-06,
      "loss": 0.7531,
      "step": 4934
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.281850223795049e-06,
      "loss": 0.8363,
      "step": 4935
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.2614883810966475e-06,
      "loss": 0.3108,
      "step": 4936
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.241158525595681e-06,
      "loss": 0.7749,
      "step": 4937
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.220860664229467e-06,
      "loss": 0.8313,
      "step": 4938
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.200594803924398e-06,
      "loss": 0.7828,
      "step": 4939
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.180360951596009e-06,
      "loss": 0.8324,
      "step": 4940
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.1601591141488555e-06,
      "loss": 0.7773,
      "step": 4941
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.1399892984765674e-06,
      "loss": 0.7544,
      "step": 4942
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.1198515114618824e-06,
      "loss": 0.8366,
      "step": 4943
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.099745759976594e-06,
      "loss": 0.776,
      "step": 4944
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.079672050881524e-06,
      "loss": 0.3288,
      "step": 4945
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.0596303910266496e-06,
      "loss": 0.832,
      "step": 4946
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.0396207872509215e-06,
      "loss": 0.7558,
      "step": 4947
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.0196432463824e-06,
      "loss": 0.7524,
      "step": 4948
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.9996977752382e-06,
      "loss": 0.7585,
      "step": 4949
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.9797843806244755e-06,
      "loss": 0.3349,
      "step": 4950
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.959903069336437e-06,
      "loss": 0.7695,
      "step": 4951
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.940053848158378e-06,
      "loss": 0.8065,
      "step": 4952
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.9202367238636145e-06,
      "loss": 0.7419,
      "step": 4953
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.900451703214483e-06,
      "loss": 0.7453,
      "step": 4954
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.880698792962447e-06,
      "loss": 0.7952,
      "step": 4955
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.860977999847905e-06,
      "loss": 0.7784,
      "step": 4956
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.841289330600408e-06,
      "loss": 0.7264,
      "step": 4957
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.821632791938459e-06,
      "loss": 0.8106,
      "step": 4958
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.802008390569613e-06,
      "loss": 0.8089,
      "step": 4959
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.782416133190515e-06,
      "loss": 0.8089,
      "step": 4960
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.762856026486774e-06,
      "loss": 0.7839,
      "step": 4961
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.7433280771330475e-06,
      "loss": 0.7739,
      "step": 4962
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.723832291793041e-06,
      "loss": 0.7832,
      "step": 4963
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.704368677119465e-06,
      "loss": 0.7645,
      "step": 4964
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.684937239754029e-06,
      "loss": 0.7785,
      "step": 4965
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.665537986327541e-06,
      "loss": 0.8303,
      "step": 4966
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.646170923459737e-06,
      "loss": 0.7393,
      "step": 4967
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.6268360577594e-06,
      "loss": 0.3508,
      "step": 4968
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.607533395824372e-06,
      "loss": 0.7594,
      "step": 4969
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.5882629442414296e-06,
      "loss": 0.761,
      "step": 4970
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.569024709586401e-06,
      "loss": 0.7719,
      "step": 4971
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.549818698424136e-06,
      "loss": 0.8037,
      "step": 4972
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.530644917308436e-06,
      "loss": 0.7161,
      "step": 4973
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.511503372782157e-06,
      "loss": 0.7923,
      "step": 4974
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.492394071377138e-06,
      "loss": 0.7933,
      "step": 4975
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.473317019614177e-06,
      "loss": 0.8132,
      "step": 4976
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.454272224003132e-06,
      "loss": 0.7885,
      "step": 4977
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.435259691042816e-06,
      "loss": 0.7733,
      "step": 4978
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.416279427221016e-06,
      "loss": 0.7817,
      "step": 4979
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.397331439014563e-06,
      "loss": 0.7689,
      "step": 4980
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.37841573288923e-06,
      "loss": 0.8189,
      "step": 4981
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.359532315299765e-06,
      "loss": 0.7877,
      "step": 4982
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.340681192689956e-06,
      "loss": 0.7747,
      "step": 4983
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.321862371492514e-06,
      "loss": 0.8369,
      "step": 4984
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.3030758581291255e-06,
      "loss": 0.7883,
      "step": 4985
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.284321659010516e-06,
      "loss": 0.7482,
      "step": 4986
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.265599780536323e-06,
      "loss": 0.8017,
      "step": 4987
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.2469102290951585e-06,
      "loss": 0.7515,
      "step": 4988
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.228253011064643e-06,
      "loss": 0.3455,
      "step": 4989
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.209628132811317e-06,
      "loss": 0.7259,
      "step": 4990
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.19103560069073e-06,
      "loss": 0.3503,
      "step": 4991
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.172475421047373e-06,
      "loss": 0.8155,
      "step": 4992
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.153947600214659e-06,
      "loss": 0.7843,
      "step": 4993
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.135452144515029e-06,
      "loss": 0.8234,
      "step": 4994
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.116989060259847e-06,
      "loss": 0.3449,
      "step": 4995
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.098558353749405e-06,
      "loss": 0.7907,
      "step": 4996
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.080160031272996e-06,
      "loss": 0.7421,
      "step": 4997
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.06179409910883e-06,
      "loss": 0.7586,
      "step": 4998
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.04346056352406e-06,
      "loss": 0.842,
      "step": 4999
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.025159430774818e-06,
      "loss": 0.7507,
      "step": 5000
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.006890707106138e-06,
      "loss": 0.7642,
      "step": 5001
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.988654398752013e-06,
      "loss": 0.765,
      "step": 5002
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.970450511935388e-06,
      "loss": 0.7175,
      "step": 5003
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.952279052868125e-06,
      "loss": 0.7986,
      "step": 5004
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.934140027751022e-06,
      "loss": 0.7912,
      "step": 5005
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.9160334427738264e-06,
      "loss": 0.7938,
      "step": 5006
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.897959304115208e-06,
      "loss": 0.72,
      "step": 5007
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.87991761794272e-06,
      "loss": 0.8357,
      "step": 5008
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.861908390412939e-06,
      "loss": 0.7631,
      "step": 5009
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.843931627671261e-06,
      "loss": 0.7807,
      "step": 5010
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.825987335852078e-06,
      "loss": 0.7268,
      "step": 5011
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.808075521078681e-06,
      "loss": 0.8998,
      "step": 5012
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.790196189463236e-06,
      "loss": 0.8115,
      "step": 5013
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.772349347106897e-06,
      "loss": 0.8073,
      "step": 5014
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.754535000099691e-06,
      "loss": 0.6939,
      "step": 5015
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.736753154520535e-06,
      "loss": 0.703,
      "step": 5016
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.7190038164373065e-06,
      "loss": 0.7162,
      "step": 5017
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.701286991906762e-06,
      "loss": 0.8064,
      "step": 5018
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.683602686974542e-06,
      "loss": 0.7504,
      "step": 5019
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.665950907675243e-06,
      "loss": 0.8081,
      "step": 5020
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.648331660032324e-06,
      "loss": 0.8009,
      "step": 5021
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.630744950058141e-06,
      "loss": 0.807,
      "step": 5022
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.613190783753973e-06,
      "loss": 0.8186,
      "step": 5023
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.595669167109984e-06,
      "loss": 0.8033,
      "step": 5024
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.57818010610519e-06,
      "loss": 0.7697,
      "step": 5025
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.560723606707573e-06,
      "loss": 0.7844,
      "step": 5026
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.543299674873957e-06,
      "loss": 0.7292,
      "step": 5027
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.525908316550031e-06,
      "loss": 0.7697,
      "step": 5028
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5085495376704365e-06,
      "loss": 0.76,
      "step": 5029
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.491223344158624e-06,
      "loss": 0.8005,
      "step": 5030
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.473929741926997e-06,
      "loss": 0.8406,
      "step": 5031
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.456668736876768e-06,
      "loss": 0.8216,
      "step": 5032
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.439440334898059e-06,
      "loss": 0.7401,
      "step": 5033
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.422244541869902e-06,
      "loss": 0.75,
      "step": 5034
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.405081363660124e-06,
      "loss": 0.7608,
      "step": 5035
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.387950806125473e-06,
      "loss": 0.7869,
      "step": 5036
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.370852875111575e-06,
      "loss": 0.8343,
      "step": 5037
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.3537875764528835e-06,
      "loss": 0.8088,
      "step": 5038
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.336754915972741e-06,
      "loss": 0.8179,
      "step": 5039
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.319754899483352e-06,
      "loss": 0.7795,
      "step": 5040
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.3027875327857906e-06,
      "loss": 0.772,
      "step": 5041
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.285852821669944e-06,
      "loss": 0.8288,
      "step": 5042
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.268950771914626e-06,
      "loss": 0.775,
      "step": 5043
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.252081389287455e-06,
      "loss": 0.7819,
      "step": 5044
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.235244679544892e-06,
      "loss": 0.3305,
      "step": 5045
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.218440648432309e-06,
      "loss": 0.7728,
      "step": 5046
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.201669301683875e-06,
      "loss": 0.788,
      "step": 5047
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.1849306450226e-06,
      "loss": 0.7877,
      "step": 5048
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.168224684160394e-06,
      "loss": 0.7677,
      "step": 5049
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.15155142479795e-06,
      "loss": 0.7819,
      "step": 5050
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.134910872624853e-06,
      "loss": 0.8093,
      "step": 5051
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.118303033319471e-06,
      "loss": 0.7708,
      "step": 5052
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.101727912549058e-06,
      "loss": 0.7581,
      "step": 5053
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.08518551596968e-06,
      "loss": 0.7488,
      "step": 5054
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.068675849226256e-06,
      "loss": 0.7658,
      "step": 5055
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.052198917952488e-06,
      "loss": 0.7519,
      "step": 5056
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.035754727770969e-06,
      "loss": 0.7533,
      "step": 5057
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.019343284293087e-06,
      "loss": 0.7543,
      "step": 5058
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.002964593119052e-06,
      "loss": 0.7389,
      "step": 5059
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.986618659837904e-06,
      "loss": 0.7997,
      "step": 5060
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.970305490027526e-06,
      "loss": 0.7104,
      "step": 5061
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.954025089254565e-06,
      "loss": 0.7932,
      "step": 5062
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.937777463074566e-06,
      "loss": 0.7894,
      "step": 5063
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.9215626170318155e-06,
      "loss": 0.7332,
      "step": 5064
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.905380556659432e-06,
      "loss": 0.7856,
      "step": 5065
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8892312874794e-06,
      "loss": 0.7882,
      "step": 5066
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.873114815002454e-06,
      "loss": 0.8001,
      "step": 5067
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.857031144728152e-06,
      "loss": 0.8087,
      "step": 5068
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8409802821448685e-06,
      "loss": 0.7618,
      "step": 5069
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.824962232729767e-06,
      "loss": 0.845,
      "step": 5070
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8089770019488325e-06,
      "loss": 0.7476,
      "step": 5071
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.7930245952568444e-06,
      "loss": 0.7424,
      "step": 5072
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.777105018097371e-06,
      "loss": 0.8175,
      "step": 5073
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.7612182759028003e-06,
      "loss": 0.8234,
      "step": 5074
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.745364374094296e-06,
      "loss": 0.8075,
      "step": 5075
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.7295433180818075e-06,
      "loss": 0.7463,
      "step": 5076
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.7137551132641168e-06,
      "loss": 0.7762,
      "step": 5077
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.6979997650287478e-06,
      "loss": 0.7462,
      "step": 5078
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.682277278752022e-06,
      "loss": 0.7637,
      "step": 5079
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.666587659799092e-06,
      "loss": 0.8324,
      "step": 5080
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.6509309135238533e-06,
      "loss": 0.7164,
      "step": 5081
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.6353070452689543e-06,
      "loss": 0.8414,
      "step": 5082
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.6197160603659073e-06,
      "loss": 0.3164,
      "step": 5083
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.6041579641349465e-06,
      "loss": 0.7131,
      "step": 5084
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.58863276188508e-06,
      "loss": 0.7663,
      "step": 5085
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.5731404589141147e-06,
      "loss": 0.7768,
      "step": 5086
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.557681060508633e-06,
      "loss": 0.3588,
      "step": 5087
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.5422545719439483e-06,
      "loss": 0.7514,
      "step": 5088
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.526860998484205e-06,
      "loss": 0.8604,
      "step": 5089
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.511500345382279e-06,
      "loss": 0.318,
      "step": 5090
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.496172617879789e-06,
      "loss": 0.7751,
      "step": 5091
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.4808778212071956e-06,
      "loss": 0.78,
      "step": 5092
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.4656159605836347e-06,
      "loss": 0.7265,
      "step": 5093
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.4503870412170513e-06,
      "loss": 0.775,
      "step": 5094
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.4351910683041555e-06,
      "loss": 0.8086,
      "step": 5095
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.4200280470303658e-06,
      "loss": 0.8325,
      "step": 5096
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.4048979825699324e-06,
      "loss": 0.7555,
      "step": 5097
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3898008800858027e-06,
      "loss": 0.7678,
      "step": 5098
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3747367447296676e-06,
      "loss": 0.8065,
      "step": 5099
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.359705581642014e-06,
      "loss": 0.792,
      "step": 5100
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3447073959520627e-06,
      "loss": 0.6964,
      "step": 5101
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3297421927777404e-06,
      "loss": 0.7315,
      "step": 5102
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3148099772257856e-06,
      "loss": 0.768,
      "step": 5103
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.299910754391633e-06,
      "loss": 0.7651,
      "step": 5104
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.28504452935946e-06,
      "loss": 0.3107,
      "step": 5105
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.27021130720222e-06,
      "loss": 0.7408,
      "step": 5106
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.255411092981575e-06,
      "loss": 0.7683,
      "step": 5107
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.2406438917479077e-06,
      "loss": 0.7685,
      "step": 5108
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.2259097085403756e-06,
      "loss": 0.7554,
      "step": 5109
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.211208548386857e-06,
      "loss": 0.709,
      "step": 5110
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.196540416303939e-06,
      "loss": 0.7926,
      "step": 5111
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.1819053172969627e-06,
      "loss": 0.7832,
      "step": 5112
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.167303256359988e-06,
      "loss": 0.7683,
      "step": 5113
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.1527342384757964e-06,
      "loss": 0.7725,
      "step": 5114
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.1381982686159106e-06,
      "loss": 0.7865,
      "step": 5115
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.123695351740541e-06,
      "loss": 0.7415,
      "step": 5116
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.1092254927986843e-06,
      "loss": 0.7761,
      "step": 5117
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.094788696727979e-06,
      "loss": 0.7321,
      "step": 5118
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.080384968454808e-06,
      "loss": 0.8023,
      "step": 5119
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.0660143128943162e-06,
      "loss": 0.3396,
      "step": 5120
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.051676734950315e-06,
      "loss": 0.7908,
      "step": 5121
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.0373722395153125e-06,
      "loss": 0.7877,
      "step": 5122
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.0231008314705934e-06,
      "loss": 0.7918,
      "step": 5123
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.0088625156860837e-06,
      "loss": 0.7925,
      "step": 5124
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.994657297020453e-06,
      "loss": 0.789,
      "step": 5125
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.98048518032108e-06,
      "loss": 0.8025,
      "step": 5126
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.966346170424028e-06,
      "loss": 0.767,
      "step": 5127
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.952240272154072e-06,
      "loss": 0.778,
      "step": 5128
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.938167490324695e-06,
      "loss": 0.7669,
      "step": 5129
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.924127829738077e-06,
      "loss": 0.7921,
      "step": 5130
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.910121295185075e-06,
      "loss": 0.8056,
      "step": 5131
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.896147891445278e-06,
      "loss": 0.7559,
      "step": 5132
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8822076232869476e-06,
      "loss": 0.8182,
      "step": 5133
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8683004954670132e-06,
      "loss": 0.8035,
      "step": 5134
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8544265127311675e-06,
      "loss": 0.7968,
      "step": 5135
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.840585679813723e-06,
      "loss": 0.7512,
      "step": 5136
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.826778001437702e-06,
      "loss": 0.7942,
      "step": 5137
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8130034823148355e-06,
      "loss": 0.738,
      "step": 5138
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7992621271455075e-06,
      "loss": 0.7505,
      "step": 5139
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7855539406188015e-06,
      "loss": 0.7391,
      "step": 5140
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7718789274124856e-06,
      "loss": 0.7851,
      "step": 5141
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7582370921929943e-06,
      "loss": 0.717,
      "step": 5142
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7446284396154485e-06,
      "loss": 0.757,
      "step": 5143
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.731052974323656e-06,
      "loss": 0.7613,
      "step": 5144
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7175107009500657e-06,
      "loss": 0.8098,
      "step": 5145
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7040016241158485e-06,
      "loss": 0.7748,
      "step": 5146
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6905257484308278e-06,
      "loss": 0.7964,
      "step": 5147
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6770830784934585e-06,
      "loss": 0.835,
      "step": 5148
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6636736188909273e-06,
      "loss": 0.8379,
      "step": 5149
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.650297374199051e-06,
      "loss": 0.8075,
      "step": 5150
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.636954348982301e-06,
      "loss": 0.7499,
      "step": 5151
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6236445477938796e-06,
      "loss": 0.7208,
      "step": 5152
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.610367975175565e-06,
      "loss": 0.7513,
      "step": 5153
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5971246356578327e-06,
      "loss": 0.8284,
      "step": 5154
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5839145337598568e-06,
      "loss": 0.8004,
      "step": 5155
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5707376739894094e-06,
      "loss": 0.7533,
      "step": 5156
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5575940608429273e-06,
      "loss": 0.726,
      "step": 5157
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5444836988055553e-06,
      "loss": 0.7463,
      "step": 5158
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5314065923510378e-06,
      "loss": 0.7752,
      "step": 5159
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5183627459417826e-06,
      "loss": 0.8327,
      "step": 5160
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5053521640288625e-06,
      "loss": 0.7821,
      "step": 5161
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4923748510519705e-06,
      "loss": 0.8304,
      "step": 5162
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4794308114395093e-06,
      "loss": 0.8838,
      "step": 5163
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.466520049608445e-06,
      "loss": 0.8226,
      "step": 5164
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4536425699644315e-06,
      "loss": 0.7637,
      "step": 5165
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4407983769017873e-06,
      "loss": 0.8609,
      "step": 5166
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.427987474803417e-06,
      "loss": 0.8051,
      "step": 5167
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.415209868040913e-06,
      "loss": 0.3548,
      "step": 5168
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.402465560974487e-06,
      "loss": 0.7527,
      "step": 5169
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3897545579529834e-06,
      "loss": 0.7819,
      "step": 5170
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3770768633138762e-06,
      "loss": 0.8297,
      "step": 5171
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.364432481383305e-06,
      "loss": 0.7785,
      "step": 5172
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.351821416476019e-06,
      "loss": 0.817,
      "step": 5173
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.339243672895375e-06,
      "loss": 0.7792,
      "step": 5174
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3266992549334178e-06,
      "loss": 0.7171,
      "step": 5175
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3141881668707675e-06,
      "loss": 0.7584,
      "step": 5176
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3017104129766764e-06,
      "loss": 0.3257,
      "step": 5177
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2892659975090715e-06,
      "loss": 0.7737,
      "step": 5178
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2768549247144445e-06,
      "loss": 0.7996,
      "step": 5179
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.26447719882793e-06,
      "loss": 0.8454,
      "step": 5180
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2521328240733053e-06,
      "loss": 0.7977,
      "step": 5181
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2398218046629227e-06,
      "loss": 0.7498,
      "step": 5182
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2275441447977995e-06,
      "loss": 0.7438,
      "step": 5183
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2152998486675404e-06,
      "loss": 0.7441,
      "step": 5184
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.203088920450347e-06,
      "loss": 0.7932,
      "step": 5185
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1909113643130975e-06,
      "loss": 0.8141,
      "step": 5186
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1787671844112347e-06,
      "loss": 0.771,
      "step": 5187
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1666563848887876e-06,
      "loss": 0.8155,
      "step": 5188
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1545789698784735e-06,
      "loss": 0.8095,
      "step": 5189
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1425349435015507e-06,
      "loss": 0.8445,
      "step": 5190
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1305243098678985e-06,
      "loss": 0.8189,
      "step": 5191
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1185470730760272e-06,
      "loss": 0.7688,
      "step": 5192
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.106603237213023e-06,
      "loss": 0.7978,
      "step": 5193
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.094692806354581e-06,
      "loss": 0.7924,
      "step": 5194
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0828157845650064e-06,
      "loss": 0.7757,
      "step": 5195
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.07097217589719e-06,
      "loss": 0.7942,
      "step": 5196
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0591619843926212e-06,
      "loss": 0.7911,
      "step": 5197
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.047385214081421e-06,
      "loss": 0.7513,
      "step": 5198
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0356418689822633e-06,
      "loss": 0.7903,
      "step": 5199
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0239319531024092e-06,
      "loss": 0.8298,
      "step": 5200
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0122554704377625e-06,
      "loss": 0.7722,
      "step": 5201
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.000612424972781e-06,
      "loss": 0.8137,
      "step": 5202
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.989002820680508e-06,
      "loss": 0.3823,
      "step": 5203
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.977426661522619e-06,
      "loss": 0.7869,
      "step": 5204
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.96588395144931e-06,
      "loss": 0.7506,
      "step": 5205
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.95437469439943e-06,
      "loss": 0.7854,
      "step": 5206
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.94289889430036e-06,
      "loss": 0.7571,
      "step": 5207
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9314565550680897e-06,
      "loss": 0.7639,
      "step": 5208
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9200476806072066e-06,
      "loss": 0.7538,
      "step": 5209
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.90867227481083e-06,
      "loss": 0.8165,
      "step": 5210
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.897330341560699e-06,
      "loss": 0.8114,
      "step": 5211
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8860218847271405e-06,
      "loss": 0.7142,
      "step": 5212
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8747469081690006e-06,
      "loss": 0.7806,
      "step": 5213
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8635054157337462e-06,
      "loss": 0.8181,
      "step": 5214
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.852297411257431e-06,
      "loss": 0.742,
      "step": 5215
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8411228985646289e-06,
      "loss": 0.7668,
      "step": 5216
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8299818814685342e-06,
      "loss": 0.8106,
      "step": 5217
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8188743637708838e-06,
      "loss": 0.8014,
      "step": 5218
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.807800349262001e-06,
      "loss": 0.8107,
      "step": 5219
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.796759841720741e-06,
      "loss": 0.7247,
      "step": 5220
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.785752844914579e-06,
      "loss": 0.808,
      "step": 5221
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7747793625995213e-06,
      "loss": 0.7812,
      "step": 5222
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7638393985201286e-06,
      "loss": 0.7763,
      "step": 5223
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7529329564095587e-06,
      "loss": 0.719,
      "step": 5224
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7420600399895016e-06,
      "loss": 0.7024,
      "step": 5225
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7312206529702112e-06,
      "loss": 0.7874,
      "step": 5226
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7204147990505182e-06,
      "loss": 0.7616,
      "step": 5227
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7096424819177726e-06,
      "loss": 0.7569,
      "step": 5228
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.698903705247934e-06,
      "loss": 0.8337,
      "step": 5229
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6881984727054712e-06,
      "loss": 0.8027,
      "step": 5230
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6775267879434063e-06,
      "loss": 0.8372,
      "step": 5231
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6668886546033712e-06,
      "loss": 0.7972,
      "step": 5232
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.656284076315473e-06,
      "loss": 0.8012,
      "step": 5233
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6457130566984169e-06,
      "loss": 0.3339,
      "step": 5234
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6351755993594509e-06,
      "loss": 0.8079,
      "step": 5235
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.624671707894343e-06,
      "loss": 0.7843,
      "step": 5236
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.614201385887426e-06,
      "loss": 0.8389,
      "step": 5237
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6037646369115977e-06,
      "loss": 0.8791,
      "step": 5238
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.5933614645282756e-06,
      "loss": 0.7932,
      "step": 5239
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.5829918722873982e-06,
      "loss": 0.8071,
      "step": 5240
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5726558637275012e-06,
      "loss": 0.7944,
      "step": 5241
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5623534423756193e-06,
      "loss": 0.757,
      "step": 5242
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5520846117473286e-06,
      "loss": 0.7939,
      "step": 5243
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.54184937534676e-06,
      "loss": 0.7641,
      "step": 5244
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5316477366665637e-06,
      "loss": 0.8621,
      "step": 5245
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5214796991879332e-06,
      "loss": 0.8042,
      "step": 5246
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5113452663806038e-06,
      "loss": 0.7913,
      "step": 5247
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5012444417028204e-06,
      "loss": 0.7982,
      "step": 5248
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4911772286013925e-06,
      "loss": 0.7778,
      "step": 5249
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4811436305116388e-06,
      "loss": 0.8045,
      "step": 5250
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4711436508573872e-06,
      "loss": 0.7926,
      "step": 5251
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4611772930510416e-06,
      "loss": 0.7632,
      "step": 5252
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4512445604935032e-06,
      "loss": 0.7522,
      "step": 5253
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4413454565741945e-06,
      "loss": 0.7346,
      "step": 5254
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4314799846710803e-06,
      "loss": 0.81,
      "step": 5255
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4216481481506338e-06,
      "loss": 0.8075,
      "step": 5256
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4118499503678606e-06,
      "loss": 0.7775,
      "step": 5257
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.402085394666286e-06,
      "loss": 0.7846,
      "step": 5258
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3923544843779447e-06,
      "loss": 0.7724,
      "step": 5259
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3826572228234136e-06,
      "loss": 0.7918,
      "step": 5260
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.372993613311757e-06,
      "loss": 0.78,
      "step": 5261
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3633636591405818e-06,
      "loss": 0.7666,
      "step": 5262
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3537673635959924e-06,
      "loss": 0.7106,
      "step": 5263
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.344204729952625e-06,
      "loss": 0.7424,
      "step": 5264
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3346757614736138e-06,
      "loss": 0.7481,
      "step": 5265
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3251804614106022e-06,
      "loss": 0.763,
      "step": 5266
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.315718833003765e-06,
      "loss": 0.8016,
      "step": 5267
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3062908794817751e-06,
      "loss": 0.766,
      "step": 5268
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2968966040618035e-06,
      "loss": 0.7682,
      "step": 5269
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2875360099495525e-06,
      "loss": 0.7805,
      "step": 5270
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2782091003392006e-06,
      "loss": 0.3167,
      "step": 5271
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.268915878413468e-06,
      "loss": 0.6598,
      "step": 5272
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2596563473435518e-06,
      "loss": 0.8139,
      "step": 5273
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2504305102891578e-06,
      "loss": 0.7871,
      "step": 5274
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.241238370398501e-06,
      "loss": 0.7521,
      "step": 5275
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2320799308083054e-06,
      "loss": 0.7319,
      "step": 5276
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2229551946437602e-06,
      "loss": 0.8873,
      "step": 5277
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2138641650186077e-06,
      "loss": 0.7177,
      "step": 5278
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2048068450350335e-06,
      "loss": 0.7745,
      "step": 5279
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.195783237783743e-06,
      "loss": 0.799,
      "step": 5280
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1867933463439617e-06,
      "loss": 0.7236,
      "step": 5281
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1778371737833693e-06,
      "loss": 0.9048,
      "step": 5282
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1689147231581543e-06,
      "loss": 0.7793,
      "step": 5283
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1600259975130257e-06,
      "loss": 0.7777,
      "step": 5284
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.151170999881135e-06,
      "loss": 0.7276,
      "step": 5285
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1423497332841537e-06,
      "loss": 0.7993,
      "step": 5286
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.133562200732252e-06,
      "loss": 0.7773,
      "step": 5287
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1248084052240649e-06,
      "loss": 0.3462,
      "step": 5288
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1160883497467133e-06,
      "loss": 0.3585,
      "step": 5289
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1074020372758397e-06,
      "loss": 0.7737,
      "step": 5290
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0987494707755396e-06,
      "loss": 0.7801,
      "step": 5291
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0901306531983957e-06,
      "loss": 0.7834,
      "step": 5292
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0815455874854996e-06,
      "loss": 0.3398,
      "step": 5293
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0729942765663746e-06,
      "loss": 0.7842,
      "step": 5294
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0644767233590868e-06,
      "loss": 0.3124,
      "step": 5295
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0559929307701555e-06,
      "loss": 0.822,
      "step": 5296
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0475429016945537e-06,
      "loss": 0.3303,
      "step": 5297
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.039126639015775e-06,
      "loss": 0.7691,
      "step": 5298
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0307441456057665e-06,
      "loss": 0.7756,
      "step": 5299
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0223954243249624e-06,
      "loss": 0.7822,
      "step": 5300
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0140804780222724e-06,
      "loss": 0.7625,
      "step": 5301
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0057993095350715e-06,
      "loss": 0.3549,
      "step": 5302
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.975519216892103e-07,
      "loss": 0.7852,
      "step": 5303
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.893383172990267e-07,
      "loss": 0.7924,
      "step": 5304
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.811584991673006e-07,
      "loss": 0.7323,
      "step": 5305
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.730124700853216e-07,
      "loss": 0.8029,
      "step": 5306
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.649002328328217e-07,
      "loss": 0.7839,
      "step": 5307
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.568217901780086e-07,
      "loss": 0.7852,
      "step": 5308
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.487771448775551e-07,
      "loss": 0.776,
      "step": 5309
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.407662996766098e-07,
      "loss": 0.7657,
      "step": 5310
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.32789257308786e-07,
      "loss": 0.7874,
      "step": 5311
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.248460204961396e-07,
      "loss": 0.7891,
      "step": 5312
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.169365919492357e-07,
      "loss": 0.7318,
      "step": 5313
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.090609743670597e-07,
      "loss": 0.7624,
      "step": 5314
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.012191704370843e-07,
      "loss": 0.8324,
      "step": 5315
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.934111828352354e-07,
      "loss": 0.7847,
      "step": 5316
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.856370142258818e-07,
      "loss": 0.7643,
      "step": 5317
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.778966672619016e-07,
      "loss": 0.3304,
      "step": 5318
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.701901445845706e-07,
      "loss": 0.8556,
      "step": 5319
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.625174488236632e-07,
      "loss": 0.3403,
      "step": 5320
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.54878582597407e-07,
      "loss": 0.8533,
      "step": 5321
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.472735485124617e-07,
      "loss": 0.7351,
      "step": 5322
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.397023491639733e-07,
      "loss": 0.7842,
      "step": 5323
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.321649871355198e-07,
      "loss": 0.788,
      "step": 5324
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.246614649991435e-07,
      "loss": 0.7987,
      "step": 5325
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.171917853153299e-07,
      "loss": 0.7852,
      "step": 5326
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.097559506330287e-07,
      "loss": 0.747,
      "step": 5327
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.023539634896216e-07,
      "loss": 0.8214,
      "step": 5328
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.949858264109655e-07,
      "loss": 0.761,
      "step": 5329
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.876515419113606e-07,
      "loss": 0.8233,
      "step": 5330
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.80351112493527e-07,
      "loss": 0.7735,
      "step": 5331
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.730845406486498e-07,
      "loss": 0.7733,
      "step": 5332
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.658518288563898e-07,
      "loss": 0.7724,
      "step": 5333
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.586529795848174e-07,
      "loss": 0.8525,
      "step": 5334
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.514879952904453e-07,
      "loss": 0.7849,
      "step": 5335
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.443568784182509e-07,
      "loss": 0.8081,
      "step": 5336
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.372596314016433e-07,
      "loss": 0.7984,
      "step": 5337
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.301962566624854e-07,
      "loss": 0.7474,
      "step": 5338
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.231667566110712e-07,
      "loss": 0.3303,
      "step": 5339
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.161711336461263e-07,
      "loss": 0.8338,
      "step": 5340
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.092093901548413e-07,
      "loss": 0.7397,
      "step": 5341
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.02281528512827e-07,
      "loss": 0.7603,
      "step": 5342
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.953875510841367e-07,
      "loss": 0.7929,
      "step": 5343
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.885274602212555e-07,
      "loss": 0.8759,
      "step": 5344
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.817012582651216e-07,
      "loss": 0.7766,
      "step": 5345
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.749089475450831e-07,
      "loss": 0.7787,
      "step": 5346
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.681505303789637e-07,
      "loss": 0.8114,
      "step": 5347
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.61426009072974e-07,
      "loss": 0.7417,
      "step": 5348
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.547353859217786e-07,
      "loss": 0.7916,
      "step": 5349
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.480786632084957e-07,
      "loss": 0.7873,
      "step": 5350
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.414558432046414e-07,
      "loss": 0.7706,
      "step": 5351
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.348669281701747e-07,
      "loss": 0.773,
      "step": 5352
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.283119203535082e-07,
      "loss": 0.7827,
      "step": 5353
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.217908219914414e-07,
      "loss": 0.7802,
      "step": 5354
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.153036353092278e-07,
      "loss": 0.7978,
      "step": 5355
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.088503625205633e-07,
      "loss": 0.7824,
      "step": 5356
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.024310058275306e-07,
      "loss": 0.7965,
      "step": 5357
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.960455674206778e-07,
      "loss": 0.7364,
      "step": 5358
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.896940494789616e-07,
      "loss": 0.7347,
      "step": 5359
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.833764541697484e-07,
      "loss": 0.3201,
      "step": 5360
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.770927836488693e-07,
      "loss": 0.7594,
      "step": 5361
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.708430400605425e-07,
      "loss": 0.7826,
      "step": 5362
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.646272255374175e-07,
      "loss": 0.8235,
      "step": 5363
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.584453422005754e-07,
      "loss": 0.7653,
      "step": 5364
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.522973921595175e-07,
      "loss": 0.3281,
      "step": 5365
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.46183377512155e-07,
      "loss": 0.7387,
      "step": 5366
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.401033003448408e-07,
      "loss": 0.8171,
      "step": 5367
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.340571627323044e-07,
      "loss": 0.7353,
      "step": 5368
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.280449667377397e-07,
      "loss": 0.7741,
      "step": 5369
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.220667144127389e-07,
      "loss": 0.3537,
      "step": 5370
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.161224077973147e-07,
      "loss": 0.764,
      "step": 5371
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.102120489198781e-07,
      "loss": 0.8334,
      "step": 5372
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.043356397972931e-07,
      "loss": 0.7616,
      "step": 5373
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.984931824348004e-07,
      "loss": 0.7271,
      "step": 5374
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.92684678826083e-07,
      "loss": 0.7643,
      "step": 5375
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.869101309532109e-07,
      "loss": 0.7529,
      "step": 5376
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.811695407867079e-07,
      "loss": 0.8241,
      "step": 5377
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.754629102854624e-07,
      "loss": 0.7573,
      "step": 5378
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.697902413968058e-07,
      "loss": 0.7691,
      "step": 5379
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.6415153605647857e-07,
      "loss": 0.3059,
      "step": 5380
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.585467961886081e-07,
      "loss": 0.7492,
      "step": 5381
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.529760237057645e-07,
      "loss": 0.7522,
      "step": 5382
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.4743922050889397e-07,
      "loss": 0.8225,
      "step": 5383
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.4193638848738506e-07,
      "loss": 0.7515,
      "step": 5384
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3646752951900237e-07,
      "loss": 0.8236,
      "step": 5385
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3103264546993094e-07,
      "loss": 0.7574,
      "step": 5386
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.256317381947761e-07,
      "loss": 0.7607,
      "step": 5387
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.2026480953653026e-07,
      "loss": 0.7558,
      "step": 5388
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.1493186132658403e-07,
      "loss": 0.8143,
      "step": 5389
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.0963289538477056e-07,
      "loss": 0.7868,
      "step": 5390
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.0436791351927684e-07,
      "loss": 0.7633,
      "step": 5391
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.991369175267323e-07,
      "loss": 0.3391,
      "step": 5392
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.939399091921425e-07,
      "loss": 0.7651,
      "step": 5393
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.887768902889444e-07,
      "loss": 0.7599,
      "step": 5394
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.836478625789508e-07,
      "loss": 0.7815,
      "step": 5395
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.785528278123729e-07,
      "loss": 0.7653,
      "step": 5396
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.734917877278532e-07,
      "loss": 0.7916,
      "step": 5397
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.6846474405239916e-07,
      "loss": 0.8051,
      "step": 5398
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.6347169850143856e-07,
      "loss": 0.7886,
      "step": 5399
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.585126527787863e-07,
      "loss": 0.8046,
      "step": 5400
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.535876085766554e-07,
      "loss": 0.7923,
      "step": 5401
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.4869656757567925e-07,
      "loss": 0.7734,
      "step": 5402
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.438395314448561e-07,
      "loss": 0.7521,
      "step": 5403
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.390165018416047e-07,
      "loss": 0.7961,
      "step": 5404
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.342274804117196e-07,
      "loss": 0.7695,
      "step": 5405
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.294724687893935e-07,
      "loss": 0.7844,
      "step": 5406
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.247514685972286e-07,
      "loss": 0.8195,
      "step": 5407
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.200644814462139e-07,
      "loss": 0.791,
      "step": 5408
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.154115089357368e-07,
      "loss": 0.3416,
      "step": 5409
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.107925526535493e-07,
      "loss": 0.8224,
      "step": 5410
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.062076141758352e-07,
      "loss": 0.7929,
      "step": 5411
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.01656695067154e-07,
      "loss": 0.7883,
      "step": 5412
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.971397968804412e-07,
      "loss": 0.8379,
      "step": 5413
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.926569211570529e-07,
      "loss": 0.8358,
      "step": 5414
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.8820806942669867e-07,
      "loss": 0.7745,
      "step": 5415
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.8379324320751967e-07,
      "loss": 0.79,
      "step": 5416
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.7941244400601087e-07,
      "loss": 0.8238,
      "step": 5417
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.750656733170764e-07,
      "loss": 0.8231,
      "step": 5418
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.707529326239966e-07,
      "loss": 0.8098,
      "step": 5419
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.6647422339844964e-07,
      "loss": 0.8059,
      "step": 5420
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.6222954710050097e-07,
      "loss": 0.782,
      "step": 5421
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5801890517859195e-07,
      "loss": 0.7698,
      "step": 5422
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.53842299069551e-07,
      "loss": 0.7835,
      "step": 5423
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.496997301986048e-07,
      "loss": 0.8306,
      "step": 5424
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.45591199979367e-07,
      "loss": 0.8109,
      "step": 5425
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.4151670981380494e-07,
      "loss": 0.7337,
      "step": 5426
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.3747626109229538e-07,
      "loss": 0.7766,
      "step": 5427
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.334698551936132e-07,
      "loss": 0.7699,
      "step": 5428
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2949749348488702e-07,
      "loss": 0.7306,
      "step": 5429
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2555917732164366e-07,
      "loss": 0.7889,
      "step": 5430
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2165490804778589e-07,
      "loss": 0.8257,
      "step": 5431
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.177846869955924e-07,
      "loss": 0.7796,
      "step": 5432
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.1394851548574014e-07,
      "loss": 0.772,
      "step": 5433
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.1014639482729304e-07,
      "loss": 0.7848,
      "step": 5434
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.0637832631765775e-07,
      "loss": 0.7549,
      "step": 5435
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.0264431124265015e-07,
      "loss": 0.7828,
      "step": 5436
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9894435087647323e-07,
      "loss": 0.7566,
      "step": 5437
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.952784464816837e-07,
      "loss": 0.7861,
      "step": 5438
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9164659930923645e-07,
      "loss": 0.8237,
      "step": 5439
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8804881059845124e-07,
      "loss": 0.8082,
      "step": 5440
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8448508157703493e-07,
      "loss": 0.7886,
      "step": 5441
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8095541346107025e-07,
      "loss": 0.7672,
      "step": 5442
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.774598074550271e-07,
      "loss": 0.7493,
      "step": 5443
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.73998264751718e-07,
      "loss": 0.7855,
      "step": 5444
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.705707865323758e-07,
      "loss": 0.8461,
      "step": 5445
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.6717737396657608e-07,
      "loss": 0.7741,
      "step": 5446
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.6381802821228143e-07,
      "loss": 0.7164,
      "step": 5447
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.6049275041584156e-07,
      "loss": 0.8193,
      "step": 5448
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.572015417119599e-07,
      "loss": 0.7479,
      "step": 5449
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5394440322372694e-07,
      "loss": 0.8012,
      "step": 5450
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.507213360626092e-07,
      "loss": 0.7989,
      "step": 5451
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.4753234132842685e-07,
      "loss": 0.7355,
      "step": 5452
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.4437742010940946e-07,
      "loss": 0.7133,
      "step": 5453
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.4125657348211807e-07,
      "loss": 0.3601,
      "step": 5454
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.3816980251151192e-07,
      "loss": 0.7486,
      "step": 5455
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.3511710825091506e-07,
      "loss": 0.8224,
      "step": 5456
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.320984917420387e-07,
      "loss": 0.7558,
      "step": 5457
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2911395401493665e-07,
      "loss": 0.7347,
      "step": 5458
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2616349608804978e-07,
      "loss": 0.7492,
      "step": 5459
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2324711896819497e-07,
      "loss": 0.8007,
      "step": 5460
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.203648236505428e-07,
      "loss": 0.8286,
      "step": 5461
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1751661111865097e-07,
      "loss": 0.7119,
      "step": 5462
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1470248234443093e-07,
      "loss": 0.7121,
      "step": 5463
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1192243828818116e-07,
      "loss": 0.7724,
      "step": 5464
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0917647989856506e-07,
      "loss": 0.7729,
      "step": 5465
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0646460811259973e-07,
      "loss": 0.7698,
      "step": 5466
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0378682385567829e-07,
      "loss": 0.3583,
      "step": 5467
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0114312804156978e-07,
      "loss": 0.8409,
      "step": 5468
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.853352157240813e-08,
      "loss": 0.7831,
      "step": 5469
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.595800533868105e-08,
      "loss": 0.347,
      "step": 5470
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.341658021926325e-08,
      "loss": 0.8209,
      "step": 5471
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.090924708138326e-08,
      "loss": 0.7752,
      "step": 5472
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.843600678064557e-08,
      "loss": 0.7694,
      "step": 5473
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.599686016100838e-08,
      "loss": 0.7476,
      "step": 5474
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.359180805479483e-08,
      "loss": 0.7849,
      "step": 5475
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.122085128272616e-08,
      "loss": 0.7838,
      "step": 5476
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.88839906538441e-08,
      "loss": 0.7969,
      "step": 5477
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.658122696557746e-08,
      "loss": 0.7121,
      "step": 5478
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.431256100371986e-08,
      "loss": 0.8242,
      "step": 5479
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.207799354244094e-08,
      "loss": 0.8355,
      "step": 5480
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.987752534424186e-08,
      "loss": 0.8483,
      "step": 5481
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.771115716002197e-08,
      "loss": 0.8014,
      "step": 5482
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.557888972901216e-08,
      "loss": 0.7802,
      "step": 5483
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.348072377884151e-08,
      "loss": 0.7569,
      "step": 5484
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.141666002548174e-08,
      "loss": 0.806,
      "step": 5485
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.9386699173269444e-08,
      "loss": 0.355,
      "step": 5486
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.739084191490607e-08,
      "loss": 0.8409,
      "step": 5487
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.542908893144683e-08,
      "loss": 0.7949,
      "step": 5488
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.35014408923229e-08,
      "loss": 0.7642,
      "step": 5489
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.160789845533032e-08,
      "loss": 0.7984,
      "step": 5490
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.974846226660779e-08,
      "loss": 0.7624,
      "step": 5491
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.792313296066997e-08,
      "loss": 0.799,
      "step": 5492
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.613191116039639e-08,
      "loss": 0.777,
      "step": 5493
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.4374797477009235e-08,
      "loss": 0.7722,
      "step": 5494
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.265179251011775e-08,
      "loss": 0.7796,
      "step": 5495
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.0962896847662744e-08,
      "loss": 0.7178,
      "step": 5496
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.93081110659721e-08,
      "loss": 0.7574,
      "step": 5497
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.7687435729705234e-08,
      "loss": 0.8532,
      "step": 5498
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.6100871391930855e-08,
      "loss": 0.7959,
      "step": 5499
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.454841859401592e-08,
      "loss": 0.7905,
      "step": 5500
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.303007786572554e-08,
      "loss": 0.7754,
      "step": 5501
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.1545849725189704e-08,
      "loss": 0.8212,
      "step": 5502
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.009573467886995e-08,
      "loss": 0.7486,
      "step": 5503
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.8679733221592673e-08,
      "loss": 0.817,
      "step": 5504
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.7297845836582457e-08,
      "loss": 0.8159,
      "step": 5505
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5950072995362118e-08,
      "loss": 0.7929,
      "step": 5506
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.4636415157863747e-08,
      "loss": 0.7835,
      "step": 5507
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.335687277233989e-08,
      "loss": 0.7857,
      "step": 5508
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2111446275441263e-08,
      "loss": 0.7577,
      "step": 5509
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.0900136092139032e-08,
      "loss": 0.815,
      "step": 5510
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9722942635791442e-08,
      "loss": 0.7529,
      "step": 5511
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8579866308088278e-08,
      "loss": 0.8,
      "step": 5512
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.7470907499106403e-08,
      "loss": 0.7504,
      "step": 5513
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.6396066587254233e-08,
      "loss": 0.8071,
      "step": 5514
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5355343939305045e-08,
      "loss": 0.3227,
      "step": 5515
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.4348739910408083e-08,
      "loss": 0.7821,
      "step": 5516
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.3376254844044145e-08,
      "loss": 0.3671,
      "step": 5517
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.2437889072069996e-08,
      "loss": 0.7809,
      "step": 5518
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1533642914696163e-08,
      "loss": 0.8092,
      "step": 5519
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0663516680464724e-08,
      "loss": 0.781,
      "step": 5520
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.827510666315932e-09,
      "loss": 0.808,
      "step": 5521
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.025625157532692e-09,
      "loss": 0.8144,
      "step": 5522
    },
    {
      "epoch": 1.0,
      "learning_rate": 8.257860427729469e-09,
      "loss": 0.887,
      "step": 5523
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.524216738907796e-09,
      "loss": 0.7431,
      "step": 5524
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.8246943414118635e-09,
      "loss": 0.844,
      "step": 5525
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.159293473950723e-09,
      "loss": 0.7643,
      "step": 5526
    },
    {
      "epoch": 1.0,
      "learning_rate": 5.528014363576084e-09,
      "loss": 0.7132,
      "step": 5527
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.930857225715624e-09,
      "loss": 0.7482,
      "step": 5528
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.3678222641285735e-09,
      "loss": 0.737,
      "step": 5529
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.838909670961233e-09,
      "loss": 0.7712,
      "step": 5530
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.3441196266914554e-09,
      "loss": 0.7601,
      "step": 5531
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.8834523001619594e-09,
      "loss": 0.7229,
      "step": 5532
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.456907848558121e-09,
      "loss": 0.7107,
      "step": 5533
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.0644864174523825e-09,
      "loss": 0.7853,
      "step": 5534
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.7061881407376413e-09,
      "loss": 0.8039,
      "step": 5535
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.3820131406938608e-09,
      "loss": 0.7513,
      "step": 5536
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0919615279325613e-09,
      "loss": 0.8124,
      "step": 5537
    },
    {
      "epoch": 1.0,
      "learning_rate": 8.360334014301252e-10,
      "loss": 0.7794,
      "step": 5538
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.14228848527798e-10,
      "loss": 0.7713,
      "step": 5539
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.265479448983811e-10,
      "loss": 0.7534,
      "step": 5540
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.7299075460174294e-10,
      "loss": 0.8005,
      "step": 5541
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5355733002930805e-10,
      "loss": 0.3318,
      "step": 5542
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.824771193736368e-11,
      "loss": 0.7502,
      "step": 5543
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.7061929435957525e-11,
      "loss": 0.8009,
      "step": 5544
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0,
      "loss": 0.3548,
      "step": 5545
    },
    {
      "epoch": 1.0,
      "step": 5545,
      "total_flos": 1805886998839296.0,
      "train_loss": 0.8035830079326767,
      "train_runtime": 81873.7924,
      "train_samples_per_second": 8.126,
      "train_steps_per_second": 0.068
    }
  ],
  "max_steps": 5545,
  "num_train_epochs": 1,
  "total_flos": 1805886998839296.0,
  "trial_name": null,
  "trial_params": null
}