{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 4813,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 2.770083102493075e-07,
      "loss": 2.4516,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.54016620498615e-07,
      "loss": 1.1419,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.310249307479226e-07,
      "loss": 1.4336,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.10803324099723e-06,
      "loss": 1.7241,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.3850415512465373e-06,
      "loss": 2.3263,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.662049861495845e-06,
      "loss": 2.0258,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.9390581717451524e-06,
      "loss": 2.3028,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.21606648199446e-06,
      "loss": 2.3708,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.4930747922437675e-06,
      "loss": 1.888,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.7700831024930747e-06,
      "loss": 1.8995,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.0470914127423827e-06,
      "loss": 2.9872,
      "step": 11
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.32409972299169e-06,
      "loss": 1.1157,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.6011080332409978e-06,
      "loss": 0.8529,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.878116343490305e-06,
      "loss": 2.4801,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.155124653739612e-06,
      "loss": 2.3516,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.43213296398892e-06,
      "loss": 1.4639,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.709141274238227e-06,
      "loss": 1.4887,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.986149584487535e-06,
      "loss": 2.1823,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.263157894736842e-06,
      "loss": 2.6648,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.540166204986149e-06,
      "loss": 2.0426,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.817174515235457e-06,
      "loss": 1.4243,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.094182825484765e-06,
      "loss": 2.533,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.3711911357340724e-06,
      "loss": 1.7595,
      "step": 23
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.64819944598338e-06,
      "loss": 1.2354,
      "step": 24
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.9252077562326875e-06,
      "loss": 2.1521,
      "step": 25
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.2022160664819955e-06,
      "loss": 1.5948,
      "step": 26
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.479224376731302e-06,
      "loss": 1.7792,
      "step": 27
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.75623268698061e-06,
      "loss": 1.9494,
      "step": 28
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.033240997229918e-06,
      "loss": 1.6268,
      "step": 29
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.310249307479224e-06,
      "loss": 2.5053,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.587257617728532e-06,
      "loss": 1.5981,
      "step": 31
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.86426592797784e-06,
      "loss": 2.0617,
      "step": 32
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.141274238227148e-06,
      "loss": 2.4635,
      "step": 33
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.418282548476454e-06,
      "loss": 2.0845,
      "step": 34
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.695290858725762e-06,
      "loss": 1.4099,
      "step": 35
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.97229916897507e-06,
      "loss": 2.0843,
      "step": 36
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0249307479224378e-05,
      "loss": 1.2684,
      "step": 37
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0526315789473684e-05,
      "loss": 1.2296,
      "step": 38
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0803324099722992e-05,
      "loss": 1.4242,
      "step": 39
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1080332409972299e-05,
      "loss": 0.916,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1357340720221607e-05,
      "loss": 2.0451,
      "step": 41
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1634349030470915e-05,
      "loss": 1.3034,
      "step": 42
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1911357340720223e-05,
      "loss": 1.1253,
      "step": 43
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.218836565096953e-05,
      "loss": 1.2185,
      "step": 44
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2465373961218837e-05,
      "loss": 1.3531,
      "step": 45
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2742382271468145e-05,
      "loss": 1.3148,
      "step": 46
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3019390581717453e-05,
      "loss": 1.6893,
      "step": 47
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.329639889196676e-05,
      "loss": 1.4025,
      "step": 48
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3573407202216069e-05,
      "loss": 1.7824,
      "step": 49
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3850415512465375e-05,
      "loss": 1.8789,
      "step": 50
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4127423822714683e-05,
      "loss": 0.9541,
      "step": 51
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4404432132963991e-05,
      "loss": 1.1318,
      "step": 52
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4681440443213299e-05,
      "loss": 1.6665,
      "step": 53
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4958448753462604e-05,
      "loss": 1.8663,
      "step": 54
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5235457063711912e-05,
      "loss": 2.8427,
      "step": 55
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.551246537396122e-05,
      "loss": 0.9955,
      "step": 56
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5789473684210526e-05,
      "loss": 1.5572,
      "step": 57
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6066481994459835e-05,
      "loss": 0.7257,
      "step": 58
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6343490304709142e-05,
      "loss": 1.522,
      "step": 59
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6620498614958448e-05,
      "loss": 0.9988,
      "step": 60
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.6897506925207758e-05,
      "loss": 1.2488,
      "step": 61
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7174515235457064e-05,
      "loss": 0.9509,
      "step": 62
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.745152354570637e-05,
      "loss": 1.106,
      "step": 63
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.772853185595568e-05,
      "loss": 1.2606,
      "step": 64
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8005540166204986e-05,
      "loss": 1.012,
      "step": 65
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8282548476454296e-05,
      "loss": 1.2144,
      "step": 66
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8559556786703602e-05,
      "loss": 1.1925,
      "step": 67
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.883656509695291e-05,
      "loss": 1.3526,
      "step": 68
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9113573407202218e-05,
      "loss": 0.7862,
      "step": 69
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9390581717451524e-05,
      "loss": 0.8912,
      "step": 70
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9667590027700834e-05,
      "loss": 1.1648,
      "step": 71
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.994459833795014e-05,
      "loss": 0.7416,
      "step": 72
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.0221606648199447e-05,
      "loss": 0.8488,
      "step": 73
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.0498614958448756e-05,
      "loss": 0.8392,
      "step": 74
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.077562326869806e-05,
      "loss": 1.2264,
      "step": 75
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.105263157894737e-05,
      "loss": 1.3924,
      "step": 76
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.1329639889196675e-05,
      "loss": 1.3261,
      "step": 77
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.1606648199445985e-05,
      "loss": 1.3085,
      "step": 78
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.188365650969529e-05,
      "loss": 0.7019,
      "step": 79
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.2160664819944597e-05,
      "loss": 1.1632,
      "step": 80
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.2437673130193907e-05,
      "loss": 1.8919,
      "step": 81
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.2714681440443213e-05,
      "loss": 0.8705,
      "step": 82
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.2991689750692523e-05,
      "loss": 0.9128,
      "step": 83
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.326869806094183e-05,
      "loss": 1.252,
      "step": 84
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.3545706371191136e-05,
      "loss": 0.5336,
      "step": 85
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.3822714681440445e-05,
      "loss": 0.6955,
      "step": 86
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.409972299168975e-05,
      "loss": 1.0534,
      "step": 87
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.437673130193906e-05,
      "loss": 1.8861,
      "step": 88
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.4653739612188367e-05,
      "loss": 1.0529,
      "step": 89
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.4930747922437674e-05,
      "loss": 0.8976,
      "step": 90
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.520775623268698e-05,
      "loss": 0.7558,
      "step": 91
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.548476454293629e-05,
      "loss": 0.8996,
      "step": 92
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.5761772853185596e-05,
      "loss": 0.8331,
      "step": 93
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.6038781163434906e-05,
      "loss": 0.7756,
      "step": 94
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.6315789473684212e-05,
      "loss": 1.5636,
      "step": 95
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.659279778393352e-05,
      "loss": 0.8584,
      "step": 96
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.6869806094182825e-05,
      "loss": 0.5704,
      "step": 97
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.7146814404432138e-05,
      "loss": 1.206,
      "step": 98
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.742382271468144e-05,
      "loss": 0.7743,
      "step": 99
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.770083102493075e-05,
      "loss": 1.0619,
      "step": 100
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.7977839335180056e-05,
      "loss": 0.9664,
      "step": 101
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.8254847645429366e-05,
      "loss": 1.0329,
      "step": 102
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.8531855955678672e-05,
      "loss": 1.6155,
      "step": 103
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.8808864265927982e-05,
      "loss": 0.8396,
      "step": 104
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.9085872576177285e-05,
      "loss": 0.4981,
      "step": 105
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.9362880886426598e-05,
      "loss": 0.8217,
      "step": 106
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.96398891966759e-05,
      "loss": 0.5921,
      "step": 107
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.9916897506925207e-05,
      "loss": 1.0478,
      "step": 108
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.0193905817174517e-05,
      "loss": 0.7171,
      "step": 109
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.0470914127423823e-05,
      "loss": 0.9125,
      "step": 110
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.074792243767313e-05,
      "loss": 1.1366,
      "step": 111
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.102493074792244e-05,
      "loss": 0.4777,
      "step": 112
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.130193905817175e-05,
      "loss": 1.1932,
      "step": 113
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.157894736842105e-05,
      "loss": 1.2255,
      "step": 114
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.185595567867036e-05,
      "loss": 1.0805,
      "step": 115
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.213296398891967e-05,
      "loss": 0.7728,
      "step": 116
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.240997229916898e-05,
      "loss": 1.409,
      "step": 117
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.2686980609418284e-05,
      "loss": 1.7325,
      "step": 118
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.296398891966759e-05,
      "loss": 1.5444,
      "step": 119
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.3240997229916896e-05,
      "loss": 0.504,
      "step": 120
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.3518005540166206e-05,
      "loss": 0.4601,
      "step": 121
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.3795013850415515e-05,
      "loss": 1.4533,
      "step": 122
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.4072022160664825e-05,
      "loss": 0.8083,
      "step": 123
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.434903047091413e-05,
      "loss": 0.6857,
      "step": 124
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.462603878116344e-05,
      "loss": 0.3884,
      "step": 125
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.490304709141274e-05,
      "loss": 0.9498,
      "step": 126
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.518005540166206e-05,
      "loss": 1.1805,
      "step": 127
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.545706371191136e-05,
      "loss": 0.9497,
      "step": 128
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.573407202216066e-05,
      "loss": 0.8614,
      "step": 129
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.601108033240997e-05,
      "loss": 1.1572,
      "step": 130
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.628808864265928e-05,
      "loss": 1.2202,
      "step": 131
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.656509695290859e-05,
      "loss": 1.2482,
      "step": 132
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.6842105263157895e-05,
      "loss": 1.2191,
      "step": 133
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.7119113573407204e-05,
      "loss": 0.9905,
      "step": 134
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.739612188365651e-05,
      "loss": 0.9565,
      "step": 135
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.767313019390582e-05,
      "loss": 0.9053,
      "step": 136
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.795013850415513e-05,
      "loss": 1.0311,
      "step": 137
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.8227146814404436e-05,
      "loss": 1.7762,
      "step": 138
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.850415512465374e-05,
      "loss": 0.861,
      "step": 139
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.878116343490305e-05,
      "loss": 1.049,
      "step": 140
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.905817174515236e-05,
      "loss": 1.0488,
      "step": 141
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.933518005540167e-05,
      "loss": 0.8391,
      "step": 142
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.961218836565097e-05,
      "loss": 1.0889,
      "step": 143
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.988919667590028e-05,
      "loss": 1.2799,
      "step": 144
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.0166204986149584e-05,
      "loss": 1.0853,
      "step": 145
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.044321329639889e-05,
      "loss": 0.6603,
      "step": 146
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.07202216066482e-05,
      "loss": 0.7617,
      "step": 147
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.099722991689751e-05,
      "loss": 0.8471,
      "step": 148
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.1274238227146816e-05,
      "loss": 0.8037,
      "step": 149
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.155124653739612e-05,
      "loss": 1.2725,
      "step": 150
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.1828254847645435e-05,
      "loss": 1.1069,
      "step": 151
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.210526315789474e-05,
      "loss": 0.7645,
      "step": 152
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.238227146814405e-05,
      "loss": 0.9414,
      "step": 153
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.265927977839335e-05,
      "loss": 0.9401,
      "step": 154
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.293628808864266e-05,
      "loss": 0.6269,
      "step": 155
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.321329639889197e-05,
      "loss": 0.7399,
      "step": 156
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.349030470914128e-05,
      "loss": 1.0247,
      "step": 157
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.376731301939058e-05,
      "loss": 1.0793,
      "step": 158
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.404432132963989e-05,
      "loss": 0.7816,
      "step": 159
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.4321329639889195e-05,
      "loss": 0.7478,
      "step": 160
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.459833795013851e-05,
      "loss": 0.7049,
      "step": 161
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.4875346260387814e-05,
      "loss": 0.7153,
      "step": 162
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.5152354570637124e-05,
      "loss": 0.9062,
      "step": 163
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.542936288088643e-05,
      "loss": 0.9352,
      "step": 164
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.5706371191135736e-05,
      "loss": 1.2948,
      "step": 165
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.5983379501385046e-05,
      "loss": 0.9519,
      "step": 166
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.6260387811634356e-05,
      "loss": 0.9326,
      "step": 167
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.653739612188366e-05,
      "loss": 0.6759,
      "step": 168
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.681440443213297e-05,
      "loss": 0.8888,
      "step": 169
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.709141274238227e-05,
      "loss": 1.1796,
      "step": 170
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.736842105263158e-05,
      "loss": 0.7405,
      "step": 171
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.764542936288089e-05,
      "loss": 1.2216,
      "step": 172
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.7922437673130193e-05,
      "loss": 1.0395,
      "step": 173
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.81994459833795e-05,
      "loss": 1.0069,
      "step": 174
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8476454293628806e-05,
      "loss": 1.2213,
      "step": 175
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.875346260387812e-05,
      "loss": 0.4854,
      "step": 176
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9030470914127425e-05,
      "loss": 1.0867,
      "step": 177
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9307479224376735e-05,
      "loss": 0.7708,
      "step": 178
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.958448753462604e-05,
      "loss": 1.1336,
      "step": 179
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.986149584487535e-05,
      "loss": 0.8622,
      "step": 180
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.013850415512466e-05,
      "loss": 0.9094,
      "step": 181
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.041551246537396e-05,
      "loss": 0.5368,
      "step": 182
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.0692520775623277e-05,
      "loss": 1.1621,
      "step": 183
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.096952908587258e-05,
      "loss": 0.8451,
      "step": 184
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.124653739612188e-05,
      "loss": 1.2182,
      "step": 185
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.152354570637119e-05,
      "loss": 0.8245,
      "step": 186
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.180055401662051e-05,
      "loss": 1.2356,
      "step": 187
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.207756232686981e-05,
      "loss": 0.8001,
      "step": 188
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.2354570637119114e-05,
      "loss": 0.6124,
      "step": 189
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.2631578947368424e-05,
      "loss": 1.258,
      "step": 190
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.2908587257617734e-05,
      "loss": 0.9576,
      "step": 191
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.318559556786704e-05,
      "loss": 0.867,
      "step": 192
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.3462603878116346e-05,
      "loss": 1.0454,
      "step": 193
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.373961218836565e-05,
      "loss": 1.2299,
      "step": 194
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.401662049861496e-05,
      "loss": 0.5111,
      "step": 195
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.4293628808864275e-05,
      "loss": 1.2188,
      "step": 196
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.457063711911358e-05,
      "loss": 0.6327,
      "step": 197
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.484764542936288e-05,
      "loss": 0.6576,
      "step": 198
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.5124653739612184e-05,
      "loss": 0.7256,
      "step": 199
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.54016620498615e-05,
      "loss": 0.8528,
      "step": 200
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.567867036011081e-05,
      "loss": 1.2068,
      "step": 201
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.595567867036011e-05,
      "loss": 1.4115,
      "step": 202
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.6232686980609416e-05,
      "loss": 1.0652,
      "step": 203
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.650969529085873e-05,
      "loss": 0.567,
      "step": 204
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.6786703601108035e-05,
      "loss": 1.5937,
      "step": 205
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.7063711911357345e-05,
      "loss": 1.0456,
      "step": 206
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.734072022160665e-05,
      "loss": 1.5754,
      "step": 207
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.7617728531855964e-05,
      "loss": 0.478,
      "step": 208
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.789473684210527e-05,
      "loss": 0.6663,
      "step": 209
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.817174515235457e-05,
      "loss": 0.7149,
      "step": 210
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.844875346260388e-05,
      "loss": 0.9025,
      "step": 211
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.8725761772853196e-05,
      "loss": 0.8972,
      "step": 212
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.90027700831025e-05,
      "loss": 0.5211,
      "step": 213
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.92797783933518e-05,
      "loss": 0.9915,
      "step": 214
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.955678670360111e-05,
      "loss": 1.1982,
      "step": 215
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.9833795013850414e-05,
      "loss": 0.6263,
      "step": 216
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.011080332409973e-05,
      "loss": 0.915,
      "step": 217
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.0387811634349034e-05,
      "loss": 0.7625,
      "step": 218
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.0664819944598337e-05,
      "loss": 0.8035,
      "step": 219
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.0941828254847646e-05,
      "loss": 0.8536,
      "step": 220
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.121883656509696e-05,
      "loss": 1.327,
      "step": 221
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.149584487534626e-05,
      "loss": 0.8058,
      "step": 222
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.177285318559557e-05,
      "loss": 0.7089,
      "step": 223
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.204986149584488e-05,
      "loss": 1.2896,
      "step": 224
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.232686980609419e-05,
      "loss": 1.0336,
      "step": 225
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.26038781163435e-05,
      "loss": 1.1276,
      "step": 226
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.28808864265928e-05,
      "loss": 0.8729,
      "step": 227
    },
    {
      "epoch": 0.09,
      "learning_rate": 6.31578947368421e-05,
      "loss": 1.1788,
      "step": 228
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.343490304709143e-05,
      "loss": 0.8849,
      "step": 229
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.371191135734072e-05,
      "loss": 0.8003,
      "step": 230
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.398891966759003e-05,
      "loss": 1.6763,
      "step": 231
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.426592797783934e-05,
      "loss": 1.5251,
      "step": 232
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.454293628808865e-05,
      "loss": 1.1443,
      "step": 233
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.481994459833796e-05,
      "loss": 1.1207,
      "step": 234
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.509695290858726e-05,
      "loss": 0.2793,
      "step": 235
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.537396121883657e-05,
      "loss": 0.98,
      "step": 236
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.565096952908588e-05,
      "loss": 0.9188,
      "step": 237
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.592797783933519e-05,
      "loss": 1.1698,
      "step": 238
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.62049861495845e-05,
      "loss": 0.7019,
      "step": 239
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.648199445983379e-05,
      "loss": 1.2051,
      "step": 240
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.67590027700831e-05,
      "loss": 1.6581,
      "step": 241
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.703601108033241e-05,
      "loss": 1.1257,
      "step": 242
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.731301939058172e-05,
      "loss": 0.9579,
      "step": 243
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.759002770083103e-05,
      "loss": 0.9176,
      "step": 244
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.786703601108033e-05,
      "loss": 0.852,
      "step": 245
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.814404432132965e-05,
      "loss": 1.4101,
      "step": 246
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.842105263157895e-05,
      "loss": 0.9897,
      "step": 247
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.869806094182826e-05,
      "loss": 0.9868,
      "step": 248
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.897506925207757e-05,
      "loss": 0.983,
      "step": 249
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.925207756232688e-05,
      "loss": 1.234,
      "step": 250
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.952908587257619e-05,
      "loss": 0.6475,
      "step": 251
    },
    {
      "epoch": 0.1,
      "learning_rate": 6.980609418282548e-05,
      "loss": 0.9303,
      "step": 252
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.008310249307479e-05,
      "loss": 0.5906,
      "step": 253
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.036011080332411e-05,
      "loss": 0.8298,
      "step": 254
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.063711911357341e-05,
      "loss": 0.7511,
      "step": 255
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.091412742382272e-05,
      "loss": 0.8627,
      "step": 256
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.119113573407203e-05,
      "loss": 0.924,
      "step": 257
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.146814404432133e-05,
      "loss": 0.9678,
      "step": 258
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.174515235457065e-05,
      "loss": 0.9091,
      "step": 259
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.202216066481994e-05,
      "loss": 1.2758,
      "step": 260
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.229916897506925e-05,
      "loss": 0.8964,
      "step": 261
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.257617728531856e-05,
      "loss": 0.7729,
      "step": 262
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.285318559556787e-05,
      "loss": 0.9729,
      "step": 263
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.313019390581718e-05,
      "loss": 0.8856,
      "step": 264
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.340720221606648e-05,
      "loss": 1.1951,
      "step": 265
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.368421052631579e-05,
      "loss": 0.8721,
      "step": 266
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.39612188365651e-05,
      "loss": 1.0066,
      "step": 267
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.423822714681441e-05,
      "loss": 0.9103,
      "step": 268
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.451523545706372e-05,
      "loss": 0.4787,
      "step": 269
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.479224376731301e-05,
      "loss": 1.2774,
      "step": 270
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.506925207756234e-05,
      "loss": 0.9079,
      "step": 271
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.534626038781163e-05,
      "loss": 0.7152,
      "step": 272
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.562326869806094e-05,
      "loss": 0.6791,
      "step": 273
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.590027700831025e-05,
      "loss": 0.6372,
      "step": 274
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.617728531855956e-05,
      "loss": 0.6888,
      "step": 275
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.645429362880887e-05,
      "loss": 0.8371,
      "step": 276
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.673130193905818e-05,
      "loss": 0.9267,
      "step": 277
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.700831024930748e-05,
      "loss": 0.5088,
      "step": 278
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.728531855955679e-05,
      "loss": 1.0857,
      "step": 279
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.75623268698061e-05,
      "loss": 1.2099,
      "step": 280
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.783933518005541e-05,
      "loss": 1.6544,
      "step": 281
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.811634349030472e-05,
      "loss": 0.5819,
      "step": 282
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.839335180055401e-05,
      "loss": 1.2996,
      "step": 283
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.867036011080334e-05,
      "loss": 0.5985,
      "step": 284
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.894736842105263e-05,
      "loss": 1.2408,
      "step": 285
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.922437673130194e-05,
      "loss": 0.8336,
      "step": 286
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.950138504155125e-05,
      "loss": 0.9302,
      "step": 287
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.977839335180056e-05,
      "loss": 0.455,
      "step": 288
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.005540166204987e-05,
      "loss": 0.8132,
      "step": 289
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.033240997229917e-05,
      "loss": 0.7687,
      "step": 290
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.060941828254848e-05,
      "loss": 0.7678,
      "step": 291
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.088642659279779e-05,
      "loss": 0.6097,
      "step": 292
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.11634349030471e-05,
      "loss": 0.696,
      "step": 293
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.14404432132964e-05,
      "loss": 0.8854,
      "step": 294
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.17174515235457e-05,
      "loss": 0.4643,
      "step": 295
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.199445983379503e-05,
      "loss": 1.197,
      "step": 296
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.227146814404432e-05,
      "loss": 0.7087,
      "step": 297
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.254847645429363e-05,
      "loss": 1.7173,
      "step": 298
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.282548476454294e-05,
      "loss": 0.7955,
      "step": 299
    },
    {
      "epoch": 0.12,
      "learning_rate": 8.310249307479224e-05,
      "loss": 0.5615,
      "step": 300
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.337950138504156e-05,
      "loss": 0.9396,
      "step": 301
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.365650969529087e-05,
      "loss": 0.9317,
      "step": 302
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.393351800554017e-05,
      "loss": 0.5371,
      "step": 303
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.421052631578948e-05,
      "loss": 1.0856,
      "step": 304
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.448753462603879e-05,
      "loss": 0.8757,
      "step": 305
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.47645429362881e-05,
      "loss": 1.1083,
      "step": 306
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.50415512465374e-05,
      "loss": 0.9584,
      "step": 307
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.53185595567867e-05,
      "loss": 0.9875,
      "step": 308
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.559556786703602e-05,
      "loss": 0.7562,
      "step": 309
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.587257617728532e-05,
      "loss": 1.5099,
      "step": 310
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.614958448753463e-05,
      "loss": 0.9142,
      "step": 311
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.642659279778394e-05,
      "loss": 0.9608,
      "step": 312
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.670360110803325e-05,
      "loss": 1.1148,
      "step": 313
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.698060941828256e-05,
      "loss": 1.0648,
      "step": 314
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.725761772853185e-05,
      "loss": 0.9291,
      "step": 315
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.753462603878116e-05,
      "loss": 0.5483,
      "step": 316
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.781163434903047e-05,
      "loss": 0.8278,
      "step": 317
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.808864265927978e-05,
      "loss": 1.0554,
      "step": 318
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.83656509695291e-05,
      "loss": 1.1652,
      "step": 319
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.864265927977839e-05,
      "loss": 0.8169,
      "step": 320
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.89196675900277e-05,
      "loss": 1.0545,
      "step": 321
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.919667590027702e-05,
      "loss": 0.8911,
      "step": 322
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.947368421052632e-05,
      "loss": 1.181,
      "step": 323
    },
    {
      "epoch": 0.13,
      "learning_rate": 8.975069252077563e-05,
      "loss": 0.8033,
      "step": 324
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.002770083102492e-05,
      "loss": 1.0105,
      "step": 325
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.030470914127425e-05,
      "loss": 1.1323,
      "step": 326
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.058171745152356e-05,
      "loss": 1.291,
      "step": 327
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.085872576177285e-05,
      "loss": 0.3639,
      "step": 328
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.113573407202216e-05,
      "loss": 0.9244,
      "step": 329
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.141274238227147e-05,
      "loss": 0.9796,
      "step": 330
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.168975069252078e-05,
      "loss": 0.6952,
      "step": 331
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.196675900277009e-05,
      "loss": 0.8255,
      "step": 332
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.224376731301939e-05,
      "loss": 1.7742,
      "step": 333
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.252077562326871e-05,
      "loss": 0.4923,
      "step": 334
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.279778393351801e-05,
      "loss": 0.9578,
      "step": 335
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.307479224376732e-05,
      "loss": 0.7029,
      "step": 336
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.335180055401663e-05,
      "loss": 1.2133,
      "step": 337
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.362880886426594e-05,
      "loss": 1.5807,
      "step": 338
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.390581717451525e-05,
      "loss": 0.9825,
      "step": 339
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.418282548476454e-05,
      "loss": 1.4704,
      "step": 340
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.445983379501385e-05,
      "loss": 0.8761,
      "step": 341
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.473684210526316e-05,
      "loss": 1.5869,
      "step": 342
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.501385041551247e-05,
      "loss": 1.0389,
      "step": 343
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.529085872576178e-05,
      "loss": 0.4874,
      "step": 344
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.556786703601108e-05,
      "loss": 0.6492,
      "step": 345
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.584487534626039e-05,
      "loss": 0.8624,
      "step": 346
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.612188365650971e-05,
      "loss": 1.104,
      "step": 347
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.6398891966759e-05,
      "loss": 1.3245,
      "step": 348
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.667590027700832e-05,
      "loss": 1.2713,
      "step": 349
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.695290858725761e-05,
      "loss": 0.8749,
      "step": 350
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.722991689750694e-05,
      "loss": 1.2783,
      "step": 351
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.750692520775624e-05,
      "loss": 0.6497,
      "step": 352
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.778393351800554e-05,
      "loss": 1.0809,
      "step": 353
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.806094182825485e-05,
      "loss": 1.755,
      "step": 354
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.833795013850416e-05,
      "loss": 2.0544,
      "step": 355
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.861495844875347e-05,
      "loss": 1.5558,
      "step": 356
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.889196675900278e-05,
      "loss": 1.8522,
      "step": 357
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.916897506925208e-05,
      "loss": 1.0079,
      "step": 358
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.94459833795014e-05,
      "loss": 1.4571,
      "step": 359
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.97229916897507e-05,
      "loss": 0.7455,
      "step": 360
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001,
      "loss": 1.287,
      "step": 361
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999999818794044e-05,
      "loss": 0.7739,
      "step": 362
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999999275176184e-05,
      "loss": 0.9937,
      "step": 363
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999998369146462e-05,
      "loss": 1.0043,
      "step": 364
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999997100704943e-05,
      "loss": 0.8218,
      "step": 365
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999995469851721e-05,
      "loss": 0.7426,
      "step": 366
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999993476586911e-05,
      "loss": 0.88,
      "step": 367
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999991120910658e-05,
      "loss": 1.4221,
      "step": 368
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999988402823136e-05,
      "loss": 1.3923,
      "step": 369
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999985322324538e-05,
      "loss": 0.6583,
      "step": 370
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.99998187941509e-05,
      "loss": 0.5681,
      "step": 371
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.99997807409504e-05,
      "loss": 0.9847,
      "step": 372
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.999973906364666e-05,
      "loss": 1.3089,
      "step": 373
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999969376224266e-05,
      "loss": 0.8442,
      "step": 374
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999964483674172e-05,
      "loss": 1.2414,
      "step": 375
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999959228714738e-05,
      "loss": 0.6574,
      "step": 376
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999953611346343e-05,
      "loss": 0.6109,
      "step": 377
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999947631569395e-05,
      "loss": 0.7767,
      "step": 378
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999941289384328e-05,
      "loss": 1.0146,
      "step": 379
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999934584791601e-05,
      "loss": 0.8703,
      "step": 380
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999927517791702e-05,
      "loss": 1.3027,
      "step": 381
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999920088385141e-05,
      "loss": 1.0246,
      "step": 382
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999912296572459e-05,
      "loss": 0.8761,
      "step": 383
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999904142354217e-05,
      "loss": 1.1485,
      "step": 384
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.99989562573101e-05,
      "loss": 0.5776,
      "step": 385
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999886746703452e-05,
      "loss": 1.1358,
      "step": 386
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999877505272189e-05,
      "loss": 1.5704,
      "step": 387
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999867901437888e-05,
      "loss": 0.4576,
      "step": 388
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999857935201248e-05,
      "loss": 1.3932,
      "step": 389
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999847606562991e-05,
      "loss": 0.8259,
      "step": 390
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999836915523863e-05,
      "loss": 0.7099,
      "step": 391
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999825862084642e-05,
      "loss": 0.6918,
      "step": 392
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999814446246127e-05,
      "loss": 0.8928,
      "step": 393
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999802668009149e-05,
      "loss": 0.7045,
      "step": 394
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999790527374557e-05,
      "loss": 0.8182,
      "step": 395
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999778024343233e-05,
      "loss": 0.5616,
      "step": 396
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.999765158916085e-05,
      "loss": 1.4836,
      "step": 397
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999751931094044e-05,
      "loss": 1.137,
      "step": 398
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999738340878066e-05,
      "loss": 1.0374,
      "step": 399
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999724388269141e-05,
      "loss": 1.3395,
      "step": 400
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999710073268279e-05,
      "loss": 0.9327,
      "step": 401
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999695395876514e-05,
      "loss": 1.426,
      "step": 402
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999680356094914e-05,
      "loss": 0.8123,
      "step": 403
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999664953924568e-05,
      "loss": 1.1999,
      "step": 404
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.99964918936659e-05,
      "loss": 1.1627,
      "step": 405
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999633062422127e-05,
      "loss": 0.6366,
      "step": 406
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999616573092345e-05,
      "loss": 0.8652,
      "step": 407
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.99959972137844e-05,
      "loss": 0.7808,
      "step": 408
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999582507281632e-05,
      "loss": 1.7164,
      "step": 409
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999564930803172e-05,
      "loss": 0.9112,
      "step": 410
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.99954699194433e-05,
      "loss": 0.5176,
      "step": 411
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999528690706409e-05,
      "loss": 1.4466,
      "step": 412
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999510027090735e-05,
      "loss": 0.9172,
      "step": 413
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.99949100109866e-05,
      "loss": 0.8974,
      "step": 414
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999471612731563e-05,
      "loss": 0.8095,
      "step": 415
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999451861990851e-05,
      "loss": 0.9953,
      "step": 416
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999431748877953e-05,
      "loss": 0.9861,
      "step": 417
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.99941127339433e-05,
      "loss": 1.5732,
      "step": 418
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999390435541463e-05,
      "loss": 0.8435,
      "step": 419
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.999369235320865e-05,
      "loss": 1.0071,
      "step": 420
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.99934767273407e-05,
      "loss": 1.2903,
      "step": 421
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999325747782644e-05,
      "loss": 0.7888,
      "step": 422
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999303460468173e-05,
      "loss": 0.9473,
      "step": 423
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999280810792274e-05,
      "loss": 1.0012,
      "step": 424
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999257798756589e-05,
      "loss": 1.1884,
      "step": 425
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999234424362787e-05,
      "loss": 0.8214,
      "step": 426
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999210687612559e-05,
      "loss": 0.7141,
      "step": 427
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999186588507627e-05,
      "loss": 1.2262,
      "step": 428
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.99916212704974e-05,
      "loss": 1.7432,
      "step": 429
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999137303240668e-05,
      "loss": 1.2902,
      "step": 430
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999112117082211e-05,
      "loss": 1.0154,
      "step": 431
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999086568576196e-05,
      "loss": 0.7786,
      "step": 432
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999060657724473e-05,
      "loss": 1.1318,
      "step": 433
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.99903438452892e-05,
      "loss": 1.424,
      "step": 434
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.999007748991445e-05,
      "loss": 1.5228,
      "step": 435
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998980751113974e-05,
      "loss": 0.9609,
      "step": 436
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998953390898464e-05,
      "loss": 0.616,
      "step": 437
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998925668346903e-05,
      "loss": 1.0843,
      "step": 438
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998897583461295e-05,
      "loss": 0.9082,
      "step": 439
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998869136243679e-05,
      "loss": 1.4522,
      "step": 440
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998840326696114e-05,
      "loss": 0.8633,
      "step": 441
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998811154820693e-05,
      "loss": 0.8465,
      "step": 442
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998781620619526e-05,
      "loss": 1.1348,
      "step": 443
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998751724094755e-05,
      "loss": 1.2992,
      "step": 444
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.998721465248546e-05,
      "loss": 0.8851,
      "step": 445
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998690844083094e-05,
      "loss": 1.0128,
      "step": 446
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998659860600618e-05,
      "loss": 1.5887,
      "step": 447
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998628514803366e-05,
      "loss": 0.7011,
      "step": 448
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998596806693605e-05,
      "loss": 1.2681,
      "step": 449
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998564736273635e-05,
      "loss": 0.8021,
      "step": 450
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998532303545783e-05,
      "loss": 1.4809,
      "step": 451
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998499508512398e-05,
      "loss": 1.3732,
      "step": 452
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998466351175857e-05,
      "loss": 0.8764,
      "step": 453
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998432831538564e-05,
      "loss": 1.45,
      "step": 454
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998398949602947e-05,
      "loss": 0.8088,
      "step": 455
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998364705371464e-05,
      "loss": 0.9615,
      "step": 456
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998330098846597e-05,
      "loss": 1.1891,
      "step": 457
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998295130030853e-05,
      "loss": 1.2467,
      "step": 458
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998259798926766e-05,
      "loss": 1.1141,
      "step": 459
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998224105536899e-05,
      "loss": 1.0638,
      "step": 460
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998188049863839e-05,
      "loss": 1.2785,
      "step": 461
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998151631910197e-05,
      "loss": 1.1698,
      "step": 462
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998114851678614e-05,
      "loss": 1.1133,
      "step": 463
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998077709171757e-05,
      "loss": 0.9759,
      "step": 464
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998040204392317e-05,
      "loss": 1.4767,
      "step": 465
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.998002337343013e-05,
      "loss": 0.7275,
      "step": 466
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.99796410802659e-05,
      "loss": 0.4613,
      "step": 467
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.997925516445817e-05,
      "loss": 1.3165,
      "step": 468
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.997886562603493e-05,
      "loss": 0.8251,
      "step": 469
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.99784724650244e-05,
      "loss": 1.3078,
      "step": 470
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997807568145512e-05,
      "loss": 0.8873,
      "step": 471
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.99776752753558e-05,
      "loss": 1.5214,
      "step": 472
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997727124675547e-05,
      "loss": 0.9913,
      "step": 473
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997686359568343e-05,
      "loss": 0.9718,
      "step": 474
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997645232216923e-05,
      "loss": 1.6416,
      "step": 475
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997603742624268e-05,
      "loss": 1.5948,
      "step": 476
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997561890793383e-05,
      "loss": 0.9131,
      "step": 477
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997519676727303e-05,
      "loss": 1.3404,
      "step": 478
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997477100429089e-05,
      "loss": 1.6455,
      "step": 479
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997434161901825e-05,
      "loss": 1.1472,
      "step": 480
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997390861148625e-05,
      "loss": 1.1566,
      "step": 481
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997347198172626e-05,
      "loss": 0.9719,
      "step": 482
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997303172976993e-05,
      "loss": 0.6265,
      "step": 483
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997258785564918e-05,
      "loss": 0.6254,
      "step": 484
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997214035939617e-05,
      "loss": 1.287,
      "step": 485
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997168924104337e-05,
      "loss": 0.9873,
      "step": 486
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997123450062342e-05,
      "loss": 1.2503,
      "step": 487
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997077613816932e-05,
      "loss": 0.6549,
      "step": 488
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.997031415371431e-05,
      "loss": 0.9606,
      "step": 489
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.996984854729183e-05,
      "loss": 1.2762,
      "step": 490
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.996937931893567e-05,
      "loss": 1.26,
      "step": 491
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.99689064686798e-05,
      "loss": 1.264,
      "step": 492
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.99684299965585e-05,
      "loss": 1.0751,
      "step": 493
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996794990260636e-05,
      "loss": 1.5657,
      "step": 494
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996746618685812e-05,
      "loss": 0.997,
      "step": 495
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996697884934885e-05,
      "loss": 0.7725,
      "step": 496
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996648789011389e-05,
      "loss": 1.3007,
      "step": 497
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996599330918882e-05,
      "loss": 1.2407,
      "step": 498
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996549510660949e-05,
      "loss": 1.6198,
      "step": 499
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.9964993282412e-05,
      "loss": 1.2881,
      "step": 500
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996448783663274e-05,
      "loss": 1.2531,
      "step": 501
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996397876930832e-05,
      "loss": 0.7795,
      "step": 502
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996346608047568e-05,
      "loss": 1.1183,
      "step": 503
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996294977017194e-05,
      "loss": 1.3853,
      "step": 504
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996242983843453e-05,
      "loss": 0.5341,
      "step": 505
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996190628530117e-05,
      "loss": 1.0163,
      "step": 506
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996137911080978e-05,
      "loss": 0.6984,
      "step": 507
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996084831499857e-05,
      "loss": 1.3504,
      "step": 508
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.996031389790601e-05,
      "loss": 1.2927,
      "step": 509
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.995977585957085e-05,
      "loss": 0.9868,
      "step": 510
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.995923420003209e-05,
      "loss": 0.9603,
      "step": 511
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.995868891932897e-05,
      "loss": 1.2706,
      "step": 512
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.995814001750103e-05,
      "loss": 0.7377,
      "step": 513
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.995758749458806e-05,
      "loss": 1.2598,
      "step": 514
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.995703135063009e-05,
      "loss": 1.0538,
      "step": 515
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.995647158566745e-05,
      "loss": 0.7755,
      "step": 516
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.99559081997407e-05,
      "loss": 0.7726,
      "step": 517
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.995534119289067e-05,
      "loss": 0.9899,
      "step": 518
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.995477056515849e-05,
      "loss": 1.2317,
      "step": 519
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.995419631658548e-05,
      "loss": 0.9583,
      "step": 520
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.995361844721329e-05,
      "loss": 1.0481,
      "step": 521
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.995303695708379e-05,
      "loss": 1.3511,
      "step": 522
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.995245184623914e-05,
      "loss": 0.9508,
      "step": 523
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.995186311472174e-05,
      "loss": 1.1754,
      "step": 524
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.995127076257427e-05,
      "loss": 1.6059,
      "step": 525
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.995067478983966e-05,
      "loss": 0.6649,
      "step": 526
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.99500751965611e-05,
      "loss": 1.3839,
      "step": 527
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994947198278207e-05,
      "loss": 0.7803,
      "step": 528
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.99488651485463e-05,
      "loss": 0.9222,
      "step": 529
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994825469389773e-05,
      "loss": 0.9513,
      "step": 530
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994764061888064e-05,
      "loss": 1.493,
      "step": 531
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994702292353953e-05,
      "loss": 0.9816,
      "step": 532
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994640160791918e-05,
      "loss": 0.5874,
      "step": 533
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994577667206461e-05,
      "loss": 0.6704,
      "step": 534
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994514811602116e-05,
      "loss": 0.7239,
      "step": 535
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994451593983433e-05,
      "loss": 1.0561,
      "step": 536
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994388014354996e-05,
      "loss": 0.9429,
      "step": 537
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994324072721417e-05,
      "loss": 1.3617,
      "step": 538
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994259769087325e-05,
      "loss": 1.5153,
      "step": 539
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994195103457386e-05,
      "loss": 0.7302,
      "step": 540
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.994130075836283e-05,
      "loss": 1.2639,
      "step": 541
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.994064686228732e-05,
      "loss": 1.0782,
      "step": 542
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993998934639472e-05,
      "loss": 0.9725,
      "step": 543
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993932821073269e-05,
      "loss": 1.4768,
      "step": 544
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993866345534912e-05,
      "loss": 1.2437,
      "step": 545
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993799508029224e-05,
      "loss": 1.0078,
      "step": 546
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993732308561047e-05,
      "loss": 0.8274,
      "step": 547
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993664747135253e-05,
      "loss": 1.406,
      "step": 548
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993596823756739e-05,
      "loss": 0.6912,
      "step": 549
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993528538430427e-05,
      "loss": 0.963,
      "step": 550
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993459891161267e-05,
      "loss": 0.7273,
      "step": 551
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993390881954235e-05,
      "loss": 0.5315,
      "step": 552
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993321510814331e-05,
      "loss": 1.5135,
      "step": 553
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993251777746586e-05,
      "loss": 0.8403,
      "step": 554
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993181682756054e-05,
      "loss": 0.7446,
      "step": 555
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993111225847814e-05,
      "loss": 1.9286,
      "step": 556
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.993040407026974e-05,
      "loss": 0.7665,
      "step": 557
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.992969226298668e-05,
      "loss": 1.2518,
      "step": 558
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.992897683668052e-05,
      "loss": 1.2373,
      "step": 559
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.992825779140316e-05,
      "loss": 0.966,
      "step": 560
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.99275351272067e-05,
      "loss": 0.8988,
      "step": 561
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.99268088441435e-05,
      "loss": 0.5575,
      "step": 562
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.992607894226622e-05,
      "loss": 1.033,
      "step": 563
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.992534542162779e-05,
      "loss": 1.1972,
      "step": 564
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.992460828228132e-05,
      "loss": 1.5913,
      "step": 565
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.992386752428029e-05,
      "loss": 0.574,
      "step": 566
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.992312314767837e-05,
      "loss": 1.081,
      "step": 567
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.992237515252952e-05,
      "loss": 1.3954,
      "step": 568
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.992162353888797e-05,
      "loss": 1.1814,
      "step": 569
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.992086830680816e-05,
      "loss": 0.8444,
      "step": 570
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.992010945634487e-05,
      "loss": 0.8178,
      "step": 571
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991934698755307e-05,
      "loss": 1.2119,
      "step": 572
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991858090048807e-05,
      "loss": 0.9955,
      "step": 573
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991781119520534e-05,
      "loss": 1.0341,
      "step": 574
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991703787176073e-05,
      "loss": 1.1999,
      "step": 575
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991626093021024e-05,
      "loss": 1.1573,
      "step": 576
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991548037061023e-05,
      "loss": 0.6923,
      "step": 577
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991469619301724e-05,
      "loss": 1.1561,
      "step": 578
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991390839748813e-05,
      "loss": 1.2644,
      "step": 579
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991311698407999e-05,
      "loss": 1.2513,
      "step": 580
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991232195285019e-05,
      "loss": 0.9585,
      "step": 581
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991152330385636e-05,
      "loss": 1.2577,
      "step": 582
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.991072103715637e-05,
      "loss": 1.017,
      "step": 583
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.990991515280841e-05,
      "loss": 1.0325,
      "step": 584
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.990910565087083e-05,
      "loss": 0.7671,
      "step": 585
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.990829253140237e-05,
      "loss": 0.9188,
      "step": 586
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.990747579446193e-05,
      "loss": 0.8552,
      "step": 587
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.990665544010871e-05,
      "loss": 0.3456,
      "step": 588
    },
    {
      "epoch": 0.24,
      "learning_rate": 9.990583146840217e-05,
      "loss": 0.888,
      "step": 589
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.990500387940205e-05,
      "loss": 1.4125,
      "step": 590
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.990417267316832e-05,
      "loss": 1.4779,
      "step": 591
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.990333784976125e-05,
      "loss": 1.1239,
      "step": 592
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.990249940924132e-05,
      "loss": 0.8218,
      "step": 593
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.99016573516693e-05,
      "loss": 1.7241,
      "step": 594
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.990081167710628e-05,
      "loss": 1.043,
      "step": 595
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989996238561348e-05,
      "loss": 1.2991,
      "step": 596
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989910947725251e-05,
      "loss": 1.0368,
      "step": 597
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989825295208518e-05,
      "loss": 0.415,
      "step": 598
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989739281017356e-05,
      "loss": 1.4931,
      "step": 599
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989652905158002e-05,
      "loss": 0.8225,
      "step": 600
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989566167636713e-05,
      "loss": 1.258,
      "step": 601
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.98947906845978e-05,
      "loss": 1.0488,
      "step": 602
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989391607633514e-05,
      "loss": 1.0103,
      "step": 603
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989303785164254e-05,
      "loss": 1.0881,
      "step": 604
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989215601058365e-05,
      "loss": 0.6532,
      "step": 605
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989127055322243e-05,
      "loss": 1.6959,
      "step": 606
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.989038147962302e-05,
      "loss": 1.1058,
      "step": 607
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.988948878984987e-05,
      "loss": 1.1321,
      "step": 608
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.988859248396768e-05,
      "loss": 1.232,
      "step": 609
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.988769256204142e-05,
      "loss": 1.4632,
      "step": 610
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.988678902413632e-05,
      "loss": 1.0904,
      "step": 611
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.98858818703179e-05,
      "loss": 1.1457,
      "step": 612
    },
    {
      "epoch": 0.25,
      "learning_rate": 9.988497110065187e-05,
      "loss": 0.8902,
      "step": 613
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.988405671520425e-05,
      "loss": 1.0681,
      "step": 614
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.988313871404133e-05,
      "loss": 1.1388,
      "step": 615
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.988221709722964e-05,
      "loss": 1.1393,
      "step": 616
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.9881291864836e-05,
      "loss": 1.4655,
      "step": 617
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.988036301692746e-05,
      "loss": 0.7697,
      "step": 618
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.987943055357134e-05,
      "loss": 0.6915,
      "step": 619
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.987849447483522e-05,
      "loss": 1.3482,
      "step": 620
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.987755478078698e-05,
      "loss": 0.7527,
      "step": 621
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.98766114714947e-05,
      "loss": 1.036,
      "step": 622
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.987566454702677e-05,
      "loss": 1.0451,
      "step": 623
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.987471400745181e-05,
      "loss": 0.63,
      "step": 624
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.987375985283873e-05,
      "loss": 1.6047,
      "step": 625
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.987280208325671e-05,
      "loss": 0.7384,
      "step": 626
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.987184069877513e-05,
      "loss": 1.1584,
      "step": 627
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.98708756994637e-05,
      "loss": 1.0661,
      "step": 628
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.986990708539234e-05,
      "loss": 1.3548,
      "step": 629
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.986893485663129e-05,
      "loss": 1.5035,
      "step": 630
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.986795901325101e-05,
      "loss": 1.2413,
      "step": 631
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.986697955532223e-05,
      "loss": 1.1699,
      "step": 632
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.986599648291592e-05,
      "loss": 0.4446,
      "step": 633
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.986500979610336e-05,
      "loss": 1.2013,
      "step": 634
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.986401949495607e-05,
      "loss": 1.2153,
      "step": 635
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.986302557954582e-05,
      "loss": 0.9389,
      "step": 636
    },
    {
      "epoch": 0.26,
      "learning_rate": 9.986202804994465e-05,
      "loss": 1.1534,
      "step": 637
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.986102690622487e-05,
      "loss": 1.3332,
      "step": 638
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.986002214845905e-05,
      "loss": 0.8681,
      "step": 639
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.985901377671999e-05,
      "loss": 1.0889,
      "step": 640
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.98580017910808e-05,
      "loss": 1.5499,
      "step": 641
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.985698619161485e-05,
      "loss": 1.3318,
      "step": 642
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.985596697839572e-05,
      "loss": 1.1389,
      "step": 643
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.98549441514973e-05,
      "loss": 0.6993,
      "step": 644
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.985391771099371e-05,
      "loss": 1.1576,
      "step": 645
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.985288765695938e-05,
      "loss": 1.0987,
      "step": 646
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.985185398946895e-05,
      "loss": 1.2714,
      "step": 647
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.985081670859734e-05,
      "loss": 1.0139,
      "step": 648
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.984977581441976e-05,
      "loss": 0.8837,
      "step": 649
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.984873130701162e-05,
      "loss": 1.4751,
      "step": 650
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.984768318644867e-05,
      "loss": 1.547,
      "step": 651
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.984663145280681e-05,
      "loss": 1.3918,
      "step": 652
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.984557610616236e-05,
      "loss": 1.22,
      "step": 653
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.984451714659176e-05,
      "loss": 1.2127,
      "step": 654
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.984345457417177e-05,
      "loss": 1.0945,
      "step": 655
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.984238838897944e-05,
      "loss": 1.5107,
      "step": 656
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.9841318591092e-05,
      "loss": 1.2592,
      "step": 657
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.984024518058703e-05,
      "loss": 1.1376,
      "step": 658
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.983916815754231e-05,
      "loss": 0.9781,
      "step": 659
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.983808752203593e-05,
      "loss": 0.9025,
      "step": 660
    },
    {
      "epoch": 0.27,
      "learning_rate": 9.98370032741462e-05,
      "loss": 0.9707,
      "step": 661
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.983591541395171e-05,
      "loss": 1.2644,
      "step": 662
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.98348239415313e-05,
      "loss": 1.3253,
      "step": 663
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.983372885696411e-05,
      "loss": 0.793,
      "step": 664
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.98326301603295e-05,
      "loss": 1.0069,
      "step": 665
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.983152785170711e-05,
      "loss": 1.946,
      "step": 666
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.983042193117682e-05,
      "loss": 1.0148,
      "step": 667
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.98293123988188e-05,
      "loss": 1.0045,
      "step": 668
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.98281992547135e-05,
      "loss": 0.5171,
      "step": 669
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.982708249894156e-05,
      "loss": 1.0548,
      "step": 670
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.982596213158394e-05,
      "loss": 1.2561,
      "step": 671
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.982483815272185e-05,
      "loss": 1.2261,
      "step": 672
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.982371056243678e-05,
      "loss": 1.151,
      "step": 673
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.982257936081042e-05,
      "loss": 1.1979,
      "step": 674
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.982144454792479e-05,
      "loss": 0.8637,
      "step": 675
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.982030612386213e-05,
      "loss": 1.172,
      "step": 676
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.981916408870496e-05,
      "loss": 0.8265,
      "step": 677
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.981801844253607e-05,
      "loss": 1.5317,
      "step": 678
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.981686918543848e-05,
      "loss": 0.7719,
      "step": 679
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.981571631749549e-05,
      "loss": 0.8323,
      "step": 680
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.981455983879069e-05,
      "loss": 1.065,
      "step": 681
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.981339974940787e-05,
      "loss": 0.6196,
      "step": 682
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.981223604943113e-05,
      "loss": 1.1609,
      "step": 683
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.981106873894483e-05,
      "loss": 1.1284,
      "step": 684
    },
    {
      "epoch": 0.28,
      "learning_rate": 9.980989781803357e-05,
      "loss": 1.13,
      "step": 685
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.980872328678221e-05,
      "loss": 1.2345,
      "step": 686
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.98075451452759e-05,
      "loss": 1.2659,
      "step": 687
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.980636339360003e-05,
      "loss": 1.0879,
      "step": 688
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.980517803184024e-05,
      "loss": 1.5741,
      "step": 689
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.980398906008248e-05,
      "loss": 1.0617,
      "step": 690
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.980279647841289e-05,
      "loss": 1.382,
      "step": 691
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.980160028691795e-05,
      "loss": 1.1371,
      "step": 692
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.980040048568434e-05,
      "loss": 0.8621,
      "step": 693
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.979919707479904e-05,
      "loss": 1.068,
      "step": 694
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.979799005434925e-05,
      "loss": 1.1513,
      "step": 695
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.979677942442247e-05,
      "loss": 0.7399,
      "step": 696
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.979556518510645e-05,
      "loss": 1.2456,
      "step": 697
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.979434733648923e-05,
      "loss": 1.5527,
      "step": 698
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.979312587865903e-05,
      "loss": 0.9568,
      "step": 699
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.979190081170443e-05,
      "loss": 0.6935,
      "step": 700
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.979067213571418e-05,
      "loss": 0.8643,
      "step": 701
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.978943985077737e-05,
      "loss": 0.8728,
      "step": 702
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.978820395698332e-05,
      "loss": 0.9427,
      "step": 703
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.978696445442161e-05,
      "loss": 0.9966,
      "step": 704
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.978572134318207e-05,
      "loss": 1.5495,
      "step": 705
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.97844746233548e-05,
      "loss": 1.9311,
      "step": 706
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.978322429503017e-05,
      "loss": 0.8942,
      "step": 707
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.978197035829883e-05,
      "loss": 1.0314,
      "step": 708
    },
    {
      "epoch": 0.29,
      "learning_rate": 9.978071281325163e-05,
      "loss": 0.6374,
      "step": 709
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.977945165997976e-05,
      "loss": 0.5481,
      "step": 710
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.977818689857458e-05,
      "loss": 0.7299,
      "step": 711
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.977691852912783e-05,
      "loss": 0.5471,
      "step": 712
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.977564655173137e-05,
      "loss": 1.2751,
      "step": 713
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.977437096647746e-05,
      "loss": 1.4971,
      "step": 714
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.977309177345852e-05,
      "loss": 1.2786,
      "step": 715
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.977180897276728e-05,
      "loss": 1.1526,
      "step": 716
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.977052256449672e-05,
      "loss": 0.828,
      "step": 717
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.976923254874008e-05,
      "loss": 1.225,
      "step": 718
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.976793892559089e-05,
      "loss": 1.8042,
      "step": 719
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.976664169514287e-05,
      "loss": 1.3964,
      "step": 720
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.976534085749006e-05,
      "loss": 1.233,
      "step": 721
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.976403641272676e-05,
      "loss": 0.5881,
      "step": 722
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.976272836094752e-05,
      "loss": 0.8648,
      "step": 723
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.976141670224714e-05,
      "loss": 1.1886,
      "step": 724
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.97601014367207e-05,
      "loss": 0.7227,
      "step": 725
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.975878256446352e-05,
      "loss": 0.8762,
      "step": 726
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.975746008557122e-05,
      "loss": 1.2463,
      "step": 727
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.975613400013964e-05,
      "loss": 1.1755,
      "step": 728
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.97548043082649e-05,
      "loss": 1.0002,
      "step": 729
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.975347101004336e-05,
      "loss": 1.6159,
      "step": 730
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.975213410557169e-05,
      "loss": 1.0661,
      "step": 731
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.97507935949468e-05,
      "loss": 1.1082,
      "step": 732
    },
    {
      "epoch": 0.3,
      "learning_rate": 9.974944947826582e-05,
      "loss": 0.9349,
      "step": 733
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.974810175562619e-05,
      "loss": 0.8732,
      "step": 734
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.97467504271256e-05,
      "loss": 1.3898,
      "step": 735
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.974539549286197e-05,
      "loss": 1.2534,
      "step": 736
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.974403695293356e-05,
      "loss": 1.5326,
      "step": 737
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.974267480743879e-05,
      "loss": 1.0471,
      "step": 738
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.974130905647643e-05,
      "loss": 0.4377,
      "step": 739
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.973993970014547e-05,
      "loss": 0.9906,
      "step": 740
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.973856673854513e-05,
      "loss": 1.5736,
      "step": 741
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.973719017177496e-05,
      "loss": 1.2486,
      "step": 742
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.97358099999347e-05,
      "loss": 1.5276,
      "step": 743
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.973442622312444e-05,
      "loss": 1.2612,
      "step": 744
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.973303884144444e-05,
      "loss": 0.7954,
      "step": 745
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.973164785499527e-05,
      "loss": 1.0939,
      "step": 746
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.973025326387775e-05,
      "loss": 0.7563,
      "step": 747
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.972885506819297e-05,
      "loss": 1.4224,
      "step": 748
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.972745326804227e-05,
      "loss": 0.687,
      "step": 749
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.972604786352726e-05,
      "loss": 0.9512,
      "step": 750
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.97246388547498e-05,
      "loss": 1.046,
      "step": 751
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.972322624181204e-05,
      "loss": 1.4294,
      "step": 752
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.972181002481634e-05,
      "loss": 1.9004,
      "step": 753
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.972039020386537e-05,
      "loss": 1.2359,
      "step": 754
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.971896677906204e-05,
      "loss": 1.2688,
      "step": 755
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.971753975050949e-05,
      "loss": 1.7539,
      "step": 756
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.97161091183112e-05,
      "loss": 1.1803,
      "step": 757
    },
    {
      "epoch": 0.31,
      "learning_rate": 9.971467488257085e-05,
      "loss": 1.0164,
      "step": 758
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.971323704339239e-05,
      "loss": 1.4858,
      "step": 759
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.971179560088004e-05,
      "loss": 1.4968,
      "step": 760
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.971035055513828e-05,
      "loss": 1.0047,
      "step": 761
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.970890190627188e-05,
      "loss": 0.9391,
      "step": 762
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.970744965438579e-05,
      "loss": 1.0358,
      "step": 763
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.97059937995853e-05,
      "loss": 0.9471,
      "step": 764
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.970453434197595e-05,
      "loss": 1.159,
      "step": 765
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.970307128166349e-05,
      "loss": 1.1594,
      "step": 766
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.970160461875398e-05,
      "loss": 0.8064,
      "step": 767
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.970013435335373e-05,
      "loss": 0.912,
      "step": 768
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.96986604855693e-05,
      "loss": 1.1174,
      "step": 769
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.969718301550754e-05,
      "loss": 1.434,
      "step": 770
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.969570194327553e-05,
      "loss": 1.1348,
      "step": 771
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.969421726898061e-05,
      "loss": 1.526,
      "step": 772
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.96927289927304e-05,
      "loss": 1.0218,
      "step": 773
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.969123711463277e-05,
      "loss": 0.8555,
      "step": 774
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.968974163479588e-05,
      "loss": 0.9914,
      "step": 775
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.968824255332809e-05,
      "loss": 1.8183,
      "step": 776
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.96867398703381e-05,
      "loss": 0.7459,
      "step": 777
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.968523358593478e-05,
      "loss": 1.0359,
      "step": 778
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.968372370022732e-05,
      "loss": 0.9833,
      "step": 779
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.96822102133252e-05,
      "loss": 1.3266,
      "step": 780
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.968069312533807e-05,
      "loss": 1.252,
      "step": 781
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.967917243637592e-05,
      "loss": 1.126,
      "step": 782
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.967764814654897e-05,
      "loss": 0.9714,
      "step": 783
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.967612025596769e-05,
      "loss": 0.8582,
      "step": 784
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.967458876474285e-05,
      "loss": 1.1435,
      "step": 785
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.967305367298543e-05,
      "loss": 1.3345,
      "step": 786
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.967151498080671e-05,
      "loss": 1.5396,
      "step": 787
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.966997268831823e-05,
      "loss": 1.3118,
      "step": 788
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.966842679563176e-05,
      "loss": 1.7285,
      "step": 789
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.966687730285936e-05,
      "loss": 1.8017,
      "step": 790
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.966532421011335e-05,
      "loss": 1.699,
      "step": 791
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.966376751750627e-05,
      "loss": 0.8836,
      "step": 792
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.966220722515097e-05,
      "loss": 1.0445,
      "step": 793
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.966064333316057e-05,
      "loss": 0.8295,
      "step": 794
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.965907584164838e-05,
      "loss": 1.0048,
      "step": 795
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.965750475072804e-05,
      "loss": 1.2424,
      "step": 796
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.965593006051341e-05,
      "loss": 1.3293,
      "step": 797
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.965435177111865e-05,
      "loss": 0.8955,
      "step": 798
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.965276988265817e-05,
      "loss": 1.853,
      "step": 799
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.965118439524659e-05,
      "loss": 1.1707,
      "step": 800
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.964959530899884e-05,
      "loss": 1.0941,
      "step": 801
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.964800262403011e-05,
      "loss": 1.0707,
      "step": 802
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.964640634045586e-05,
      "loss": 1.7526,
      "step": 803
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.964480645839175e-05,
      "loss": 1.7632,
      "step": 804
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.964320297795378e-05,
      "loss": 0.9543,
      "step": 805
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.964159589925815e-05,
      "loss": 1.1821,
      "step": 806
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.963998522242135e-05,
      "loss": 0.8391,
      "step": 807
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.963837094756013e-05,
      "loss": 1.1324,
      "step": 808
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.96367530747915e-05,
      "loss": 1.452,
      "step": 809
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.963513160423272e-05,
      "loss": 1.1043,
      "step": 810
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.963350653600132e-05,
      "loss": 0.9177,
      "step": 811
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.96318778702151e-05,
      "loss": 1.2958,
      "step": 812
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.963024560699211e-05,
      "loss": 0.8503,
      "step": 813
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.962860974645063e-05,
      "loss": 1.4663,
      "step": 814
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.962697028870925e-05,
      "loss": 0.731,
      "step": 815
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.962532723388682e-05,
      "loss": 0.7078,
      "step": 816
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.962368058210241e-05,
      "loss": 1.1129,
      "step": 817
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.962203033347537e-05,
      "loss": 0.7671,
      "step": 818
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.962037648812533e-05,
      "loss": 0.8314,
      "step": 819
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.961871904617215e-05,
      "loss": 0.8157,
      "step": 820
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.961705800773599e-05,
      "loss": 1.8031,
      "step": 821
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.961539337293721e-05,
      "loss": 1.4299,
      "step": 822
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.96137251418965e-05,
      "loss": 1.084,
      "step": 823
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.961205331473476e-05,
      "loss": 0.9545,
      "step": 824
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.961037789157317e-05,
      "loss": 0.9594,
      "step": 825
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.960869887253317e-05,
      "loss": 0.685,
      "step": 826
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.960701625773645e-05,
      "loss": 0.5898,
      "step": 827
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.960533004730499e-05,
      "loss": 0.3748,
      "step": 828
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.9603640241361e-05,
      "loss": 0.944,
      "step": 829
    },
    {
      "epoch": 0.34,
      "learning_rate": 9.960194684002695e-05,
      "loss": 1.3102,
      "step": 830
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.960024984342559e-05,
      "loss": 0.6937,
      "step": 831
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.959854925167995e-05,
      "loss": 1.7224,
      "step": 832
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.959684506491324e-05,
      "loss": 1.0259,
      "step": 833
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.959513728324903e-05,
      "loss": 0.7729,
      "step": 834
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.959342590681107e-05,
      "loss": 0.7291,
      "step": 835
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.959171093572343e-05,
      "loss": 1.4775,
      "step": 836
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.958999237011041e-05,
      "loss": 1.3844,
      "step": 837
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.958827021009656e-05,
      "loss": 0.9956,
      "step": 838
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.958654445580672e-05,
      "loss": 0.8586,
      "step": 839
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.958481510736599e-05,
      "loss": 1.0153,
      "step": 840
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.958308216489969e-05,
      "loss": 0.741,
      "step": 841
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.958134562853342e-05,
      "loss": 1.4884,
      "step": 842
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.957960549839309e-05,
      "loss": 1.3806,
      "step": 843
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.95778617746048e-05,
      "loss": 1.2423,
      "step": 844
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.957611445729494e-05,
      "loss": 1.0033,
      "step": 845
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.957436354659018e-05,
      "loss": 0.8687,
      "step": 846
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.95726090426174e-05,
      "loss": 1.0185,
      "step": 847
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.95708509455038e-05,
      "loss": 0.7968,
      "step": 848
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.956908925537678e-05,
      "loss": 0.9734,
      "step": 849
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.956732397236406e-05,
      "loss": 0.6137,
      "step": 850
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.956555509659358e-05,
      "loss": 0.8893,
      "step": 851
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.956378262819354e-05,
      "loss": 1.2645,
      "step": 852
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.956200656729244e-05,
      "loss": 1.2021,
      "step": 853
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.956022691401899e-05,
      "loss": 0.5049,
      "step": 854
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.95584436685022e-05,
      "loss": 0.7578,
      "step": 855
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.95566568308713e-05,
      "loss": 1.1112,
      "step": 856
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.955486640125584e-05,
      "loss": 1.015,
      "step": 857
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.955307237978557e-05,
      "loss": 1.4002,
      "step": 858
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.955127476659053e-05,
      "loss": 1.3222,
      "step": 859
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.9549473561801e-05,
      "loss": 0.7072,
      "step": 860
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.954766876554758e-05,
      "loss": 0.6467,
      "step": 861
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.954586037796102e-05,
      "loss": 1.0206,
      "step": 862
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.954404839917247e-05,
      "loss": 0.6349,
      "step": 863
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.954223282931319e-05,
      "loss": 1.2141,
      "step": 864
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.954041366851485e-05,
      "loss": 0.9477,
      "step": 865
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.953859091690926e-05,
      "loss": 1.5755,
      "step": 866
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.953676457462855e-05,
      "loss": 1.1425,
      "step": 867
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.95349346418051e-05,
      "loss": 1.1923,
      "step": 868
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.953310111857155e-05,
      "loss": 1.0386,
      "step": 869
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.95312640050608e-05,
      "loss": 1.3826,
      "step": 870
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.952942330140597e-05,
      "loss": 0.8811,
      "step": 871
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.952757900774054e-05,
      "loss": 0.7122,
      "step": 872
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.952573112419816e-05,
      "loss": 0.6666,
      "step": 873
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.952387965091277e-05,
      "loss": 1.0029,
      "step": 874
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.952202458801857e-05,
      "loss": 1.2713,
      "step": 875
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.952016593565e-05,
      "loss": 1.3955,
      "step": 876
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.951830369394181e-05,
      "loss": 2.1807,
      "step": 877
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.951643786302896e-05,
      "loss": 1.1137,
      "step": 878
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.95145684430467e-05,
      "loss": 1.5083,
      "step": 879
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.951269543413053e-05,
      "loss": 0.68,
      "step": 880
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.95108188364162e-05,
      "loss": 0.9071,
      "step": 881
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.950893865003974e-05,
      "loss": 1.3305,
      "step": 882
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.950705487513744e-05,
      "loss": 1.448,
      "step": 883
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.950516751184582e-05,
      "loss": 1.28,
      "step": 884
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.950327656030168e-05,
      "loss": 1.194,
      "step": 885
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.95013820206421e-05,
      "loss": 1.0805,
      "step": 886
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.949948389300439e-05,
      "loss": 0.9947,
      "step": 887
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.949758217752612e-05,
      "loss": 1.0635,
      "step": 888
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.949567687434516e-05,
      "loss": 1.4685,
      "step": 889
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.949376798359957e-05,
      "loss": 0.8237,
      "step": 890
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.949185550542776e-05,
      "loss": 1.657,
      "step": 891
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.94899394399683e-05,
      "loss": 0.5713,
      "step": 892
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.948801978736011e-05,
      "loss": 0.7943,
      "step": 893
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.948609654774231e-05,
      "loss": 0.8406,
      "step": 894
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.948416972125431e-05,
      "loss": 1.5165,
      "step": 895
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.948223930803577e-05,
      "loss": 0.7982,
      "step": 896
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.94803053082266e-05,
      "loss": 1.0695,
      "step": 897
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.947836772196701e-05,
      "loss": 1.0359,
      "step": 898
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.947642654939741e-05,
      "loss": 1.4777,
      "step": 899
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.947448179065852e-05,
      "loss": 1.2066,
      "step": 900
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.947253344589129e-05,
      "loss": 1.5291,
      "step": 901
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.947058151523693e-05,
      "loss": 0.9287,
      "step": 902
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.946862599883695e-05,
      "loss": 1.3863,
      "step": 903
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.946666689683307e-05,
      "loss": 0.471,
      "step": 904
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.946470420936729e-05,
      "loss": 1.0581,
      "step": 905
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.946273793658188e-05,
      "loss": 1.0473,
      "step": 906
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.946076807861936e-05,
      "loss": 0.4576,
      "step": 907
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.94587946356225e-05,
      "loss": 1.7349,
      "step": 908
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.945681760773436e-05,
      "loss": 1.2862,
      "step": 909
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.94548369950982e-05,
      "loss": 0.5314,
      "step": 910
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.945285279785761e-05,
      "loss": 1.6043,
      "step": 911
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.945086501615642e-05,
      "loss": 0.9184,
      "step": 912
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.944887365013867e-05,
      "loss": 1.0944,
      "step": 913
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.944687869994874e-05,
      "loss": 0.6829,
      "step": 914
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.944488016573121e-05,
      "loss": 0.9706,
      "step": 915
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.944287804763093e-05,
      "loss": 1.1893,
      "step": 916
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.944087234579304e-05,
      "loss": 1.4894,
      "step": 917
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.94388630603629e-05,
      "loss": 0.9959,
      "step": 918
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.943685019148614e-05,
      "loss": 2.1717,
      "step": 919
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.943483373930868e-05,
      "loss": 1.2038,
      "step": 920
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.943281370397668e-05,
      "loss": 0.9158,
      "step": 921
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.943079008563654e-05,
      "loss": 0.4257,
      "step": 922
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.942876288443493e-05,
      "loss": 1.4138,
      "step": 923
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.94267321005188e-05,
      "loss": 1.2227,
      "step": 924
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.942469773403536e-05,
      "loss": 1.1004,
      "step": 925
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.942265978513203e-05,
      "loss": 0.5936,
      "step": 926
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.942061825395657e-05,
      "loss": 0.9576,
      "step": 927
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.94185731406569e-05,
      "loss": 1.1438,
      "step": 928
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.94165244453813e-05,
      "loss": 1.2187,
      "step": 929
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.941447216827827e-05,
      "loss": 0.5806,
      "step": 930
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.941241630949651e-05,
      "loss": 0.8657,
      "step": 931
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.941035686918508e-05,
      "loss": 1.1697,
      "step": 932
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.940829384749323e-05,
      "loss": 1.492,
      "step": 933
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.940622724457052e-05,
      "loss": 1.3968,
      "step": 934
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.940415706056673e-05,
      "loss": 1.2035,
      "step": 935
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.94020832956319e-05,
      "loss": 0.7929,
      "step": 936
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.940000594991634e-05,
      "loss": 1.2811,
      "step": 937
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.939792502357064e-05,
      "loss": 1.6425,
      "step": 938
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.939584051674559e-05,
      "loss": 0.5716,
      "step": 939
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.939375242959235e-05,
      "loss": 1.2241,
      "step": 940
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.93916607622622e-05,
      "loss": 0.7763,
      "step": 941
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.93895655149068e-05,
      "loss": 1.4929,
      "step": 942
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.938746668767799e-05,
      "loss": 1.4949,
      "step": 943
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.938536428072791e-05,
      "loss": 1.2009,
      "step": 944
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.938325829420893e-05,
      "loss": 0.7104,
      "step": 945
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.93811487282737e-05,
      "loss": 1.3436,
      "step": 946
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.937903558307517e-05,
      "loss": 1.3054,
      "step": 947
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.937691885876645e-05,
      "loss": 0.8848,
      "step": 948
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.937479855550098e-05,
      "loss": 0.8869,
      "step": 949
    },
    {
      "epoch": 0.39,
      "learning_rate": 9.937267467343247e-05,
      "loss": 0.7426,
      "step": 950
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.937054721271484e-05,
      "loss": 1.5019,
      "step": 951
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.936841617350229e-05,
      "loss": 1.3013,
      "step": 952
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.936628155594929e-05,
      "loss": 1.3086,
      "step": 953
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.936414336021058e-05,
      "loss": 1.3706,
      "step": 954
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.936200158644112e-05,
      "loss": 0.7076,
      "step": 955
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.935985623479614e-05,
      "loss": 1.3003,
      "step": 956
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.935770730543118e-05,
      "loss": 0.9868,
      "step": 957
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.935555479850197e-05,
      "loss": 0.8065,
      "step": 958
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.935339871416452e-05,
      "loss": 1.1384,
      "step": 959
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.935123905257513e-05,
      "loss": 0.9941,
      "step": 960
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.934907581389032e-05,
      "loss": 1.0941,
      "step": 961
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.934690899826692e-05,
      "loss": 0.8136,
      "step": 962
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.934473860586195e-05,
      "loss": 0.8793,
      "step": 963
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.934256463683274e-05,
      "loss": 1.051,
      "step": 964
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.934038709133686e-05,
      "loss": 1.3232,
      "step": 965
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.933820596953216e-05,
      "loss": 0.8045,
      "step": 966
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.93360212715767e-05,
      "loss": 1.424,
      "step": 967
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.933383299762886e-05,
      "loss": 0.8243,
      "step": 968
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.933164114784726e-05,
      "loss": 1.492,
      "step": 969
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.932944572239073e-05,
      "loss": 0.7309,
      "step": 970
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.932724672141842e-05,
      "loss": 1.5969,
      "step": 971
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.932504414508974e-05,
      "loss": 1.0973,
      "step": 972
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.93228379935643e-05,
      "loss": 1.7657,
      "step": 973
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.932062826700204e-05,
      "loss": 0.959,
      "step": 974
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.93184149655631e-05,
      "loss": 1.0445,
      "step": 975
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.931619808940794e-05,
      "loss": 1.0737,
      "step": 976
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.931397763869721e-05,
      "loss": 1.0674,
      "step": 977
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.931175361359186e-05,
      "loss": 1.0106,
      "step": 978
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.930952601425312e-05,
      "loss": 1.4708,
      "step": 979
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.930729484084242e-05,
      "loss": 1.1476,
      "step": 980
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.930506009352148e-05,
      "loss": 1.0001,
      "step": 981
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.93028217724523e-05,
      "loss": 0.897,
      "step": 982
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.930057987779711e-05,
      "loss": 0.9552,
      "step": 983
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.92983344097184e-05,
      "loss": 1.1291,
      "step": 984
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.929608536837896e-05,
      "loss": 1.1322,
      "step": 985
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.929383275394174e-05,
      "loss": 0.5714,
      "step": 986
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.929157656657009e-05,
      "loss": 0.9944,
      "step": 987
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.92893168064275e-05,
      "loss": 0.9557,
      "step": 988
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.928705347367777e-05,
      "loss": 1.1833,
      "step": 989
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.928478656848497e-05,
      "loss": 1.5862,
      "step": 990
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.928251609101337e-05,
      "loss": 1.3692,
      "step": 991
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.928024204142758e-05,
      "loss": 0.749,
      "step": 992
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.92779644198924e-05,
      "loss": 1.5582,
      "step": 993
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.927568322657293e-05,
      "loss": 1.1404,
      "step": 994
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.927339846163452e-05,
      "loss": 1.0357,
      "step": 995
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.927111012524279e-05,
      "loss": 1.3642,
      "step": 996
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.926881821756358e-05,
      "loss": 1.2304,
      "step": 997
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.9266522738763e-05,
      "loss": 0.9742,
      "step": 998
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.926422368900747e-05,
      "loss": 0.9416,
      "step": 999
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.926192106846359e-05,
      "loss": 0.8632,
      "step": 1000
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.92596148772983e-05,
      "loss": 1.5058,
      "step": 1001
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.925730511567872e-05,
      "loss": 0.7877,
      "step": 1002
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.92549917837723e-05,
      "loss": 1.1285,
      "step": 1003
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.92526748817467e-05,
      "loss": 1.1989,
      "step": 1004
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.925035440976987e-05,
      "loss": 1.4861,
      "step": 1005
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.924803036800996e-05,
      "loss": 1.0762,
      "step": 1006
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.924570275663547e-05,
      "loss": 0.9575,
      "step": 1007
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.924337157581509e-05,
      "loss": 1.0656,
      "step": 1008
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.924103682571779e-05,
      "loss": 1.1118,
      "step": 1009
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.92386985065128e-05,
      "loss": 0.4379,
      "step": 1010
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.923635661836962e-05,
      "loss": 1.2157,
      "step": 1011
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.923401116145797e-05,
      "loss": 1.0579,
      "step": 1012
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.923166213594786e-05,
      "loss": 1.3061,
      "step": 1013
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.922930954200957e-05,
      "loss": 0.8935,
      "step": 1014
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.922695337981363e-05,
      "loss": 1.0971,
      "step": 1015
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.922459364953078e-05,
      "loss": 1.1252,
      "step": 1016
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.922223035133208e-05,
      "loss": 0.6858,
      "step": 1017
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.921986348538884e-05,
      "loss": 0.8246,
      "step": 1018
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.921749305187259e-05,
      "loss": 0.7449,
      "step": 1019
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.921511905095516e-05,
      "loss": 1.296,
      "step": 1020
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.921274148280864e-05,
      "loss": 1.1888,
      "step": 1021
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.921036034760532e-05,
      "loss": 1.2126,
      "step": 1022
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.920797564551784e-05,
      "loss": 1.0199,
      "step": 1023
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.920558737671901e-05,
      "loss": 1.3225,
      "step": 1024
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.920319554138196e-05,
      "loss": 1.3046,
      "step": 1025
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.920080013968003e-05,
      "loss": 1.3716,
      "step": 1026
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.919840117178687e-05,
      "loss": 0.949,
      "step": 1027
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.919599863787635e-05,
      "loss": 0.754,
      "step": 1028
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.919359253812261e-05,
      "loss": 1.147,
      "step": 1029
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.919118287270009e-05,
      "loss": 0.5768,
      "step": 1030
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.918876964178337e-05,
      "loss": 1.1333,
      "step": 1031
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.918635284554743e-05,
      "loss": 1.1223,
      "step": 1032
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.918393248416743e-05,
      "loss": 0.9234,
      "step": 1033
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.918150855781879e-05,
      "loss": 0.9046,
      "step": 1034
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.917908106667721e-05,
      "loss": 1.2014,
      "step": 1035
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.917665001091865e-05,
      "loss": 0.7608,
      "step": 1036
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.917421539071932e-05,
      "loss": 0.8675,
      "step": 1037
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.917177720625566e-05,
      "loss": 1.3332,
      "step": 1038
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.916933545770442e-05,
      "loss": 0.9023,
      "step": 1039
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.916689014524257e-05,
      "loss": 1.2829,
      "step": 1040
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.916444126904735e-05,
      "loss": 1.27,
      "step": 1041
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.916198882929628e-05,
      "loss": 0.4931,
      "step": 1042
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.91595328261671e-05,
      "loss": 0.8926,
      "step": 1043
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.915707325983785e-05,
      "loss": 0.5463,
      "step": 1044
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.915461013048678e-05,
      "loss": 1.4035,
      "step": 1045
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.915214343829243e-05,
      "loss": 0.5114,
      "step": 1046
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.91496731834336e-05,
      "loss": 1.0977,
      "step": 1047
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.914719936608934e-05,
      "loss": 1.3436,
      "step": 1048
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.914472198643893e-05,
      "loss": 0.7776,
      "step": 1049
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.9142241044662e-05,
      "loss": 1.3885,
      "step": 1050
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.91397565409383e-05,
      "loss": 1.0247,
      "step": 1051
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.913726847544795e-05,
      "loss": 0.7926,
      "step": 1052
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.91347768483713e-05,
      "loss": 1.2663,
      "step": 1053
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.913228165988892e-05,
      "loss": 0.9279,
      "step": 1054
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.912978291018172e-05,
      "loss": 1.2046,
      "step": 1055
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.912728059943075e-05,
      "loss": 0.8019,
      "step": 1056
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.91247747278174e-05,
      "loss": 1.5348,
      "step": 1057
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.912226529552335e-05,
      "loss": 1.5953,
      "step": 1058
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.911975230273043e-05,
      "loss": 0.8471,
      "step": 1059
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.911723574962081e-05,
      "loss": 1.3134,
      "step": 1060
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.911471563637691e-05,
      "loss": 1.3425,
      "step": 1061
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.911219196318138e-05,
      "loss": 0.8982,
      "step": 1062
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.910966473021713e-05,
      "loss": 0.7481,
      "step": 1063
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.910713393766737e-05,
      "loss": 1.1054,
      "step": 1064
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.91045995857155e-05,
      "loss": 0.766,
      "step": 1065
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.910206167454526e-05,
      "loss": 0.9972,
      "step": 1066
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.909952020434056e-05,
      "loss": 0.8217,
      "step": 1067
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.909697517528564e-05,
      "loss": 1.1242,
      "step": 1068
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.909442658756497e-05,
      "loss": 0.7128,
      "step": 1069
    },
    {
      "epoch": 0.44,
      "learning_rate": 9.909187444136325e-05,
      "loss": 1.1622,
      "step": 1070
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.908931873686551e-05,
      "loss": 1.1037,
      "step": 1071
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.908675947425697e-05,
      "loss": 0.6895,
      "step": 1072
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.90841966537231e-05,
      "loss": 0.7042,
      "step": 1073
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.90816302754497e-05,
      "loss": 1.5569,
      "step": 1074
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.90790603396228e-05,
      "loss": 0.8289,
      "step": 1075
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.907648684642863e-05,
      "loss": 0.829,
      "step": 1076
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.907390979605375e-05,
      "loss": 1.8898,
      "step": 1077
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.907132918868495e-05,
      "loss": 0.948,
      "step": 1078
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.906874502450927e-05,
      "loss": 1.0177,
      "step": 1079
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.906615730371403e-05,
      "loss": 0.5747,
      "step": 1080
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.906356602648676e-05,
      "loss": 0.9446,
      "step": 1081
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.906097119301534e-05,
      "loss": 0.916,
      "step": 1082
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.90583728034878e-05,
      "loss": 1.6114,
      "step": 1083
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.905577085809249e-05,
      "loss": 1.9282,
      "step": 1084
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.905316535701801e-05,
      "loss": 1.1031,
      "step": 1085
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.905055630045321e-05,
      "loss": 0.8379,
      "step": 1086
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.904794368858721e-05,
      "loss": 1.2479,
      "step": 1087
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.904532752160937e-05,
      "loss": 1.208,
      "step": 1088
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.904270779970932e-05,
      "loss": 1.0471,
      "step": 1089
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.904008452307693e-05,
      "loss": 1.4658,
      "step": 1090
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.903745769190236e-05,
      "loss": 1.3575,
      "step": 1091
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.903482730637601e-05,
      "loss": 0.9512,
      "step": 1092
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.903219336668852e-05,
      "loss": 1.2994,
      "step": 1093
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.90295558730308e-05,
      "loss": 1.0761,
      "step": 1094
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.902691482559406e-05,
      "loss": 1.4257,
      "step": 1095
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.902427022456967e-05,
      "loss": 0.4968,
      "step": 1096
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.902162207014938e-05,
      "loss": 0.7172,
      "step": 1097
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.901897036252508e-05,
      "loss": 0.5307,
      "step": 1098
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.9016315101889e-05,
      "loss": 0.7369,
      "step": 1099
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.901365628843361e-05,
      "loss": 0.6659,
      "step": 1100
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.901099392235161e-05,
      "loss": 1.1859,
      "step": 1101
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.900832800383598e-05,
      "loss": 1.4144,
      "step": 1102
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.900565853307995e-05,
      "loss": 0.3864,
      "step": 1103
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.900298551027701e-05,
      "loss": 1.1005,
      "step": 1104
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.90003089356209e-05,
      "loss": 1.3981,
      "step": 1105
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.899762880930564e-05,
      "loss": 1.4543,
      "step": 1106
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.899494513152547e-05,
      "loss": 0.9658,
      "step": 1107
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.899225790247493e-05,
      "loss": 0.3592,
      "step": 1108
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.89895671223488e-05,
      "loss": 1.0306,
      "step": 1109
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.898687279134209e-05,
      "loss": 1.4755,
      "step": 1110
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.89841749096501e-05,
      "loss": 1.0183,
      "step": 1111
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.898147347746841e-05,
      "loss": 1.2145,
      "step": 1112
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.897876849499277e-05,
      "loss": 1.8392,
      "step": 1113
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.897605996241929e-05,
      "loss": 0.7522,
      "step": 1114
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.897334787994427e-05,
      "loss": 0.8121,
      "step": 1115
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.89706322477643e-05,
      "loss": 0.5325,
      "step": 1116
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.89679130660762e-05,
      "loss": 0.9924,
      "step": 1117
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.896519033507707e-05,
      "loss": 1.5609,
      "step": 1118
    },
    {
      "epoch": 0.46,
      "learning_rate": 9.896246405496426e-05,
      "loss": 1.1992,
      "step": 1119
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.895973422593538e-05,
      "loss": 1.2152,
      "step": 1120
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.895700084818831e-05,
      "loss": 0.9742,
      "step": 1121
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.895426392192114e-05,
      "loss": 1.0506,
      "step": 1122
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.895152344733227e-05,
      "loss": 0.7581,
      "step": 1123
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.894877942462033e-05,
      "loss": 0.8878,
      "step": 1124
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.894603185398421e-05,
      "loss": 1.1879,
      "step": 1125
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.894328073562307e-05,
      "loss": 0.9666,
      "step": 1126
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.89405260697363e-05,
      "loss": 0.6377,
      "step": 1127
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.893776785652358e-05,
      "loss": 0.5469,
      "step": 1128
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.893500609618482e-05,
      "loss": 0.7269,
      "step": 1129
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.893224078892022e-05,
      "loss": 1.254,
      "step": 1130
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.892947193493021e-05,
      "loss": 0.9957,
      "step": 1131
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.892669953441545e-05,
      "loss": 1.2844,
      "step": 1132
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.892392358757694e-05,
      "loss": 0.7688,
      "step": 1133
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.892114409461585e-05,
      "loss": 1.5051,
      "step": 1134
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.891836105573367e-05,
      "loss": 1.3412,
      "step": 1135
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.891557447113209e-05,
      "loss": 0.9004,
      "step": 1136
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.891278434101312e-05,
      "loss": 0.8088,
      "step": 1137
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.890999066557897e-05,
      "loss": 1.2679,
      "step": 1138
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.890719344503216e-05,
      "loss": 1.1258,
      "step": 1139
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.890439267957543e-05,
      "loss": 1.0213,
      "step": 1140
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.890158836941177e-05,
      "loss": 1.0103,
      "step": 1141
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.889878051474446e-05,
      "loss": 0.7637,
      "step": 1142
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.8895969115777e-05,
      "loss": 0.8668,
      "step": 1143
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.889315417271319e-05,
      "loss": 0.8934,
      "step": 1144
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.889033568575705e-05,
      "loss": 1.4817,
      "step": 1145
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.888751365511287e-05,
      "loss": 1.3401,
      "step": 1146
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.888468808098521e-05,
      "loss": 1.1959,
      "step": 1147
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.888185896357887e-05,
      "loss": 1.488,
      "step": 1148
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.88790263030989e-05,
      "loss": 0.992,
      "step": 1149
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.887619009975062e-05,
      "loss": 0.9644,
      "step": 1150
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.887335035373962e-05,
      "loss": 0.8841,
      "step": 1151
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.887050706527172e-05,
      "loss": 1.12,
      "step": 1152
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.886766023455301e-05,
      "loss": 1.5886,
      "step": 1153
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.886480986178985e-05,
      "loss": 0.9104,
      "step": 1154
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.886195594718881e-05,
      "loss": 1.39,
      "step": 1155
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.885909849095676e-05,
      "loss": 1.695,
      "step": 1156
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.885623749330084e-05,
      "loss": 0.6468,
      "step": 1157
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.885337295442839e-05,
      "loss": 0.9829,
      "step": 1158
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.885050487454705e-05,
      "loss": 0.8668,
      "step": 1159
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.884763325386472e-05,
      "loss": 0.4727,
      "step": 1160
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.884475809258952e-05,
      "loss": 1.5509,
      "step": 1161
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.884187939092986e-05,
      "loss": 1.0193,
      "step": 1162
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.883899714909438e-05,
      "loss": 1.2406,
      "step": 1163
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.883611136729202e-05,
      "loss": 0.6134,
      "step": 1164
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.883322204573192e-05,
      "loss": 0.7997,
      "step": 1165
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.883032918462353e-05,
      "loss": 1.1848,
      "step": 1166
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.88274327841765e-05,
      "loss": 0.9887,
      "step": 1167
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.88245328446008e-05,
      "loss": 1.2125,
      "step": 1168
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.882162936610662e-05,
      "loss": 1.2363,
      "step": 1169
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.88187223489044e-05,
      "loss": 1.0586,
      "step": 1170
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.881581179320484e-05,
      "loss": 1.0584,
      "step": 1171
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.88128976992189e-05,
      "loss": 0.993,
      "step": 1172
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.880998006715785e-05,
      "loss": 1.5924,
      "step": 1173
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.880705889723311e-05,
      "loss": 1.6305,
      "step": 1174
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.880413418965644e-05,
      "loss": 1.3579,
      "step": 1175
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.880120594463982e-05,
      "loss": 1.2155,
      "step": 1176
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.879827416239552e-05,
      "loss": 0.923,
      "step": 1177
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.879533884313598e-05,
      "loss": 0.9215,
      "step": 1178
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.879239998707403e-05,
      "loss": 1.2723,
      "step": 1179
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.878945759442266e-05,
      "loss": 1.3883,
      "step": 1180
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.878651166539514e-05,
      "loss": 0.6212,
      "step": 1181
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.878356220020499e-05,
      "loss": 1.1763,
      "step": 1182
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.8780609199066e-05,
      "loss": 1.4452,
      "step": 1183
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.877765266219223e-05,
      "loss": 0.8639,
      "step": 1184
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.877469258979793e-05,
      "loss": 1.0509,
      "step": 1185
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.87717289820977e-05,
      "loss": 1.2525,
      "step": 1186
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.876876183930632e-05,
      "loss": 1.6002,
      "step": 1187
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.876579116163888e-05,
      "loss": 1.0428,
      "step": 1188
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.876281694931069e-05,
      "loss": 1.0756,
      "step": 1189
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.87598392025373e-05,
      "loss": 1.5023,
      "step": 1190
    },
    {
      "epoch": 0.49,
      "learning_rate": 9.87568579215346e-05,
      "loss": 0.7724,
      "step": 1191
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.875387310651865e-05,
      "loss": 1.6034,
      "step": 1192
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.875088475770578e-05,
      "loss": 0.8979,
      "step": 1193
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.874789287531262e-05,
      "loss": 1.009,
      "step": 1194
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.874489745955601e-05,
      "loss": 1.312,
      "step": 1195
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.87418985106531e-05,
      "loss": 1.2171,
      "step": 1196
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.87388960288212e-05,
      "loss": 1.0322,
      "step": 1197
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.873589001427798e-05,
      "loss": 0.9128,
      "step": 1198
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.873288046724134e-05,
      "loss": 1.2735,
      "step": 1199
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.872986738792938e-05,
      "loss": 0.8964,
      "step": 1200
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.872685077656049e-05,
      "loss": 0.6892,
      "step": 1201
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.872383063335337e-05,
      "loss": 0.6708,
      "step": 1202
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.872080695852687e-05,
      "loss": 1.3117,
      "step": 1203
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.871777975230021e-05,
      "loss": 1.5649,
      "step": 1204
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.871474901489277e-05,
      "loss": 1.1225,
      "step": 1205
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.871171474652422e-05,
      "loss": 1.0763,
      "step": 1206
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.870867694741452e-05,
      "loss": 1.1602,
      "step": 1207
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.870563561778384e-05,
      "loss": 1.401,
      "step": 1208
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.870259075785262e-05,
      "loss": 1.1325,
      "step": 1209
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.869954236784158e-05,
      "loss": 1.0669,
      "step": 1210
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.869649044797164e-05,
      "loss": 1.7789,
      "step": 1211
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.869343499846405e-05,
      "loss": 1.193,
      "step": 1212
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.869037601954023e-05,
      "loss": 1.1076,
      "step": 1213
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.868731351142195e-05,
      "loss": 1.3788,
      "step": 1214
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.868424747433115e-05,
      "loss": 0.7027,
      "step": 1215
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.868117790849009e-05,
      "loss": 1.4434,
      "step": 1216
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.867810481412126e-05,
      "loss": 0.8159,
      "step": 1217
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.867502819144736e-05,
      "loss": 0.9966,
      "step": 1218
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.867194804069145e-05,
      "loss": 0.9548,
      "step": 1219
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.866886436207676e-05,
      "loss": 1.4326,
      "step": 1220
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.866577715582678e-05,
      "loss": 0.8055,
      "step": 1221
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.866268642216534e-05,
      "loss": 1.1675,
      "step": 1222
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.86595921613164e-05,
      "loss": 0.6975,
      "step": 1223
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.865649437350428e-05,
      "loss": 1.2753,
      "step": 1224
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.865339305895349e-05,
      "loss": 0.4357,
      "step": 1225
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.865028821788884e-05,
      "loss": 0.947,
      "step": 1226
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.864717985053535e-05,
      "loss": 1.1541,
      "step": 1227
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.864406795711835e-05,
      "loss": 1.3642,
      "step": 1228
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.864095253786338e-05,
      "loss": 1.1577,
      "step": 1229
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.863783359299628e-05,
      "loss": 1.1032,
      "step": 1230
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.863471112274308e-05,
      "loss": 1.5778,
      "step": 1231
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.863158512733013e-05,
      "loss": 0.6415,
      "step": 1232
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.862845560698399e-05,
      "loss": 0.9474,
      "step": 1233
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.862532256193152e-05,
      "loss": 1.3849,
      "step": 1234
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.862218599239979e-05,
      "loss": 0.6968,
      "step": 1235
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.861904589861616e-05,
      "loss": 0.7776,
      "step": 1236
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.861590228080822e-05,
      "loss": 0.8668,
      "step": 1237
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.861275513920382e-05,
      "loss": 0.9488,
      "step": 1238
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.860960447403111e-05,
      "loss": 0.9345,
      "step": 1239
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.860645028551841e-05,
      "loss": 1.811,
      "step": 1240
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.860329257389438e-05,
      "loss": 0.5921,
      "step": 1241
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.860013133938788e-05,
      "loss": 0.7987,
      "step": 1242
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.859696658222804e-05,
      "loss": 1.0626,
      "step": 1243
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.859379830264425e-05,
      "loss": 1.0481,
      "step": 1244
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.859062650086619e-05,
      "loss": 1.1505,
      "step": 1245
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.858745117712371e-05,
      "loss": 0.877,
      "step": 1246
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.858427233164698e-05,
      "loss": 1.4248,
      "step": 1247
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.858108996466642e-05,
      "loss": 0.7537,
      "step": 1248
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.85779040764127e-05,
      "loss": 1.0887,
      "step": 1249
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.857471466711671e-05,
      "loss": 1.3096,
      "step": 1250
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.857152173700968e-05,
      "loss": 0.9068,
      "step": 1251
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.8568325286323e-05,
      "loss": 0.925,
      "step": 1252
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.856512531528837e-05,
      "loss": 1.0395,
      "step": 1253
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.85619218241377e-05,
      "loss": 0.7998,
      "step": 1254
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.855871481310324e-05,
      "loss": 1.2018,
      "step": 1255
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.85555042824174e-05,
      "loss": 0.7852,
      "step": 1256
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.855229023231292e-05,
      "loss": 0.6296,
      "step": 1257
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.854907266302273e-05,
      "loss": 1.2584,
      "step": 1258
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.854585157478007e-05,
      "loss": 0.8289,
      "step": 1259
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.85426269678184e-05,
      "loss": 1.2396,
      "step": 1260
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.853939884237144e-05,
      "loss": 1.0722,
      "step": 1261
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.853616719867321e-05,
      "loss": 1.2676,
      "step": 1262
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.85329320369579e-05,
      "loss": 1.59,
      "step": 1263
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.852969335746003e-05,
      "loss": 1.0316,
      "step": 1264
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.852645116041434e-05,
      "loss": 1.3362,
      "step": 1265
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.852320544605582e-05,
      "loss": 0.9036,
      "step": 1266
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.851995621461976e-05,
      "loss": 1.1871,
      "step": 1267
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.851670346634163e-05,
      "loss": 1.1798,
      "step": 1268
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.851344720145723e-05,
      "loss": 0.6916,
      "step": 1269
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.851018742020257e-05,
      "loss": 0.8193,
      "step": 1270
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.850692412281393e-05,
      "loss": 0.7642,
      "step": 1271
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.850365730952784e-05,
      "loss": 1.0293,
      "step": 1272
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.850038698058108e-05,
      "loss": 1.264,
      "step": 1273
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.849711313621069e-05,
      "loss": 1.0546,
      "step": 1274
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.849383577665399e-05,
      "loss": 1.3445,
      "step": 1275
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.84905549021485e-05,
      "loss": 1.0747,
      "step": 1276
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.848727051293202e-05,
      "loss": 1.2008,
      "step": 1277
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.848398260924266e-05,
      "loss": 0.9837,
      "step": 1278
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.848069119131869e-05,
      "loss": 1.0764,
      "step": 1279
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.847739625939869e-05,
      "loss": 0.583,
      "step": 1280
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.84740978137215e-05,
      "loss": 0.9003,
      "step": 1281
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.847079585452619e-05,
      "loss": 1.4365,
      "step": 1282
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.846749038205208e-05,
      "loss": 0.9002,
      "step": 1283
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.846418139653877e-05,
      "loss": 0.5473,
      "step": 1284
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.84608688982261e-05,
      "loss": 1.181,
      "step": 1285
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.845755288735418e-05,
      "loss": 1.057,
      "step": 1286
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.845423336416335e-05,
      "loss": 1.5769,
      "step": 1287
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.845091032889422e-05,
      "loss": 1.2061,
      "step": 1288
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.844758378178766e-05,
      "loss": 1.6596,
      "step": 1289
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.844425372308477e-05,
      "loss": 1.1799,
      "step": 1290
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.844092015302693e-05,
      "loss": 1.2264,
      "step": 1291
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.843758307185575e-05,
      "loss": 1.0789,
      "step": 1292
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.843424247981314e-05,
      "loss": 1.669,
      "step": 1293
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.843089837714121e-05,
      "loss": 1.247,
      "step": 1294
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.842755076408237e-05,
      "loss": 1.0607,
      "step": 1295
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.842419964087923e-05,
      "loss": 1.1449,
      "step": 1296
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.842084500777471e-05,
      "loss": 0.7156,
      "step": 1297
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.841748686501196e-05,
      "loss": 1.1906,
      "step": 1298
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.841412521283437e-05,
      "loss": 1.3198,
      "step": 1299
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.841076005148564e-05,
      "loss": 0.9773,
      "step": 1300
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.840739138120965e-05,
      "loss": 1.0596,
      "step": 1301
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.840401920225058e-05,
      "loss": 1.1387,
      "step": 1302
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.840064351485285e-05,
      "loss": 1.2758,
      "step": 1303
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.839726431926113e-05,
      "loss": 0.9365,
      "step": 1304
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.839388161572037e-05,
      "loss": 1.1233,
      "step": 1305
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.839049540447575e-05,
      "loss": 1.2032,
      "step": 1306
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.838710568577272e-05,
      "loss": 1.189,
      "step": 1307
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.838371245985695e-05,
      "loss": 1.5459,
      "step": 1308
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.838031572697441e-05,
      "loss": 1.2757,
      "step": 1309
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.837691548737129e-05,
      "loss": 1.1922,
      "step": 1310
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.837351174129405e-05,
      "loss": 1.3073,
      "step": 1311
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.837010448898942e-05,
      "loss": 1.0524,
      "step": 1312
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.836669373070434e-05,
      "loss": 0.6714,
      "step": 1313
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.836327946668605e-05,
      "loss": 0.7901,
      "step": 1314
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.8359861697182e-05,
      "loss": 1.0084,
      "step": 1315
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.835644042243995e-05,
      "loss": 0.9282,
      "step": 1316
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.835301564270787e-05,
      "loss": 1.5038,
      "step": 1317
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.834958735823398e-05,
      "loss": 0.8624,
      "step": 1318
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.834615556926679e-05,
      "loss": 0.9219,
      "step": 1319
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.834272027605503e-05,
      "loss": 0.7232,
      "step": 1320
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.83392814788477e-05,
      "loss": 1.435,
      "step": 1321
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.833583917789406e-05,
      "loss": 0.8736,
      "step": 1322
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.833239337344362e-05,
      "loss": 1.0563,
      "step": 1323
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.832894406574614e-05,
      "loss": 0.6836,
      "step": 1324
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.832549125505159e-05,
      "loss": 1.5351,
      "step": 1325
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.832203494161031e-05,
      "loss": 0.8341,
      "step": 1326
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.831857512567277e-05,
      "loss": 1.2002,
      "step": 1327
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.831511180748978e-05,
      "loss": 1.0965,
      "step": 1328
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.831164498731234e-05,
      "loss": 0.7322,
      "step": 1329
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.830817466539175e-05,
      "loss": 0.8522,
      "step": 1330
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.830470084197953e-05,
      "loss": 0.7699,
      "step": 1331
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.83012235173275e-05,
      "loss": 1.1437,
      "step": 1332
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.829774269168767e-05,
      "loss": 0.9244,
      "step": 1333
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.829425836531237e-05,
      "loss": 0.5293,
      "step": 1334
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.829077053845415e-05,
      "loss": 1.064,
      "step": 1335
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.82872792113658e-05,
      "loss": 0.5555,
      "step": 1336
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.828378438430037e-05,
      "loss": 1.9056,
      "step": 1337
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.828028605751119e-05,
      "loss": 0.7969,
      "step": 1338
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.827678423125184e-05,
      "loss": 1.5462,
      "step": 1339
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.827327890577612e-05,
      "loss": 1.752,
      "step": 1340
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.826977008133811e-05,
      "loss": 0.8053,
      "step": 1341
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.826625775819213e-05,
      "loss": 1.0906,
      "step": 1342
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.826274193659279e-05,
      "loss": 0.8553,
      "step": 1343
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.825922261679488e-05,
      "loss": 0.778,
      "step": 1344
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.825569979905353e-05,
      "loss": 0.9803,
      "step": 1345
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.825217348362405e-05,
      "loss": 1.3144,
      "step": 1346
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.824864367076207e-05,
      "loss": 1.484,
      "step": 1347
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.824511036072343e-05,
      "loss": 1.3674,
      "step": 1348
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.82415735537642e-05,
      "loss": 1.2864,
      "step": 1349
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.823803325014078e-05,
      "loss": 0.9288,
      "step": 1350
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.823448945010974e-05,
      "loss": 1.0438,
      "step": 1351
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.823094215392798e-05,
      "loss": 0.7183,
      "step": 1352
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.822739136185259e-05,
      "loss": 1.1801,
      "step": 1353
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.822383707414094e-05,
      "loss": 0.9658,
      "step": 1354
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.822027929105067e-05,
      "loss": 0.8763,
      "step": 1355
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.821671801283965e-05,
      "loss": 1.0155,
      "step": 1356
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.821315323976601e-05,
      "loss": 1.4441,
      "step": 1357
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.820958497208814e-05,
      "loss": 0.8321,
      "step": 1358
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.820601321006466e-05,
      "loss": 1.4242,
      "step": 1359
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.820243795395448e-05,
      "loss": 0.6153,
      "step": 1360
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.81988592040167e-05,
      "loss": 0.7282,
      "step": 1361
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.819527696051078e-05,
      "loss": 1.1493,
      "step": 1362
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.819169122369632e-05,
      "loss": 1.7738,
      "step": 1363
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.818810199383325e-05,
      "loss": 0.7511,
      "step": 1364
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.81845092711817e-05,
      "loss": 0.9926,
      "step": 1365
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.818091305600213e-05,
      "loss": 1.0244,
      "step": 1366
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.817731334855513e-05,
      "loss": 0.6335,
      "step": 1367
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.817371014910166e-05,
      "loss": 1.0961,
      "step": 1368
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.817010345790289e-05,
      "loss": 1.0537,
      "step": 1369
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.816649327522024e-05,
      "loss": 0.7389,
      "step": 1370
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.816287960131535e-05,
      "loss": 0.9916,
      "step": 1371
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.815926243645022e-05,
      "loss": 1.2415,
      "step": 1372
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.815564178088694e-05,
      "loss": 0.8219,
      "step": 1373
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.815201763488802e-05,
      "loss": 1.3507,
      "step": 1374
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.814838999871611e-05,
      "loss": 1.5466,
      "step": 1375
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.814475887263414e-05,
      "loss": 0.8171,
      "step": 1376
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.814112425690534e-05,
      "loss": 1.3411,
      "step": 1377
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.813748615179312e-05,
      "loss": 1.4085,
      "step": 1378
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.81338445575612e-05,
      "loss": 0.7753,
      "step": 1379
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.813019947447353e-05,
      "loss": 0.6202,
      "step": 1380
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.81265509027943e-05,
      "loss": 1.1676,
      "step": 1381
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.812289884278798e-05,
      "loss": 1.3807,
      "step": 1382
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.811924329471926e-05,
      "loss": 0.6659,
      "step": 1383
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.811558425885315e-05,
      "loss": 1.5427,
      "step": 1384
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.81119217354548e-05,
      "loss": 0.5468,
      "step": 1385
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.810825572478974e-05,
      "loss": 0.9225,
      "step": 1386
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.810458622712364e-05,
      "loss": 1.3172,
      "step": 1387
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.810091324272252e-05,
      "loss": 1.2072,
      "step": 1388
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.809723677185256e-05,
      "loss": 1.6078,
      "step": 1389
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.809355681478029e-05,
      "loss": 0.8604,
      "step": 1390
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.808987337177241e-05,
      "loss": 0.6314,
      "step": 1391
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.808618644309589e-05,
      "loss": 1.0177,
      "step": 1392
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.808249602901802e-05,
      "loss": 1.1677,
      "step": 1393
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.807880212980623e-05,
      "loss": 1.8194,
      "step": 1394
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.80751047457283e-05,
      "loss": 0.6266,
      "step": 1395
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.807140387705222e-05,
      "loss": 1.0381,
      "step": 1396
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.806769952404624e-05,
      "loss": 1.1612,
      "step": 1397
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.806399168697885e-05,
      "loss": 1.8251,
      "step": 1398
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.806028036611878e-05,
      "loss": 0.8887,
      "step": 1399
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.80565655617351e-05,
      "loss": 0.9781,
      "step": 1400
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.805284727409701e-05,
      "loss": 1.0997,
      "step": 1401
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.804912550347405e-05,
      "loss": 0.7924,
      "step": 1402
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.804540025013596e-05,
      "loss": 1.1751,
      "step": 1403
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.804167151435278e-05,
      "loss": 0.7893,
      "step": 1404
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.803793929639477e-05,
      "loss": 0.8783,
      "step": 1405
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.803420359653243e-05,
      "loss": 0.6387,
      "step": 1406
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.803046441503656e-05,
      "loss": 1.121,
      "step": 1407
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.802672175217818e-05,
      "loss": 0.9843,
      "step": 1408
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.802297560822855e-05,
      "loss": 0.5497,
      "step": 1409
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.801922598345921e-05,
      "loss": 0.55,
      "step": 1410
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.801547287814195e-05,
      "loss": 0.995,
      "step": 1411
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.801171629254879e-05,
      "loss": 0.6527,
      "step": 1412
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.800795622695202e-05,
      "loss": 1.0574,
      "step": 1413
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.800419268162417e-05,
      "loss": 1.2567,
      "step": 1414
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.800042565683806e-05,
      "loss": 0.8715,
      "step": 1415
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.79966551528667e-05,
      "loss": 1.1835,
      "step": 1416
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.799288116998342e-05,
      "loss": 0.8517,
      "step": 1417
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.798910370846173e-05,
      "loss": 1.3771,
      "step": 1418
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.798532276857545e-05,
      "loss": 1.4137,
      "step": 1419
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.798153835059862e-05,
      "loss": 0.5575,
      "step": 1420
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.797775045480556e-05,
      "loss": 1.1729,
      "step": 1421
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.797395908147082e-05,
      "loss": 1.3954,
      "step": 1422
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.79701642308692e-05,
      "loss": 0.8542,
      "step": 1423
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.796636590327575e-05,
      "loss": 1.219,
      "step": 1424
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.796256409896583e-05,
      "loss": 1.318,
      "step": 1425
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.795875881821494e-05,
      "loss": 0.9548,
      "step": 1426
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.795495006129894e-05,
      "loss": 0.9886,
      "step": 1427
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.795113782849388e-05,
      "loss": 0.7141,
      "step": 1428
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.794732212007609e-05,
      "loss": 0.9759,
      "step": 1429
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.794350293632211e-05,
      "loss": 1.2214,
      "step": 1430
    },
    {
      "epoch": 0.59,
      "learning_rate": 9.793968027750882e-05,
      "loss": 1.4847,
      "step": 1431
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.793585414391325e-05,
      "loss": 1.3493,
      "step": 1432
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.793202453581274e-05,
      "loss": 1.047,
      "step": 1433
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.792819145348488e-05,
      "loss": 1.5809,
      "step": 1434
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.792435489720748e-05,
      "loss": 0.9223,
      "step": 1435
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.792051486725863e-05,
      "loss": 1.346,
      "step": 1436
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.791667136391669e-05,
      "loss": 0.6076,
      "step": 1437
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.791282438746021e-05,
      "loss": 1.2415,
      "step": 1438
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.790897393816804e-05,
      "loss": 0.4542,
      "step": 1439
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.790512001631929e-05,
      "loss": 1.2678,
      "step": 1440
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.790126262219326e-05,
      "loss": 0.7801,
      "step": 1441
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.789740175606959e-05,
      "loss": 1.2695,
      "step": 1442
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.789353741822809e-05,
      "loss": 0.8707,
      "step": 1443
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.788966960894887e-05,
      "loss": 0.5229,
      "step": 1444
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.788579832851228e-05,
      "loss": 1.9298,
      "step": 1445
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.78819235771989e-05,
      "loss": 1.4821,
      "step": 1446
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.787804535528962e-05,
      "loss": 1.6006,
      "step": 1447
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.78741636630655e-05,
      "loss": 1.6154,
      "step": 1448
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.787027850080792e-05,
      "loss": 1.0367,
      "step": 1449
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.786638986879849e-05,
      "loss": 1.0049,
      "step": 1450
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.786249776731905e-05,
      "loss": 0.7207,
      "step": 1451
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.785860219665173e-05,
      "loss": 1.2081,
      "step": 1452
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.785470315707886e-05,
      "loss": 0.9467,
      "step": 1453
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.785080064888308e-05,
      "loss": 0.7889,
      "step": 1454
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.784689467234722e-05,
      "loss": 0.845,
      "step": 1455
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.784298522775443e-05,
      "loss": 1.2675,
      "step": 1456
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.783907231538808e-05,
      "loss": 1.3039,
      "step": 1457
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.783515593553174e-05,
      "loss": 1.3197,
      "step": 1458
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.783123608846931e-05,
      "loss": 0.7745,
      "step": 1459
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.782731277448492e-05,
      "loss": 1.6534,
      "step": 1460
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.782338599386293e-05,
      "loss": 1.4997,
      "step": 1461
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.781945574688796e-05,
      "loss": 1.0093,
      "step": 1462
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.781552203384487e-05,
      "loss": 1.4809,
      "step": 1463
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.781158485501881e-05,
      "loss": 0.9086,
      "step": 1464
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.780764421069514e-05,
      "loss": 1.2186,
      "step": 1465
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.78037001011595e-05,
      "loss": 0.968,
      "step": 1466
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.779975252669777e-05,
      "loss": 0.73,
      "step": 1467
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.779580148759605e-05,
      "loss": 0.8861,
      "step": 1468
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.779184698414074e-05,
      "loss": 1.1082,
      "step": 1469
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.778788901661849e-05,
      "loss": 1.2972,
      "step": 1470
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.778392758531616e-05,
      "loss": 0.9516,
      "step": 1471
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.77799626905209e-05,
      "loss": 1.2814,
      "step": 1472
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.777599433252007e-05,
      "loss": 1.2004,
      "step": 1473
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.777202251160133e-05,
      "loss": 1.0882,
      "step": 1474
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.776804722805255e-05,
      "loss": 1.1657,
      "step": 1475
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.776406848216188e-05,
      "loss": 0.7469,
      "step": 1476
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.776008627421772e-05,
      "loss": 1.1913,
      "step": 1477
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.775610060450868e-05,
      "loss": 1.3964,
      "step": 1478
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.775211147332368e-05,
      "loss": 1.2058,
      "step": 1479
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.774811888095184e-05,
      "loss": 1.0229,
      "step": 1480
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.774412282768255e-05,
      "loss": 1.0941,
      "step": 1481
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.774012331380547e-05,
      "loss": 1.1467,
      "step": 1482
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.77361203396105e-05,
      "loss": 0.6582,
      "step": 1483
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.773211390538776e-05,
      "loss": 0.7552,
      "step": 1484
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.772810401142765e-05,
      "loss": 0.7737,
      "step": 1485
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.772409065802085e-05,
      "loss": 0.8262,
      "step": 1486
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.772007384545822e-05,
      "loss": 1.2875,
      "step": 1487
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.771605357403092e-05,
      "loss": 1.3107,
      "step": 1488
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.771202984403037e-05,
      "loss": 1.0701,
      "step": 1489
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.770800265574817e-05,
      "loss": 0.8476,
      "step": 1490
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.770397200947628e-05,
      "loss": 1.9378,
      "step": 1491
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.76999379055068e-05,
      "loss": 1.2172,
      "step": 1492
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.769590034413217e-05,
      "loss": 0.7995,
      "step": 1493
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.7691859325645e-05,
      "loss": 0.8076,
      "step": 1494
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.768781485033822e-05,
      "loss": 1.5861,
      "step": 1495
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.7683766918505e-05,
      "loss": 0.8326,
      "step": 1496
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.76797155304387e-05,
      "loss": 1.42,
      "step": 1497
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.767566068643303e-05,
      "loss": 0.9559,
      "step": 1498
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.767160238678183e-05,
      "loss": 1.1644,
      "step": 1499
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.766754063177932e-05,
      "loss": 1.6485,
      "step": 1500
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.766347542171986e-05,
      "loss": 1.1684,
      "step": 1501
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.765940675689811e-05,
      "loss": 0.64,
      "step": 1502
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.7655334637609e-05,
      "loss": 1.3017,
      "step": 1503
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.765125906414767e-05,
      "loss": 1.5974,
      "step": 1504
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.764718003680953e-05,
      "loss": 1.0435,
      "step": 1505
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.764309755589025e-05,
      "loss": 1.3546,
      "step": 1506
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.76390116216857e-05,
      "loss": 1.0583,
      "step": 1507
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.763492223449208e-05,
      "loss": 1.3467,
      "step": 1508
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.76308293946058e-05,
      "loss": 1.3772,
      "step": 1509
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.762673310232348e-05,
      "loss": 0.8977,
      "step": 1510
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.762263335794206e-05,
      "loss": 0.819,
      "step": 1511
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.761853016175868e-05,
      "loss": 1.1758,
      "step": 1512
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.761442351407077e-05,
      "loss": 1.09,
      "step": 1513
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.761031341517597e-05,
      "loss": 1.7908,
      "step": 1514
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.76061998653722e-05,
      "loss": 1.3178,
      "step": 1515
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.760208286495762e-05,
      "loss": 0.6496,
      "step": 1516
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.759796241423065e-05,
      "loss": 1.064,
      "step": 1517
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.759383851348994e-05,
      "loss": 0.9929,
      "step": 1518
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.758971116303438e-05,
      "loss": 0.9149,
      "step": 1519
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.758558036316316e-05,
      "loss": 1.4125,
      "step": 1520
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.758144611417568e-05,
      "loss": 0.8205,
      "step": 1521
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.75773084163716e-05,
      "loss": 1.1403,
      "step": 1522
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.757316727005083e-05,
      "loss": 1.0376,
      "step": 1523
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.756902267551353e-05,
      "loss": 0.999,
      "step": 1524
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.756487463306013e-05,
      "loss": 1.1976,
      "step": 1525
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.756072314299123e-05,
      "loss": 1.058,
      "step": 1526
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.755656820560781e-05,
      "loss": 0.7424,
      "step": 1527
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.755240982121099e-05,
      "loss": 0.9441,
      "step": 1528
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.754824799010218e-05,
      "loss": 1.0586,
      "step": 1529
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.754408271258307e-05,
      "loss": 0.8661,
      "step": 1530
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.753991398895554e-05,
      "loss": 1.2476,
      "step": 1531
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.753574181952175e-05,
      "loss": 1.0341,
      "step": 1532
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.753156620458413e-05,
      "loss": 1.0798,
      "step": 1533
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.752738714444532e-05,
      "loss": 0.5193,
      "step": 1534
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.752320463940821e-05,
      "loss": 0.7797,
      "step": 1535
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.7519018689776e-05,
      "loss": 1.7497,
      "step": 1536
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.751482929585208e-05,
      "loss": 0.9404,
      "step": 1537
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.751063645794011e-05,
      "loss": 1.1871,
      "step": 1538
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.750644017634398e-05,
      "loss": 0.6839,
      "step": 1539
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.750224045136787e-05,
      "loss": 1.0963,
      "step": 1540
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.749803728331617e-05,
      "loss": 0.8174,
      "step": 1541
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.749383067249354e-05,
      "loss": 0.8001,
      "step": 1542
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.748962061920489e-05,
      "loss": 1.4983,
      "step": 1543
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.748540712375535e-05,
      "loss": 1.4872,
      "step": 1544
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.748119018645038e-05,
      "loss": 1.1885,
      "step": 1545
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.747696980759557e-05,
      "loss": 0.6873,
      "step": 1546
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.747274598749688e-05,
      "loss": 1.2855,
      "step": 1547
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.746851872646041e-05,
      "loss": 0.9611,
      "step": 1548
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.746428802479259e-05,
      "loss": 0.7602,
      "step": 1549
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.746005388280006e-05,
      "loss": 0.79,
      "step": 1550
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.745581630078975e-05,
      "loss": 0.4968,
      "step": 1551
    },
    {
      "epoch": 0.64,
      "learning_rate": 9.745157527906877e-05,
      "loss": 0.6259,
      "step": 1552
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.744733081794455e-05,
      "loss": 0.9694,
      "step": 1553
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.744308291772471e-05,
      "loss": 0.8717,
      "step": 1554
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.743883157871717e-05,
      "loss": 1.077,
      "step": 1555
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.743457680123007e-05,
      "loss": 0.7497,
      "step": 1556
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.743031858557182e-05,
      "loss": 0.7723,
      "step": 1557
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.742605693205105e-05,
      "loss": 1.0682,
      "step": 1558
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.742179184097665e-05,
      "loss": 1.6674,
      "step": 1559
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.741752331265778e-05,
      "loss": 1.0748,
      "step": 1560
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.741325134740382e-05,
      "loss": 0.5224,
      "step": 1561
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.740897594552441e-05,
      "loss": 1.2491,
      "step": 1562
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.740469710732947e-05,
      "loss": 1.1478,
      "step": 1563
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.740041483312909e-05,
      "loss": 0.8254,
      "step": 1564
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.73961291232337e-05,
      "loss": 1.2563,
      "step": 1565
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.739183997795394e-05,
      "loss": 0.5917,
      "step": 1566
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.738754739760067e-05,
      "loss": 1.3615,
      "step": 1567
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.738325138248506e-05,
      "loss": 1.2398,
      "step": 1568
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.737895193291847e-05,
      "loss": 1.3073,
      "step": 1569
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.737464904921252e-05,
      "loss": 0.8856,
      "step": 1570
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.737034273167914e-05,
      "loss": 1.3088,
      "step": 1571
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.736603298063041e-05,
      "loss": 0.7887,
      "step": 1572
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.736171979637876e-05,
      "loss": 1.4707,
      "step": 1573
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.735740317923678e-05,
      "loss": 0.738,
      "step": 1574
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.735308312951738e-05,
      "loss": 0.9632,
      "step": 1575
    },
    {
      "epoch": 0.65,
      "learning_rate": 9.734875964753367e-05,
      "loss": 1.2701,
      "step": 1576
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.734443273359902e-05,
      "loss": 0.7024,
      "step": 1577
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.734010238802707e-05,
      "loss": 0.9572,
      "step": 1578
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.733576861113169e-05,
      "loss": 1.1709,
      "step": 1579
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.7331431403227e-05,
      "loss": 2.0283,
      "step": 1580
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.732709076462737e-05,
      "loss": 1.4736,
      "step": 1581
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.732274669564743e-05,
      "loss": 1.2938,
      "step": 1582
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.731839919660202e-05,
      "loss": 1.237,
      "step": 1583
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.731404826780631e-05,
      "loss": 1.2042,
      "step": 1584
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.73096939095756e-05,
      "loss": 0.8927,
      "step": 1585
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.730533612222556e-05,
      "loss": 1.3003,
      "step": 1586
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.730097490607202e-05,
      "loss": 1.9445,
      "step": 1587
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.72966102614311e-05,
      "loss": 1.1104,
      "step": 1588
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.729224218861917e-05,
      "loss": 0.7476,
      "step": 1589
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.728787068795283e-05,
      "loss": 0.767,
      "step": 1590
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.728349575974891e-05,
      "loss": 1.202,
      "step": 1591
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.727911740432456e-05,
      "loss": 0.9595,
      "step": 1592
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.727473562199712e-05,
      "loss": 1.0988,
      "step": 1593
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.727035041308419e-05,
      "loss": 1.4318,
      "step": 1594
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.726596177790361e-05,
      "loss": 0.8124,
      "step": 1595
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.726156971677349e-05,
      "loss": 1.0742,
      "step": 1596
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.725717423001217e-05,
      "loss": 0.8761,
      "step": 1597
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.725277531793824e-05,
      "loss": 1.7509,
      "step": 1598
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.724837298087057e-05,
      "loss": 1.4523,
      "step": 1599
    },
    {
      "epoch": 0.66,
      "learning_rate": 9.724396721912823e-05,
      "loss": 0.9962,
      "step": 1600
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.723955803303056e-05,
      "loss": 1.3194,
      "step": 1601
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.723514542289715e-05,
      "loss": 0.3947,
      "step": 1602
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.723072938904785e-05,
      "loss": 0.7461,
      "step": 1603
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.722630993180272e-05,
      "loss": 0.8193,
      "step": 1604
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.722188705148212e-05,
      "loss": 1.4517,
      "step": 1605
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.72174607484066e-05,
      "loss": 1.3442,
      "step": 1606
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.721303102289703e-05,
      "loss": 1.1357,
      "step": 1607
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.720859787527444e-05,
      "loss": 0.6152,
      "step": 1608
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.720416130586019e-05,
      "loss": 1.348,
      "step": 1609
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.719972131497585e-05,
      "loss": 1.2373,
      "step": 1610
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.719527790294323e-05,
      "loss": 1.4909,
      "step": 1611
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.719083107008438e-05,
      "loss": 0.7157,
      "step": 1612
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.718638081672166e-05,
      "loss": 1.2389,
      "step": 1613
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.718192714317762e-05,
      "loss": 1.153,
      "step": 1614
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.717747004977504e-05,
      "loss": 1.2192,
      "step": 1615
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.717300953683703e-05,
      "loss": 1.2349,
      "step": 1616
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.716854560468688e-05,
      "loss": 1.1932,
      "step": 1617
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.716407825364813e-05,
      "loss": 0.8164,
      "step": 1618
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.71596074840446e-05,
      "loss": 0.7252,
      "step": 1619
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.715513329620035e-05,
      "loss": 1.4285,
      "step": 1620
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.715065569043964e-05,
      "loss": 0.7427,
      "step": 1621
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.714617466708708e-05,
      "loss": 0.8994,
      "step": 1622
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.714169022646741e-05,
      "loss": 0.3128,
      "step": 1623
    },
    {
      "epoch": 0.67,
      "learning_rate": 9.71372023689057e-05,
      "loss": 0.8044,
      "step": 1624
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.713271109472724e-05,
      "loss": 0.7431,
      "step": 1625
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.712821640425756e-05,
      "loss": 1.1159,
      "step": 1626
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.712371829782245e-05,
      "loss": 0.9064,
      "step": 1627
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.711921677574794e-05,
      "loss": 1.3196,
      "step": 1628
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.711471183836032e-05,
      "loss": 0.932,
      "step": 1629
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.71102034859861e-05,
      "loss": 1.3747,
      "step": 1630
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.710569171895207e-05,
      "loss": 1.3088,
      "step": 1631
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.710117653758525e-05,
      "loss": 1.3412,
      "step": 1632
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.709665794221294e-05,
      "loss": 0.5368,
      "step": 1633
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.70921359331626e-05,
      "loss": 1.0651,
      "step": 1634
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.708761051076205e-05,
      "loss": 1.2471,
      "step": 1635
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.708308167533927e-05,
      "loss": 1.4674,
      "step": 1636
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.707854942722254e-05,
      "loss": 0.9631,
      "step": 1637
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.707401376674035e-05,
      "loss": 1.2996,
      "step": 1638
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.706947469422147e-05,
      "loss": 0.8663,
      "step": 1639
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.706493220999491e-05,
      "loss": 0.6748,
      "step": 1640
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.70603863143899e-05,
      "loss": 1.4382,
      "step": 1641
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.705583700773593e-05,
      "loss": 1.0381,
      "step": 1642
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.705128429036278e-05,
      "loss": 1.7218,
      "step": 1643
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.704672816260041e-05,
      "loss": 1.9546,
      "step": 1644
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.704216862477908e-05,
      "loss": 1.432,
      "step": 1645
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.703760567722926e-05,
      "loss": 0.7859,
      "step": 1646
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.70330393202817e-05,
      "loss": 1.1043,
      "step": 1647
    },
    {
      "epoch": 0.68,
      "learning_rate": 9.702846955426737e-05,
      "loss": 0.6456,
      "step": 1648
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.70238963795175e-05,
      "loss": 1.0836,
      "step": 1649
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.701931979636354e-05,
      "loss": 0.3533,
      "step": 1650
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.701473980513727e-05,
      "loss": 1.3638,
      "step": 1651
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.701015640617059e-05,
      "loss": 0.8832,
      "step": 1652
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.700556959979578e-05,
      "loss": 0.8174,
      "step": 1653
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.700097938634524e-05,
      "loss": 1.2738,
      "step": 1654
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.699638576615175e-05,
      "loss": 1.1821,
      "step": 1655
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.69917887395482e-05,
      "loss": 1.2628,
      "step": 1656
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.698718830686783e-05,
      "loss": 1.2545,
      "step": 1657
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.698258446844409e-05,
      "loss": 0.6001,
      "step": 1658
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.697797722461066e-05,
      "loss": 1.1444,
      "step": 1659
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.697336657570148e-05,
      "loss": 0.9161,
      "step": 1660
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.696875252205077e-05,
      "loss": 1.1814,
      "step": 1661
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.696413506399295e-05,
      "loss": 1.4664,
      "step": 1662
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.69595142018627e-05,
      "loss": 1.4405,
      "step": 1663
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.695488993599497e-05,
      "loss": 0.6902,
      "step": 1664
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.695026226672491e-05,
      "loss": 1.4115,
      "step": 1665
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.694563119438796e-05,
      "loss": 1.0056,
      "step": 1666
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.69409967193198e-05,
      "loss": 0.9323,
      "step": 1667
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.693635884185633e-05,
      "loss": 0.6286,
      "step": 1668
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.693171756233373e-05,
      "loss": 1.3471,
      "step": 1669
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.69270728810884e-05,
      "loss": 0.8424,
      "step": 1670
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.6922424798457e-05,
      "loss": 1.1916,
      "step": 1671
    },
    {
      "epoch": 0.69,
      "learning_rate": 9.691777331477644e-05,
      "loss": 1.6538,
      "step": 1672
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.691311843038387e-05,
      "loss": 0.8725,
      "step": 1673
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.690846014561667e-05,
      "loss": 0.9443,
      "step": 1674
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.69037984608125e-05,
      "loss": 1.1968,
      "step": 1675
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.689913337630925e-05,
      "loss": 0.9288,
      "step": 1676
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.689446489244505e-05,
      "loss": 1.2182,
      "step": 1677
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.688979300955829e-05,
      "loss": 1.1751,
      "step": 1678
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.688511772798759e-05,
      "loss": 1.0742,
      "step": 1679
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.688043904807182e-05,
      "loss": 1.1561,
      "step": 1680
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.687575697015013e-05,
      "loss": 1.4642,
      "step": 1681
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.687107149456187e-05,
      "loss": 0.9524,
      "step": 1682
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.686638262164665e-05,
      "loss": 0.7986,
      "step": 1683
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.686169035174434e-05,
      "loss": 0.6981,
      "step": 1684
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.685699468519505e-05,
      "loss": 0.905,
      "step": 1685
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.685229562233911e-05,
      "loss": 1.2062,
      "step": 1686
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.684759316351714e-05,
      "loss": 1.1121,
      "step": 1687
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.684288730906999e-05,
      "loss": 1.5352,
      "step": 1688
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.683817805933873e-05,
      "loss": 0.7926,
      "step": 1689
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.68334654146647e-05,
      "loss": 1.572,
      "step": 1690
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.68287493753895e-05,
      "loss": 0.9168,
      "step": 1691
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.682402994185498e-05,
      "loss": 0.6759,
      "step": 1692
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.681930711440317e-05,
      "loss": 1.2771,
      "step": 1693
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.681458089337641e-05,
      "loss": 0.8684,
      "step": 1694
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.680985127911725e-05,
      "loss": 1.7649,
      "step": 1695
    },
    {
      "epoch": 0.7,
      "learning_rate": 9.680511827196855e-05,
      "loss": 1.0475,
      "step": 1696
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.680038187227333e-05,
      "loss": 1.1321,
      "step": 1697
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.679564208037491e-05,
      "loss": 1.1734,
      "step": 1698
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.679089889661684e-05,
      "loss": 1.2014,
      "step": 1699
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.678615232134292e-05,
      "loss": 1.2033,
      "step": 1700
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.678140235489717e-05,
      "loss": 1.3008,
      "step": 1701
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.677664899762391e-05,
      "loss": 1.2362,
      "step": 1702
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.677189224986767e-05,
      "loss": 0.85,
      "step": 1703
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.676713211197321e-05,
      "loss": 1.5262,
      "step": 1704
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.676236858428558e-05,
      "loss": 0.78,
      "step": 1705
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.675760166715004e-05,
      "loss": 1.2432,
      "step": 1706
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.675283136091212e-05,
      "loss": 0.8194,
      "step": 1707
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.674805766591755e-05,
      "loss": 1.1869,
      "step": 1708
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.674328058251236e-05,
      "loss": 1.005,
      "step": 1709
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.673850011104282e-05,
      "loss": 1.3726,
      "step": 1710
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.673371625185542e-05,
      "loss": 1.3096,
      "step": 1711
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.672892900529687e-05,
      "loss": 1.0661,
      "step": 1712
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.672413837171421e-05,
      "loss": 1.5925,
      "step": 1713
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.671934435145467e-05,
      "loss": 1.0477,
      "step": 1714
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.67145469448657e-05,
      "loss": 1.2067,
      "step": 1715
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.670974615229506e-05,
      "loss": 1.2249,
      "step": 1716
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.67049419740907e-05,
      "loss": 0.4507,
      "step": 1717
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.670013441060086e-05,
      "loss": 1.243,
      "step": 1718
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.669532346217398e-05,
      "loss": 1.5173,
      "step": 1719
    },
    {
      "epoch": 0.71,
      "learning_rate": 9.669050912915879e-05,
      "loss": 1.0932,
      "step": 1720
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.668569141190424e-05,
      "loss": 1.4783,
      "step": 1721
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.668087031075952e-05,
      "loss": 0.7208,
      "step": 1722
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.667604582607408e-05,
      "loss": 0.7164,
      "step": 1723
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.66712179581976e-05,
      "loss": 1.6039,
      "step": 1724
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.666638670748004e-05,
      "loss": 0.9474,
      "step": 1725
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.666155207427157e-05,
      "loss": 0.9585,
      "step": 1726
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.665671405892259e-05,
      "loss": 0.4041,
      "step": 1727
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.665187266178381e-05,
      "loss": 1.2826,
      "step": 1728
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.664702788320612e-05,
      "loss": 1.3249,
      "step": 1729
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.664217972354071e-05,
      "loss": 0.675,
      "step": 1730
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.663732818313894e-05,
      "loss": 1.0454,
      "step": 1731
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.663247326235249e-05,
      "loss": 1.5665,
      "step": 1732
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.662761496153327e-05,
      "loss": 1.3688,
      "step": 1733
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.662275328103341e-05,
      "loss": 0.8588,
      "step": 1734
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.661788822120527e-05,
      "loss": 1.0628,
      "step": 1735
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.661301978240151e-05,
      "loss": 1.0029,
      "step": 1736
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.6608147964975e-05,
      "loss": 0.7738,
      "step": 1737
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.660327276927885e-05,
      "loss": 1.2898,
      "step": 1738
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.659839419566645e-05,
      "loss": 1.3033,
      "step": 1739
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.659351224449137e-05,
      "loss": 0.9674,
      "step": 1740
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.658862691610753e-05,
      "loss": 1.5141,
      "step": 1741
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.658373821086897e-05,
      "loss": 0.9603,
      "step": 1742
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.657884612913004e-05,
      "loss": 1.4845,
      "step": 1743
    },
    {
      "epoch": 0.72,
      "learning_rate": 9.657395067124538e-05,
      "loss": 1.4152,
      "step": 1744
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.656905183756976e-05,
      "loss": 1.1455,
      "step": 1745
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.65641496284583e-05,
      "loss": 1.3269,
      "step": 1746
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.655924404426633e-05,
      "loss": 1.4928,
      "step": 1747
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.655433508534938e-05,
      "loss": 0.6809,
      "step": 1748
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.65494227520633e-05,
      "loss": 1.0084,
      "step": 1749
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.654450704476412e-05,
      "loss": 1.2589,
      "step": 1750
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.653958796380815e-05,
      "loss": 1.4843,
      "step": 1751
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.653466550955196e-05,
      "loss": 0.8525,
      "step": 1752
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.65297396823523e-05,
      "loss": 1.2448,
      "step": 1753
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.652481048256625e-05,
      "loss": 1.4266,
      "step": 1754
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.651987791055106e-05,
      "loss": 1.3585,
      "step": 1755
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.651494196666426e-05,
      "loss": 1.9291,
      "step": 1756
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.651000265126362e-05,
      "loss": 1.2272,
      "step": 1757
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.650505996470716e-05,
      "loss": 1.3988,
      "step": 1758
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.650011390735314e-05,
      "loss": 1.1741,
      "step": 1759
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.649516447956003e-05,
      "loss": 1.2795,
      "step": 1760
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.649021168168664e-05,
      "loss": 1.1641,
      "step": 1761
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.64852555140919e-05,
      "loss": 0.624,
      "step": 1762
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.648029597713506e-05,
      "loss": 0.9927,
      "step": 1763
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.647533307117562e-05,
      "loss": 1.1938,
      "step": 1764
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.647036679657329e-05,
      "loss": 1.0365,
      "step": 1765
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.646539715368803e-05,
      "loss": 0.9092,
      "step": 1766
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.646042414288006e-05,
      "loss": 1.0215,
      "step": 1767
    },
    {
      "epoch": 0.73,
      "learning_rate": 9.645544776450982e-05,
      "loss": 0.6764,
      "step": 1768
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.645046801893804e-05,
      "loss": 1.1478,
      "step": 1769
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.644548490652563e-05,
      "loss": 0.7652,
      "step": 1770
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.644049842763382e-05,
      "loss": 1.0224,
      "step": 1771
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.643550858262401e-05,
      "loss": 0.7186,
      "step": 1772
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.643051537185787e-05,
      "loss": 0.8693,
      "step": 1773
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.642551879569733e-05,
      "loss": 0.8565,
      "step": 1774
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.642051885450458e-05,
      "loss": 1.1532,
      "step": 1775
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.641551554864199e-05,
      "loss": 0.5801,
      "step": 1776
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.641050887847223e-05,
      "loss": 1.0698,
      "step": 1777
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.640549884435818e-05,
      "loss": 0.8496,
      "step": 1778
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.6400485446663e-05,
      "loss": 1.132,
      "step": 1779
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.639546868575006e-05,
      "loss": 0.9053,
      "step": 1780
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.639044856198298e-05,
      "loss": 0.6956,
      "step": 1781
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.638542507572566e-05,
      "loss": 1.1719,
      "step": 1782
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.638039822734218e-05,
      "loss": 0.7463,
      "step": 1783
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.637536801719691e-05,
      "loss": 1.38,
      "step": 1784
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.637033444565448e-05,
      "loss": 1.4399,
      "step": 1785
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.636529751307969e-05,
      "loss": 0.9603,
      "step": 1786
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.636025721983765e-05,
      "loss": 0.5487,
      "step": 1787
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.635521356629369e-05,
      "loss": 0.9938,
      "step": 1788
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.63501665528134e-05,
      "loss": 1.0984,
      "step": 1789
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.634511617976256e-05,
      "loss": 0.7757,
      "step": 1790
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.634006244750728e-05,
      "loss": 1.5615,
      "step": 1791
    },
    {
      "epoch": 0.74,
      "learning_rate": 9.633500535641384e-05,
      "loss": 0.4092,
      "step": 1792
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.632994490684882e-05,
      "loss": 0.5941,
      "step": 1793
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.632488109917895e-05,
      "loss": 1.4374,
      "step": 1794
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.631981393377133e-05,
      "loss": 0.8039,
      "step": 1795
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.631474341099322e-05,
      "loss": 1.5264,
      "step": 1796
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.630966953121212e-05,
      "loss": 1.0283,
      "step": 1797
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.630459229479583e-05,
      "loss": 0.8283,
      "step": 1798
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.629951170211236e-05,
      "loss": 1.665,
      "step": 1799
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.629442775352993e-05,
      "loss": 0.5446,
      "step": 1800
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.628934044941707e-05,
      "loss": 0.9427,
      "step": 1801
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.62842497901425e-05,
      "loss": 1.1176,
      "step": 1802
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.62791557760752e-05,
      "loss": 0.8257,
      "step": 1803
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.627405840758443e-05,
      "loss": 0.6897,
      "step": 1804
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.626895768503962e-05,
      "loss": 0.7332,
      "step": 1805
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.62638536088105e-05,
      "loss": 1.0738,
      "step": 1806
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.625874617926703e-05,
      "loss": 1.5795,
      "step": 1807
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.625363539677941e-05,
      "loss": 1.0341,
      "step": 1808
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.624852126171807e-05,
      "loss": 1.4248,
      "step": 1809
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.624340377445368e-05,
      "loss": 1.3546,
      "step": 1810
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.623828293535721e-05,
      "loss": 1.2014,
      "step": 1811
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.623315874479981e-05,
      "loss": 1.0217,
      "step": 1812
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.62280312031529e-05,
      "loss": 0.8076,
      "step": 1813
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.62229003107881e-05,
      "loss": 1.0762,
      "step": 1814
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.621776606807735e-05,
      "loss": 0.9111,
      "step": 1815
    },
    {
      "epoch": 0.75,
      "learning_rate": 9.621262847539276e-05,
      "loss": 1.7017,
      "step": 1816
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.620748753310675e-05,
      "loss": 0.6666,
      "step": 1817
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.620234324159194e-05,
      "loss": 1.6075,
      "step": 1818
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.619719560122119e-05,
      "loss": 1.6268,
      "step": 1819
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.619204461236761e-05,
      "loss": 1.0813,
      "step": 1820
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.618689027540458e-05,
      "loss": 0.6997,
      "step": 1821
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.618173259070565e-05,
      "loss": 0.9371,
      "step": 1822
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.61765715586447e-05,
      "loss": 1.0575,
      "step": 1823
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.617140717959582e-05,
      "loss": 1.0534,
      "step": 1824
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.61662394539333e-05,
      "loss": 0.8559,
      "step": 1825
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.616106838203176e-05,
      "loss": 1.3803,
      "step": 1826
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.615589396426597e-05,
      "loss": 1.8794,
      "step": 1827
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.615071620101098e-05,
      "loss": 1.4522,
      "step": 1828
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.614553509264212e-05,
      "loss": 1.1575,
      "step": 1829
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.61403506395349e-05,
      "loss": 0.7856,
      "step": 1830
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.613516284206514e-05,
      "loss": 0.957,
      "step": 1831
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.612997170060881e-05,
      "loss": 1.0452,
      "step": 1832
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.612477721554221e-05,
      "loss": 1.1479,
      "step": 1833
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.611957938724186e-05,
      "loss": 1.5978,
      "step": 1834
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.611437821608448e-05,
      "loss": 0.9693,
      "step": 1835
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.610917370244708e-05,
      "loss": 1.4468,
      "step": 1836
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.610396584670689e-05,
      "loss": 1.0284,
      "step": 1837
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.609875464924139e-05,
      "loss": 1.1417,
      "step": 1838
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.609354011042831e-05,
      "loss": 0.7451,
      "step": 1839
    },
    {
      "epoch": 0.76,
      "learning_rate": 9.608832223064559e-05,
      "loss": 0.6956,
      "step": 1840
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.608310101027144e-05,
      "loss": 0.8374,
      "step": 1841
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.607787644968433e-05,
      "loss": 0.6906,
      "step": 1842
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.607264854926293e-05,
      "loss": 0.9185,
      "step": 1843
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.606741730938617e-05,
      "loss": 1.084,
      "step": 1844
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.606218273043323e-05,
      "loss": 1.0437,
      "step": 1845
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.605694481278351e-05,
      "loss": 0.8956,
      "step": 1846
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.60517035568167e-05,
      "loss": 0.6221,
      "step": 1847
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.604645896291265e-05,
      "loss": 1.0974,
      "step": 1848
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.604121103145153e-05,
      "loss": 1.0218,
      "step": 1849
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.603595976281374e-05,
      "loss": 1.1868,
      "step": 1850
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.603070515737986e-05,
      "loss": 1.1089,
      "step": 1851
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.602544721553079e-05,
      "loss": 0.822,
      "step": 1852
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.602018593764764e-05,
      "loss": 1.1104,
      "step": 1853
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.601492132411172e-05,
      "loss": 1.0973,
      "step": 1854
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.600965337530466e-05,
      "loss": 0.6624,
      "step": 1855
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.60043820916083e-05,
      "loss": 0.8249,
      "step": 1856
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.599910747340469e-05,
      "loss": 0.9738,
      "step": 1857
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.599382952107614e-05,
      "loss": 0.982,
      "step": 1858
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.598854823500522e-05,
      "loss": 1.353,
      "step": 1859
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.598326361557476e-05,
      "loss": 1.3119,
      "step": 1860
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.597797566316776e-05,
      "loss": 1.0571,
      "step": 1861
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.597268437816754e-05,
      "loss": 0.6158,
      "step": 1862
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.596738976095758e-05,
      "loss": 1.0768,
      "step": 1863
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.596209181192167e-05,
      "loss": 0.803,
      "step": 1864
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.595679053144383e-05,
      "loss": 1.021,
      "step": 1865
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.59514859199083e-05,
      "loss": 0.8071,
      "step": 1866
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.594617797769956e-05,
      "loss": 1.4832,
      "step": 1867
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.594086670520235e-05,
      "loss": 1.1528,
      "step": 1868
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.593555210280165e-05,
      "loss": 1.5456,
      "step": 1869
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.593023417088267e-05,
      "loss": 1.257,
      "step": 1870
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.592491290983087e-05,
      "loss": 0.35,
      "step": 1871
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.591958832003193e-05,
      "loss": 0.6821,
      "step": 1872
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.591426040187181e-05,
      "loss": 0.843,
      "step": 1873
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.59089291557367e-05,
      "loss": 1.1465,
      "step": 1874
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.590359458201299e-05,
      "loss": 0.9156,
      "step": 1875
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.589825668108736e-05,
      "loss": 0.9676,
      "step": 1876
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.589291545334671e-05,
      "loss": 0.7788,
      "step": 1877
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.588757089917818e-05,
      "loss": 1.3762,
      "step": 1878
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.588222301896917e-05,
      "loss": 1.0425,
      "step": 1879
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.587687181310729e-05,
      "loss": 1.0585,
      "step": 1880
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.587151728198042e-05,
      "loss": 0.7762,
      "step": 1881
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.586615942597667e-05,
      "loss": 0.968,
      "step": 1882
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.58607982454844e-05,
      "loss": 1.1656,
      "step": 1883
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.585543374089217e-05,
      "loss": 1.5376,
      "step": 1884
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.585006591258882e-05,
      "loss": 0.8168,
      "step": 1885
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.584469476096345e-05,
      "loss": 1.2225,
      "step": 1886
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.583932028640536e-05,
      "loss": 0.8732,
      "step": 1887
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.583394248930409e-05,
      "loss": 1.0735,
      "step": 1888
    },
    {
      "epoch": 0.78,
      "learning_rate": 9.582856137004945e-05,
      "loss": 1.1285,
      "step": 1889
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.582317692903148e-05,
      "loss": 1.0537,
      "step": 1890
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.581778916664044e-05,
      "loss": 1.1498,
      "step": 1891
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.581239808326687e-05,
      "loss": 1.1563,
      "step": 1892
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.58070036793015e-05,
      "loss": 0.6628,
      "step": 1893
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.580160595513537e-05,
      "loss": 1.0568,
      "step": 1894
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.579620491115968e-05,
      "loss": 1.721,
      "step": 1895
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.579080054776593e-05,
      "loss": 0.6237,
      "step": 1896
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.578539286534582e-05,
      "loss": 0.4592,
      "step": 1897
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.577998186429136e-05,
      "loss": 0.5776,
      "step": 1898
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.577456754499469e-05,
      "loss": 0.8446,
      "step": 1899
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.576914990784831e-05,
      "loss": 1.2723,
      "step": 1900
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.576372895324486e-05,
      "loss": 0.5175,
      "step": 1901
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.575830468157729e-05,
      "loss": 0.898,
      "step": 1902
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.575287709323876e-05,
      "loss": 0.6725,
      "step": 1903
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.574744618862267e-05,
      "loss": 0.9798,
      "step": 1904
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.574201196812265e-05,
      "loss": 0.7188,
      "step": 1905
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.573657443213262e-05,
      "loss": 1.1674,
      "step": 1906
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.573113358104668e-05,
      "loss": 1.2478,
      "step": 1907
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.572568941525918e-05,
      "loss": 1.308,
      "step": 1908
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.572024193516478e-05,
      "loss": 0.745,
      "step": 1909
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.571479114115828e-05,
      "loss": 1.1278,
      "step": 1910
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.570933703363477e-05,
      "loss": 1.4007,
      "step": 1911
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.57038796129896e-05,
      "loss": 1.0424,
      "step": 1912
    },
    {
      "epoch": 0.79,
      "learning_rate": 9.569841887961832e-05,
      "loss": 1.4786,
      "step": 1913
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.569295483391673e-05,
      "loss": 0.7018,
      "step": 1914
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.56874874762809e-05,
      "loss": 1.2881,
      "step": 1915
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.568201680710711e-05,
      "loss": 0.9292,
      "step": 1916
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.567654282679188e-05,
      "loss": 1.0761,
      "step": 1917
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.567106553573199e-05,
      "loss": 1.0584,
      "step": 1918
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.566558493432442e-05,
      "loss": 0.37,
      "step": 1919
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.566010102296643e-05,
      "loss": 1.2316,
      "step": 1920
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.565461380205552e-05,
      "loss": 1.406,
      "step": 1921
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.564912327198941e-05,
      "loss": 0.7409,
      "step": 1922
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.564362943316606e-05,
      "loss": 0.9597,
      "step": 1923
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.563813228598369e-05,
      "loss": 0.6959,
      "step": 1924
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.563263183084071e-05,
      "loss": 1.5125,
      "step": 1925
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.562712806813585e-05,
      "loss": 0.3795,
      "step": 1926
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.562162099826802e-05,
      "loss": 0.7085,
      "step": 1927
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.561611062163639e-05,
      "loss": 0.8804,
      "step": 1928
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.561059693864035e-05,
      "loss": 0.3957,
      "step": 1929
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.560507994967956e-05,
      "loss": 1.1713,
      "step": 1930
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.55995596551539e-05,
      "loss": 0.7422,
      "step": 1931
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.559403605546349e-05,
      "loss": 1.3043,
      "step": 1932
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.55885091510087e-05,
      "loss": 0.7886,
      "step": 1933
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.558297894219012e-05,
      "loss": 1.4348,
      "step": 1934
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.55774454294086e-05,
      "loss": 0.6592,
      "step": 1935
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.557190861306524e-05,
      "loss": 0.9429,
      "step": 1936
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.556636849356135e-05,
      "loss": 1.3102,
      "step": 1937
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.556082507129849e-05,
      "loss": 1.0279,
      "step": 1938
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.555527834667843e-05,
      "loss": 0.9167,
      "step": 1939
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.554972832010327e-05,
      "loss": 0.7746,
      "step": 1940
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.554417499197523e-05,
      "loss": 0.8019,
      "step": 1941
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.553861836269687e-05,
      "loss": 1.3664,
      "step": 1942
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.553305843267093e-05,
      "loss": 0.9441,
      "step": 1943
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.552749520230043e-05,
      "loss": 1.6481,
      "step": 1944
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.552192867198856e-05,
      "loss": 1.0434,
      "step": 1945
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.551635884213884e-05,
      "loss": 0.6164,
      "step": 1946
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.551078571315495e-05,
      "loss": 0.7683,
      "step": 1947
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.550520928544087e-05,
      "loss": 1.6769,
      "step": 1948
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.549962955940077e-05,
      "loss": 1.0245,
      "step": 1949
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.549404653543911e-05,
      "loss": 1.5692,
      "step": 1950
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.548846021396055e-05,
      "loss": 1.0055,
      "step": 1951
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.548287059536998e-05,
      "loss": 0.7293,
      "step": 1952
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.547727768007256e-05,
      "loss": 1.1493,
      "step": 1953
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.54716814684737e-05,
      "loss": 1.1368,
      "step": 1954
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.546608196097899e-05,
      "loss": 1.1629,
      "step": 1955
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.546047915799433e-05,
      "loss": 1.3241,
      "step": 1956
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.545487305992581e-05,
      "loss": 1.1299,
      "step": 1957
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.544926366717976e-05,
      "loss": 0.628,
      "step": 1958
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.544365098016279e-05,
      "loss": 1.2361,
      "step": 1959
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.54380349992817e-05,
      "loss": 1.0835,
      "step": 1960
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.543241572494354e-05,
      "loss": 0.7481,
      "step": 1961
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.542679315755564e-05,
      "loss": 1.3029,
      "step": 1962
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.542116729752553e-05,
      "loss": 1.3694,
      "step": 1963
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.541553814526096e-05,
      "loss": 1.1149,
      "step": 1964
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.540990570116997e-05,
      "loss": 0.8877,
      "step": 1965
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.540426996566078e-05,
      "loss": 0.8332,
      "step": 1966
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.539863093914193e-05,
      "loss": 1.0722,
      "step": 1967
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.539298862202212e-05,
      "loss": 1.205,
      "step": 1968
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.538734301471032e-05,
      "loss": 0.7669,
      "step": 1969
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.538169411761575e-05,
      "loss": 0.7405,
      "step": 1970
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.537604193114783e-05,
      "loss": 1.2179,
      "step": 1971
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.537038645571628e-05,
      "loss": 0.7863,
      "step": 1972
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.536472769173099e-05,
      "loss": 1.272,
      "step": 1973
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.535906563960213e-05,
      "loss": 1.2952,
      "step": 1974
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.535340029974011e-05,
      "loss": 1.4968,
      "step": 1975
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.534773167255556e-05,
      "loss": 0.7791,
      "step": 1976
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.534205975845935e-05,
      "loss": 0.9366,
      "step": 1977
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.533638455786261e-05,
      "loss": 0.8795,
      "step": 1978
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.533070607117667e-05,
      "loss": 0.9362,
      "step": 1979
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.532502429881314e-05,
      "loss": 0.9871,
      "step": 1980
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.531933924118384e-05,
      "loss": 0.8744,
      "step": 1981
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.531365089870084e-05,
      "loss": 0.5849,
      "step": 1982
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.530795927177642e-05,
      "loss": 1.0579,
      "step": 1983
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.530226436082317e-05,
      "loss": 1.1794,
      "step": 1984
    },
    {
      "epoch": 0.82,
      "learning_rate": 9.529656616625383e-05,
      "loss": 1.3668,
      "step": 1985
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.529086468848144e-05,
      "loss": 1.2036,
      "step": 1986
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.528515992791925e-05,
      "loss": 1.4691,
      "step": 1987
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.527945188498075e-05,
      "loss": 1.0694,
      "step": 1988
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.527374056007967e-05,
      "loss": 1.2621,
      "step": 1989
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.526802595363e-05,
      "loss": 0.6036,
      "step": 1990
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.526230806604595e-05,
      "loss": 0.5078,
      "step": 1991
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.525658689774192e-05,
      "loss": 1.0411,
      "step": 1992
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.525086244913264e-05,
      "loss": 0.8468,
      "step": 1993
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.524513472063302e-05,
      "loss": 1.4252,
      "step": 1994
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.523940371265822e-05,
      "loss": 0.8959,
      "step": 1995
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.523366942562363e-05,
      "loss": 1.0602,
      "step": 1996
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.522793185994488e-05,
      "loss": 1.7066,
      "step": 1997
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.522219101603787e-05,
      "loss": 1.4114,
      "step": 1998
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.521644689431867e-05,
      "loss": 0.8197,
      "step": 1999
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.521069949520367e-05,
      "loss": 1.3888,
      "step": 2000
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.520494881910941e-05,
      "loss": 1.1457,
      "step": 2001
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.519919486645276e-05,
      "loss": 0.4047,
      "step": 2002
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.519343763765075e-05,
      "loss": 0.9803,
      "step": 2003
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.518767713312069e-05,
      "loss": 0.8227,
      "step": 2004
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.51819133532801e-05,
      "loss": 1.0023,
      "step": 2005
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.517614629854676e-05,
      "loss": 1.0858,
      "step": 2006
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.517037596933868e-05,
      "loss": 1.1071,
      "step": 2007
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.516460236607413e-05,
      "loss": 0.8351,
      "step": 2008
    },
    {
      "epoch": 0.83,
      "learning_rate": 9.515882548917157e-05,
      "loss": 0.9167,
      "step": 2009
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.515304533904971e-05,
      "loss": 0.4189,
      "step": 2010
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.514726191612753e-05,
      "loss": 0.9034,
      "step": 2011
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.514147522082423e-05,
      "loss": 1.2324,
      "step": 2012
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.513568525355922e-05,
      "loss": 0.7966,
      "step": 2013
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.51298920147522e-05,
      "loss": 1.2565,
      "step": 2014
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.512409550482306e-05,
      "loss": 0.8473,
      "step": 2015
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.511829572419194e-05,
      "loss": 1.3877,
      "step": 2016
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.511249267327924e-05,
      "loss": 0.9193,
      "step": 2017
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.510668635250556e-05,
      "loss": 0.8626,
      "step": 2018
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.510087676229176e-05,
      "loss": 1.0697,
      "step": 2019
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.509506390305896e-05,
      "loss": 0.6231,
      "step": 2020
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.508924777522845e-05,
      "loss": 0.8508,
      "step": 2021
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.508342837922182e-05,
      "loss": 1.1706,
      "step": 2022
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.507760571546086e-05,
      "loss": 0.9009,
      "step": 2023
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.507177978436762e-05,
      "loss": 0.7576,
      "step": 2024
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.506595058636438e-05,
      "loss": 0.5227,
      "step": 2025
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.506011812187365e-05,
      "loss": 1.3732,
      "step": 2026
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.505428239131818e-05,
      "loss": 1.1463,
      "step": 2027
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.504844339512095e-05,
      "loss": 1.4318,
      "step": 2028
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.504260113370522e-05,
      "loss": 0.6235,
      "step": 2029
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.50367556074944e-05,
      "loss": 0.6181,
      "step": 2030
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.503090681691223e-05,
      "loss": 1.1932,
      "step": 2031
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.502505476238261e-05,
      "loss": 0.9491,
      "step": 2032
    },
    {
      "epoch": 0.84,
      "learning_rate": 9.501919944432974e-05,
      "loss": 0.9702,
      "step": 2033
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.5013340863178e-05,
      "loss": 1.1196,
      "step": 2034
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.500747901935205e-05,
      "loss": 0.6633,
      "step": 2035
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.500161391327678e-05,
      "loss": 0.9382,
      "step": 2036
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.499574554537727e-05,
      "loss": 0.9099,
      "step": 2037
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.498987391607892e-05,
      "loss": 0.6194,
      "step": 2038
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.49839990258073e-05,
      "loss": 1.1644,
      "step": 2039
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.497812087498823e-05,
      "loss": 0.8653,
      "step": 2040
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.497223946404778e-05,
      "loss": 0.9331,
      "step": 2041
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.496635479341223e-05,
      "loss": 1.3053,
      "step": 2042
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.496046686350815e-05,
      "loss": 1.4485,
      "step": 2043
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.49545756747623e-05,
      "loss": 0.5477,
      "step": 2044
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.494868122760166e-05,
      "loss": 0.5795,
      "step": 2045
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.494278352245349e-05,
      "loss": 1.0718,
      "step": 2046
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.493688255974527e-05,
      "loss": 0.9719,
      "step": 2047
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.493097833990474e-05,
      "loss": 0.8322,
      "step": 2048
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.492507086335981e-05,
      "loss": 0.7493,
      "step": 2049
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.49191601305387e-05,
      "loss": 1.1695,
      "step": 2050
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.49132461418698e-05,
      "loss": 1.1433,
      "step": 2051
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.490732889778182e-05,
      "loss": 1.2467,
      "step": 2052
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.490140839870363e-05,
      "loss": 1.5114,
      "step": 2053
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.489548464506433e-05,
      "loss": 1.9416,
      "step": 2054
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.488955763729335e-05,
      "loss": 0.631,
      "step": 2055
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.488362737582023e-05,
      "loss": 0.5063,
      "step": 2056
    },
    {
      "epoch": 0.85,
      "learning_rate": 9.487769386107487e-05,
      "loss": 1.2681,
      "step": 2057
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.48717570934873e-05,
      "loss": 0.4865,
      "step": 2058
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.486581707348786e-05,
      "loss": 0.4411,
      "step": 2059
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.485987380150707e-05,
      "loss": 1.3301,
      "step": 2060
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.485392727797573e-05,
      "loss": 0.8876,
      "step": 2061
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.484797750332486e-05,
      "loss": 1.2812,
      "step": 2062
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.48420244779857e-05,
      "loss": 1.5538,
      "step": 2063
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.483606820238975e-05,
      "loss": 1.2642,
      "step": 2064
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.483010867696875e-05,
      "loss": 1.0062,
      "step": 2065
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.482414590215463e-05,
      "loss": 1.3071,
      "step": 2066
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.481817987837959e-05,
      "loss": 1.3616,
      "step": 2067
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.481221060607608e-05,
      "loss": 1.1983,
      "step": 2068
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.480623808567677e-05,
      "loss": 0.8944,
      "step": 2069
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.480026231761454e-05,
      "loss": 0.8134,
      "step": 2070
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.479428330232253e-05,
      "loss": 0.3604,
      "step": 2071
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.478830104023412e-05,
      "loss": 0.9354,
      "step": 2072
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.478231553178295e-05,
      "loss": 0.6725,
      "step": 2073
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.477632677740283e-05,
      "loss": 1.0761,
      "step": 2074
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.477033477752782e-05,
      "loss": 0.676,
      "step": 2075
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.476433953259227e-05,
      "loss": 0.5597,
      "step": 2076
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.475834104303072e-05,
      "loss": 1.0501,
      "step": 2077
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.475233930927794e-05,
      "loss": 1.2029,
      "step": 2078
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.474633433176898e-05,
      "loss": 0.838,
      "step": 2079
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.474032611093908e-05,
      "loss": 1.0744,
      "step": 2080
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.473431464722371e-05,
      "loss": 1.3041,
      "step": 2081
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.472829994105864e-05,
      "loss": 0.9771,
      "step": 2082
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.472228199287978e-05,
      "loss": 1.0197,
      "step": 2083
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.471626080312335e-05,
      "loss": 1.2619,
      "step": 2084
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.471023637222578e-05,
      "loss": 1.0499,
      "step": 2085
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.470420870062374e-05,
      "loss": 0.6902,
      "step": 2086
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.469817778875412e-05,
      "loss": 1.0607,
      "step": 2087
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.469214363705407e-05,
      "loss": 1.1159,
      "step": 2088
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.468610624596093e-05,
      "loss": 0.8966,
      "step": 2089
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.468006561591234e-05,
      "loss": 0.6681,
      "step": 2090
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.467402174734612e-05,
      "loss": 0.9858,
      "step": 2091
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.466797464070035e-05,
      "loss": 0.7509,
      "step": 2092
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.466192429641331e-05,
      "loss": 0.8205,
      "step": 2093
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.465587071492359e-05,
      "loss": 0.4885,
      "step": 2094
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.464981389666993e-05,
      "loss": 0.7805,
      "step": 2095
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.464375384209138e-05,
      "loss": 1.0646,
      "step": 2096
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.463769055162715e-05,
      "loss": 0.9179,
      "step": 2097
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.463162402571674e-05,
      "loss": 0.8832,
      "step": 2098
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.462555426479988e-05,
      "loss": 1.0566,
      "step": 2099
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.46194812693165e-05,
      "loss": 0.6927,
      "step": 2100
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.461340503970677e-05,
      "loss": 0.7042,
      "step": 2101
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.460732557641114e-05,
      "loss": 1.0475,
      "step": 2102
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.460124287987027e-05,
      "loss": 1.2277,
      "step": 2103
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.459515695052502e-05,
      "loss": 1.4783,
      "step": 2104
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.458906778881651e-05,
      "loss": 1.145,
      "step": 2105
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.458297539518613e-05,
      "loss": 1.4768,
      "step": 2106
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.457687977007546e-05,
      "loss": 1.4075,
      "step": 2107
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.457078091392631e-05,
      "loss": 1.2592,
      "step": 2108
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.456467882718073e-05,
      "loss": 1.1023,
      "step": 2109
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.455857351028105e-05,
      "loss": 1.9609,
      "step": 2110
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.455246496366978e-05,
      "loss": 1.1271,
      "step": 2111
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.454635318778967e-05,
      "loss": 0.8907,
      "step": 2112
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.454023818308374e-05,
      "loss": 0.9429,
      "step": 2113
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.453411994999521e-05,
      "loss": 1.3862,
      "step": 2114
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.452799848896753e-05,
      "loss": 1.1359,
      "step": 2115
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.452187380044442e-05,
      "loss": 1.5964,
      "step": 2116
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.45157458848698e-05,
      "loss": 1.3131,
      "step": 2117
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.450961474268782e-05,
      "loss": 1.3633,
      "step": 2118
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.450348037434291e-05,
      "loss": 1.3188,
      "step": 2119
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.44973427802797e-05,
      "loss": 0.8223,
      "step": 2120
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.449120196094306e-05,
      "loss": 1.1121,
      "step": 2121
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.448505791677805e-05,
      "loss": 0.6351,
      "step": 2122
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.447891064823004e-05,
      "loss": 1.6962,
      "step": 2123
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.447276015574462e-05,
      "loss": 0.7619,
      "step": 2124
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.446660643976754e-05,
      "loss": 0.921,
      "step": 2125
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.446044950074488e-05,
      "loss": 1.1938,
      "step": 2126
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.445428933912288e-05,
      "loss": 0.9262,
      "step": 2127
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.444812595534805e-05,
      "loss": 0.7771,
      "step": 2128
    },
    {
      "epoch": 0.88,
      "learning_rate": 9.444195934986715e-05,
      "loss": 0.7976,
      "step": 2129
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.443578952312712e-05,
      "loss": 1.0976,
      "step": 2130
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.442961647557518e-05,
      "loss": 0.7644,
      "step": 2131
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.442344020765875e-05,
      "loss": 0.8304,
      "step": 2132
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.441726071982553e-05,
      "loss": 0.9891,
      "step": 2133
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.44110780125234e-05,
      "loss": 0.5841,
      "step": 2134
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.44048920862005e-05,
      "loss": 1.0251,
      "step": 2135
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.439870294130522e-05,
      "loss": 1.0867,
      "step": 2136
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.439251057828613e-05,
      "loss": 1.6218,
      "step": 2137
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.43863149975921e-05,
      "loss": 0.5606,
      "step": 2138
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.438011619967217e-05,
      "loss": 0.5343,
      "step": 2139
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.437391418497568e-05,
      "loss": 1.1781,
      "step": 2140
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.436770895395214e-05,
      "loss": 0.7914,
      "step": 2141
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.436150050705133e-05,
      "loss": 0.7467,
      "step": 2142
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.435528884472327e-05,
      "loss": 0.9602,
      "step": 2143
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.434907396741814e-05,
      "loss": 0.6448,
      "step": 2144
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.434285587558647e-05,
      "loss": 0.8455,
      "step": 2145
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.433663456967893e-05,
      "loss": 1.5747,
      "step": 2146
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.433041005014647e-05,
      "loss": 0.9029,
      "step": 2147
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.432418231744025e-05,
      "loss": 1.3249,
      "step": 2148
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.431795137201165e-05,
      "loss": 1.04,
      "step": 2149
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.431171721431234e-05,
      "loss": 1.4959,
      "step": 2150
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.430547984479419e-05,
      "loss": 0.6432,
      "step": 2151
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.429923926390926e-05,
      "loss": 1.5168,
      "step": 2152
    },
    {
      "epoch": 0.89,
      "learning_rate": 9.42929954721099e-05,
      "loss": 1.2631,
      "step": 2153
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.428674846984869e-05,
      "loss": 1.7973,
      "step": 2154
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.428049825757842e-05,
      "loss": 1.4569,
      "step": 2155
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.427424483575211e-05,
      "loss": 0.9821,
      "step": 2156
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.426798820482304e-05,
      "loss": 1.1863,
      "step": 2157
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.426172836524468e-05,
      "loss": 1.1091,
      "step": 2158
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.425546531747078e-05,
      "loss": 1.1254,
      "step": 2159
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.424919906195529e-05,
      "loss": 1.1969,
      "step": 2160
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.424292959915241e-05,
      "loss": 1.0204,
      "step": 2161
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.423665692951656e-05,
      "loss": 1.3046,
      "step": 2162
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.423038105350242e-05,
      "loss": 0.6702,
      "step": 2163
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.422410197156484e-05,
      "loss": 0.9876,
      "step": 2164
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.421781968415896e-05,
      "loss": 1.0101,
      "step": 2165
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.421153419174014e-05,
      "loss": 0.8415,
      "step": 2166
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.420524549476399e-05,
      "loss": 0.579,
      "step": 2167
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.419895359368628e-05,
      "loss": 0.824,
      "step": 2168
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.41926584889631e-05,
      "loss": 1.0388,
      "step": 2169
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.418636018105071e-05,
      "loss": 0.6752,
      "step": 2170
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.418005867040566e-05,
      "loss": 1.1693,
      "step": 2171
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.417375395748467e-05,
      "loss": 1.2726,
      "step": 2172
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.416744604274473e-05,
      "loss": 1.5525,
      "step": 2173
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.416113492664304e-05,
      "loss": 0.8587,
      "step": 2174
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.415482060963705e-05,
      "loss": 1.1425,
      "step": 2175
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.414850309218446e-05,
      "loss": 1.4192,
      "step": 2176
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.414218237474315e-05,
      "loss": 0.8983,
      "step": 2177
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.413585845777126e-05,
      "loss": 0.7101,
      "step": 2178
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.41295313417272e-05,
      "loss": 1.1038,
      "step": 2179
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.412320102706954e-05,
      "loss": 0.7056,
      "step": 2180
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.411686751425712e-05,
      "loss": 1.1101,
      "step": 2181
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.411053080374901e-05,
      "loss": 1.1344,
      "step": 2182
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.410419089600452e-05,
      "loss": 0.8268,
      "step": 2183
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.409784779148317e-05,
      "loss": 0.9359,
      "step": 2184
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.409150149064472e-05,
      "loss": 1.0235,
      "step": 2185
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.408515199394918e-05,
      "loss": 1.1234,
      "step": 2186
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.407879930185677e-05,
      "loss": 0.8542,
      "step": 2187
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.407244341482794e-05,
      "loss": 1.3409,
      "step": 2188
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.406608433332341e-05,
      "loss": 1.2164,
      "step": 2189
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.405972205780407e-05,
      "loss": 1.1709,
      "step": 2190
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.405335658873107e-05,
      "loss": 0.9656,
      "step": 2191
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.40469879265658e-05,
      "loss": 0.8562,
      "step": 2192
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.404061607176991e-05,
      "loss": 0.9451,
      "step": 2193
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.40342410248052e-05,
      "loss": 1.1004,
      "step": 2194
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.402786278613378e-05,
      "loss": 0.8408,
      "step": 2195
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.402148135621793e-05,
      "loss": 0.8077,
      "step": 2196
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.401509673552023e-05,
      "loss": 1.0264,
      "step": 2197
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.400870892450341e-05,
      "loss": 1.4377,
      "step": 2198
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.40023179236305e-05,
      "loss": 0.8322,
      "step": 2199
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.399592373336474e-05,
      "loss": 1.3173,
      "step": 2200
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.398952635416957e-05,
      "loss": 1.3253,
      "step": 2201
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.398312578650872e-05,
      "loss": 1.0334,
      "step": 2202
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.397672203084609e-05,
      "loss": 1.2013,
      "step": 2203
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.397031508764587e-05,
      "loss": 0.6272,
      "step": 2204
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.396390495737242e-05,
      "loss": 0.86,
      "step": 2205
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.395749164049037e-05,
      "loss": 0.532,
      "step": 2206
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.395107513746458e-05,
      "loss": 1.3953,
      "step": 2207
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.394465544876012e-05,
      "loss": 1.1964,
      "step": 2208
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.393823257484233e-05,
      "loss": 0.732,
      "step": 2209
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.393180651617673e-05,
      "loss": 1.0523,
      "step": 2210
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.39253772732291e-05,
      "loss": 0.9847,
      "step": 2211
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.391894484646546e-05,
      "loss": 0.8166,
      "step": 2212
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.391250923635204e-05,
      "loss": 0.8822,
      "step": 2213
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.39060704433553e-05,
      "loss": 1.124,
      "step": 2214
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.389962846794194e-05,
      "loss": 0.8293,
      "step": 2215
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.389318331057892e-05,
      "loss": 0.9944,
      "step": 2216
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.388673497173336e-05,
      "loss": 0.5481,
      "step": 2217
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.388028345187268e-05,
      "loss": 0.9439,
      "step": 2218
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.387382875146448e-05,
      "loss": 1.0929,
      "step": 2219
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.386737087097662e-05,
      "loss": 0.7651,
      "step": 2220
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.386090981087719e-05,
      "loss": 1.2196,
      "step": 2221
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.385444557163449e-05,
      "loss": 1.0389,
      "step": 2222
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.384797815371709e-05,
      "loss": 0.6253,
      "step": 2223
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.384150755759372e-05,
      "loss": 0.7906,
      "step": 2224
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.383503378373342e-05,
      "loss": 0.9886,
      "step": 2225
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.382855683260541e-05,
      "loss": 0.6595,
      "step": 2226
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.382207670467915e-05,
      "loss": 1.1758,
      "step": 2227
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.381559340042436e-05,
      "loss": 0.5329,
      "step": 2228
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.380910692031095e-05,
      "loss": 0.8857,
      "step": 2229
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.380261726480906e-05,
      "loss": 0.6842,
      "step": 2230
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.37961244343891e-05,
      "loss": 1.0888,
      "step": 2231
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.378962842952167e-05,
      "loss": 1.1395,
      "step": 2232
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.378312925067763e-05,
      "loss": 1.3472,
      "step": 2233
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.377662689832803e-05,
      "loss": 0.865,
      "step": 2234
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.377012137294421e-05,
      "loss": 1.4253,
      "step": 2235
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.376361267499768e-05,
      "loss": 0.6841,
      "step": 2236
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.375710080496022e-05,
      "loss": 0.518,
      "step": 2237
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.375058576330382e-05,
      "loss": 1.4962,
      "step": 2238
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.37440675505007e-05,
      "loss": 0.9157,
      "step": 2239
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.373754616702332e-05,
      "loss": 1.417,
      "step": 2240
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.373102161334439e-05,
      "loss": 0.4498,
      "step": 2241
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.372449388993679e-05,
      "loss": 1.0191,
      "step": 2242
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.371796299727368e-05,
      "loss": 1.4882,
      "step": 2243
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.371142893582842e-05,
      "loss": 1.2569,
      "step": 2244
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.370489170607464e-05,
      "loss": 0.7704,
      "step": 2245
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.369835130848616e-05,
      "loss": 1.1144,
      "step": 2246
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.369180774353705e-05,
      "loss": 1.2385,
      "step": 2247
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.368526101170159e-05,
      "loss": 0.8404,
      "step": 2248
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.367871111345432e-05,
      "loss": 0.6929,
      "step": 2249
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.367215804926998e-05,
      "loss": 0.6802,
      "step": 2250
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.366560181962353e-05,
      "loss": 0.9203,
      "step": 2251
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.365904242499024e-05,
      "loss": 0.6575,
      "step": 2252
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.36524798658455e-05,
      "loss": 0.6635,
      "step": 2253
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.3645914142665e-05,
      "loss": 0.6525,
      "step": 2254
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.363934525592464e-05,
      "loss": 1.0346,
      "step": 2255
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.363277320610053e-05,
      "loss": 0.7504,
      "step": 2256
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.362619799366905e-05,
      "loss": 1.5736,
      "step": 2257
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.36196196191068e-05,
      "loss": 1.1861,
      "step": 2258
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.361303808289053e-05,
      "loss": 1.0603,
      "step": 2259
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.360645338549736e-05,
      "loss": 1.0116,
      "step": 2260
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.359986552740453e-05,
      "loss": 1.6359,
      "step": 2261
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.359327450908954e-05,
      "loss": 0.9171,
      "step": 2262
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.358668033103015e-05,
      "loss": 1.3575,
      "step": 2263
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.358008299370428e-05,
      "loss": 1.2917,
      "step": 2264
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.357348249759017e-05,
      "loss": 0.9911,
      "step": 2265
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.35668788431662e-05,
      "loss": 1.0621,
      "step": 2266
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.356027203091104e-05,
      "loss": 0.9634,
      "step": 2267
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.355366206130354e-05,
      "loss": 1.0932,
      "step": 2268
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.354704893482285e-05,
      "loss": 0.9366,
      "step": 2269
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.354043265194827e-05,
      "loss": 1.0733,
      "step": 2270
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.353381321315937e-05,
      "loss": 0.3766,
      "step": 2271
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.352719061893596e-05,
      "loss": 0.8584,
      "step": 2272
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.352056486975804e-05,
      "loss": 0.8898,
      "step": 2273
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.351393596610587e-05,
      "loss": 0.6569,
      "step": 2274
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.350730390845995e-05,
      "loss": 1.433,
      "step": 2275
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.350066869730094e-05,
      "loss": 1.7089,
      "step": 2276
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.34940303331098e-05,
      "loss": 0.654,
      "step": 2277
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.348738881636771e-05,
      "loss": 0.6648,
      "step": 2278
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.348074414755604e-05,
      "loss": 1.5465,
      "step": 2279
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.347409632715642e-05,
      "loss": 0.8412,
      "step": 2280
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.34674453556507e-05,
      "loss": 1.0155,
      "step": 2281
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.346079123352097e-05,
      "loss": 1.0684,
      "step": 2282
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.345413396124951e-05,
      "loss": 0.8827,
      "step": 2283
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.344747353931888e-05,
      "loss": 0.6004,
      "step": 2284
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.344080996821183e-05,
      "loss": 1.2571,
      "step": 2285
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.343414324841136e-05,
      "loss": 1.13,
      "step": 2286
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.342747338040067e-05,
      "loss": 1.0933,
      "step": 2287
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.342080036466324e-05,
      "loss": 1.4767,
      "step": 2288
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.341412420168272e-05,
      "loss": 0.7353,
      "step": 2289
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.340744489194302e-05,
      "loss": 1.2481,
      "step": 2290
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.340076243592828e-05,
      "loss": 1.1833,
      "step": 2291
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.339407683412287e-05,
      "loss": 0.7046,
      "step": 2292
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.338738808701134e-05,
      "loss": 1.1952,
      "step": 2293
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.338069619507855e-05,
      "loss": 0.9685,
      "step": 2294
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.33740011588095e-05,
      "loss": 0.6216,
      "step": 2295
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.33673029786895e-05,
      "loss": 0.8425,
      "step": 2296
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.336060165520404e-05,
      "loss": 1.5874,
      "step": 2297
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.335389718883883e-05,
      "loss": 0.8419,
      "step": 2298
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.334718958007985e-05,
      "loss": 1.2715,
      "step": 2299
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.334047882941328e-05,
      "loss": 0.6213,
      "step": 2300
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.333376493732551e-05,
      "loss": 1.4953,
      "step": 2301
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.33270479043032e-05,
      "loss": 1.4029,
      "step": 2302
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.332032773083322e-05,
      "loss": 0.7813,
      "step": 2303
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.331360441740265e-05,
      "loss": 0.6154,
      "step": 2304
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.330687796449882e-05,
      "loss": 0.6666,
      "step": 2305
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.330014837260925e-05,
      "loss": 0.6293,
      "step": 2306
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.329341564222176e-05,
      "loss": 0.5183,
      "step": 2307
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.328667977382435e-05,
      "loss": 0.9864,
      "step": 2308
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.327994076790523e-05,
      "loss": 0.652,
      "step": 2309
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.327319862495286e-05,
      "loss": 0.8806,
      "step": 2310
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.326645334545594e-05,
      "loss": 0.9219,
      "step": 2311
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.325970492990337e-05,
      "loss": 0.7789,
      "step": 2312
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.32529533787843e-05,
      "loss": 0.973,
      "step": 2313
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.324619869258812e-05,
      "loss": 1.1699,
      "step": 2314
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.323944087180439e-05,
      "loss": 0.5326,
      "step": 2315
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.323267991692293e-05,
      "loss": 1.0445,
      "step": 2316
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.322591582843383e-05,
      "loss": 0.753,
      "step": 2317
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.321914860682734e-05,
      "loss": 1.49,
      "step": 2318
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.321237825259396e-05,
      "loss": 1.2909,
      "step": 2319
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.320560476622444e-05,
      "loss": 1.1843,
      "step": 2320
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.319882814820971e-05,
      "loss": 1.2391,
      "step": 2321
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.319204839904099e-05,
      "loss": 1.4893,
      "step": 2322
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.318526551920967e-05,
      "loss": 0.9819,
      "step": 2323
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.31784795092074e-05,
      "loss": 0.7467,
      "step": 2324
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.317169036952604e-05,
      "loss": 1.0592,
      "step": 2325
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.316489810065769e-05,
      "loss": 0.6735,
      "step": 2326
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.315810270309466e-05,
      "loss": 1.0611,
      "step": 2327
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.315130417732949e-05,
      "loss": 0.9364,
      "step": 2328
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.314450252385498e-05,
      "loss": 0.6736,
      "step": 2329
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.313769774316411e-05,
      "loss": 0.6807,
      "step": 2330
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.313088983575011e-05,
      "loss": 0.6364,
      "step": 2331
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.312407880210643e-05,
      "loss": 1.0478,
      "step": 2332
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.311726464272677e-05,
      "loss": 1.5332,
      "step": 2333
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.3110447358105e-05,
      "loss": 0.5967,
      "step": 2334
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.310362694873528e-05,
      "loss": 1.2524,
      "step": 2335
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.309680341511198e-05,
      "loss": 1.4138,
      "step": 2336
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.308997675772966e-05,
      "loss": 0.9069,
      "step": 2337
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.308314697708314e-05,
      "loss": 1.0634,
      "step": 2338
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.307631407366747e-05,
      "loss": 0.924,
      "step": 2339
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.306947804797787e-05,
      "loss": 1.4142,
      "step": 2340
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.30626389005099e-05,
      "loss": 1.5018,
      "step": 2341
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.305579663175924e-05,
      "loss": 1.4315,
      "step": 2342
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.304895124222184e-05,
      "loss": 1.179,
      "step": 2343
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.304210273239386e-05,
      "loss": 0.413,
      "step": 2344
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.303525110277172e-05,
      "loss": 1.1463,
      "step": 2345
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.3028396353852e-05,
      "loss": 0.7139,
      "step": 2346
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.30215384861316e-05,
      "loss": 0.5033,
      "step": 2347
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.301467750010757e-05,
      "loss": 1.1746,
      "step": 2348
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.30078133962772e-05,
      "loss": 0.6468,
      "step": 2349
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.300094617513803e-05,
      "loss": 1.0015,
      "step": 2350
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.299407583718782e-05,
      "loss": 0.8784,
      "step": 2351
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.298720238292456e-05,
      "loss": 1.0966,
      "step": 2352
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.29803258128464e-05,
      "loss": 1.394,
      "step": 2353
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.29734461274518e-05,
      "loss": 0.6013,
      "step": 2354
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.296656332723945e-05,
      "loss": 0.9131,
      "step": 2355
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.295967741270816e-05,
      "loss": 1.2088,
      "step": 2356
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.295278838435712e-05,
      "loss": 0.9401,
      "step": 2357
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.29458962426856e-05,
      "loss": 0.9589,
      "step": 2358
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.29390009881932e-05,
      "loss": 1.197,
      "step": 2359
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.293210262137965e-05,
      "loss": 0.6047,
      "step": 2360
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.292520114274503e-05,
      "loss": 0.8485,
      "step": 2361
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.291829655278952e-05,
      "loss": 1.2291,
      "step": 2362
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.291138885201362e-05,
      "loss": 0.966,
      "step": 2363
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.290447804091798e-05,
      "loss": 1.1629,
      "step": 2364
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.289756412000355e-05,
      "loss": 0.9579,
      "step": 2365
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.289064708977144e-05,
      "loss": 1.1652,
      "step": 2366
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.288372695072302e-05,
      "loss": 1.4348,
      "step": 2367
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.287680370335987e-05,
      "loss": 1.1529,
      "step": 2368
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.286987734818382e-05,
      "loss": 0.8721,
      "step": 2369
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.286294788569691e-05,
      "loss": 1.4012,
      "step": 2370
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.285601531640139e-05,
      "loss": 0.8828,
      "step": 2371
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.284907964079977e-05,
      "loss": 0.5792,
      "step": 2372
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.284214085939472e-05,
      "loss": 1.4052,
      "step": 2373
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.283519897268922e-05,
      "loss": 1.1849,
      "step": 2374
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.282825398118644e-05,
      "loss": 1.5093,
      "step": 2375
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.282130588538973e-05,
      "loss": 0.628,
      "step": 2376
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.281435468580272e-05,
      "loss": 0.9389,
      "step": 2377
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.280740038292928e-05,
      "loss": 1.7086,
      "step": 2378
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.280044297727344e-05,
      "loss": 1.2582,
      "step": 2379
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.27934824693395e-05,
      "loss": 1.2299,
      "step": 2380
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.278651885963199e-05,
      "loss": 1.1864,
      "step": 2381
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.277955214865561e-05,
      "loss": 1.033,
      "step": 2382
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.277258233691534e-05,
      "loss": 1.0124,
      "step": 2383
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.276560942491641e-05,
      "loss": 1.4809,
      "step": 2384
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.275863341316417e-05,
      "loss": 0.8865,
      "step": 2385
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.275165430216431e-05,
      "loss": 0.975,
      "step": 2386
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.274467209242265e-05,
      "loss": 1.2693,
      "step": 2387
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.273768678444532e-05,
      "loss": 0.8972,
      "step": 2388
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.273069837873859e-05,
      "loss": 0.9775,
      "step": 2389
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.272370687580903e-05,
      "loss": 0.5566,
      "step": 2390
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.271671227616337e-05,
      "loss": 0.7194,
      "step": 2391
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.270971458030862e-05,
      "loss": 1.0877,
      "step": 2392
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.270271378875198e-05,
      "loss": 1.0762,
      "step": 2393
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.269570990200086e-05,
      "loss": 1.1277,
      "step": 2394
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.268870292056297e-05,
      "loss": 1.4221,
      "step": 2395
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.268169284494617e-05,
      "loss": 0.9119,
      "step": 2396
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.267467967565854e-05,
      "loss": 0.7998,
      "step": 2397
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.266766341320845e-05,
      "loss": 2.0255,
      "step": 2398
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.266064405810444e-05,
      "loss": 1.0327,
      "step": 2399
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.265362161085529e-05,
      "loss": 0.9148,
      "step": 2400
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.264659607197e-05,
      "loss": 1.0649,
      "step": 2401
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.263956744195781e-05,
      "loss": 0.5221,
      "step": 2402
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.263253572132816e-05,
      "loss": 1.0479,
      "step": 2403
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.26255009105907e-05,
      "loss": 1.5084,
      "step": 2404
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.261846301025538e-05,
      "loss": 1.1166,
      "step": 2405
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.26114220208323e-05,
      "loss": 1.414,
      "step": 2406
    },
    {
      "epoch": 1.0,
      "eval_loss": 1.1093326807022095,
      "eval_runtime": 340.3735,
      "eval_samples_per_second": 6.942,
      "eval_steps_per_second": 1.736,
      "step": 2406
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.260437794283183e-05,
      "loss": 1.1388,
      "step": 2407
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.25973307767645e-05,
      "loss": 0.492,
      "step": 2408
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.259028052314111e-05,
      "loss": 0.6312,
      "step": 2409
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.258322718247272e-05,
      "loss": 0.8885,
      "step": 2410
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.257617075527054e-05,
      "loss": 0.5899,
      "step": 2411
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.256911124204605e-05,
      "loss": 0.9793,
      "step": 2412
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.256204864331094e-05,
      "loss": 0.4544,
      "step": 2413
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.255498295957712e-05,
      "loss": 0.8411,
      "step": 2414
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.254791419135671e-05,
      "loss": 0.5192,
      "step": 2415
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.254084233916209e-05,
      "loss": 0.6222,
      "step": 2416
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.253376740350587e-05,
      "loss": 1.0724,
      "step": 2417
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.252668938490081e-05,
      "loss": 0.7525,
      "step": 2418
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.251960828385997e-05,
      "loss": 0.4408,
      "step": 2419
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.251252410089659e-05,
      "loss": 0.6371,
      "step": 2420
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.250543683652416e-05,
      "loss": 0.8514,
      "step": 2421
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.249834649125638e-05,
      "loss": 1.0752,
      "step": 2422
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.249125306560717e-05,
      "loss": 0.7912,
      "step": 2423
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.248415656009069e-05,
      "loss": 0.4549,
      "step": 2424
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.247705697522128e-05,
      "loss": 0.7096,
      "step": 2425
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.24699543115136e-05,
      "loss": 0.7318,
      "step": 2426
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.246284856948239e-05,
      "loss": 0.5925,
      "step": 2427
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.245573974964275e-05,
      "loss": 0.477,
      "step": 2428
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.244862785250991e-05,
      "loss": 0.6783,
      "step": 2429
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.244151287859936e-05,
      "loss": 0.8235,
      "step": 2430
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.243439482842685e-05,
      "loss": 0.5053,
      "step": 2431
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.242727370250825e-05,
      "loss": 0.7831,
      "step": 2432
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.242014950135976e-05,
      "loss": 0.6826,
      "step": 2433
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.241302222549775e-05,
      "loss": 0.4453,
      "step": 2434
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.240589187543883e-05,
      "loss": 0.4571,
      "step": 2435
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.23987584516998e-05,
      "loss": 0.4959,
      "step": 2436
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.239162195479773e-05,
      "loss": 0.7522,
      "step": 2437
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.238448238524987e-05,
      "loss": 1.0797,
      "step": 2438
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.237733974357373e-05,
      "loss": 0.5499,
      "step": 2439
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.237019403028702e-05,
      "loss": 0.4424,
      "step": 2440
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.236304524590768e-05,
      "loss": 0.7186,
      "step": 2441
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.235589339095389e-05,
      "loss": 1.1061,
      "step": 2442
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.234873846594398e-05,
      "loss": 0.922,
      "step": 2443
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.234158047139662e-05,
      "loss": 0.8278,
      "step": 2444
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.233441940783058e-05,
      "loss": 0.9188,
      "step": 2445
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.232725527576496e-05,
      "loss": 0.4709,
      "step": 2446
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.232008807571901e-05,
      "loss": 0.6649,
      "step": 2447
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.231291780821222e-05,
      "loss": 0.6347,
      "step": 2448
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.230574447376432e-05,
      "loss": 0.7758,
      "step": 2449
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.229856807289526e-05,
      "loss": 0.7476,
      "step": 2450
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.229138860612516e-05,
      "loss": 0.2909,
      "step": 2451
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.228420607397445e-05,
      "loss": 0.7219,
      "step": 2452
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.227702047696372e-05,
      "loss": 0.6533,
      "step": 2453
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.226983181561382e-05,
      "loss": 1.1085,
      "step": 2454
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.226264009044577e-05,
      "loss": 0.3457,
      "step": 2455
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.225544530198084e-05,
      "loss": 0.3473,
      "step": 2456
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.224824745074055e-05,
      "loss": 0.586,
      "step": 2457
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.22410465372466e-05,
      "loss": 0.4406,
      "step": 2458
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.223384256202095e-05,
      "loss": 0.5697,
      "step": 2459
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.222663552558575e-05,
      "loss": 0.4296,
      "step": 2460
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.221942542846335e-05,
      "loss": 0.6199,
      "step": 2461
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.221221227117643e-05,
      "loss": 0.7882,
      "step": 2462
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.220499605424775e-05,
      "loss": 1.0547,
      "step": 2463
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.219777677820038e-05,
      "loss": 1.0891,
      "step": 2464
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.219055444355762e-05,
      "loss": 0.6205,
      "step": 2465
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.218332905084291e-05,
      "loss": 0.8556,
      "step": 2466
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.217610060058e-05,
      "loss": 0.6477,
      "step": 2467
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.216886909329279e-05,
      "loss": 0.3721,
      "step": 2468
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.216163452950548e-05,
      "loss": 0.701,
      "step": 2469
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.215439690974243e-05,
      "loss": 0.7562,
      "step": 2470
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.214715623452823e-05,
      "loss": 0.6508,
      "step": 2471
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.213991250438772e-05,
      "loss": 0.6872,
      "step": 2472
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.213266571984594e-05,
      "loss": 0.4958,
      "step": 2473
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.212541588142811e-05,
      "loss": 0.8532,
      "step": 2474
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.211816298965979e-05,
      "loss": 0.3307,
      "step": 2475
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.211090704506664e-05,
      "loss": 0.5595,
      "step": 2476
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.21036480481746e-05,
      "loss": 0.5883,
      "step": 2477
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.20963859995098e-05,
      "loss": 0.8123,
      "step": 2478
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.208912089959865e-05,
      "loss": 0.6159,
      "step": 2479
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.20818527489677e-05,
      "loss": 1.0293,
      "step": 2480
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.207458154814379e-05,
      "loss": 0.3658,
      "step": 2481
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.206730729765396e-05,
      "loss": 0.5057,
      "step": 2482
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.206002999802544e-05,
      "loss": 0.6132,
      "step": 2483
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.205274964978572e-05,
      "loss": 0.2993,
      "step": 2484
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.204546625346249e-05,
      "loss": 0.9525,
      "step": 2485
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.203817980958367e-05,
      "loss": 0.8594,
      "step": 2486
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.20308903186774e-05,
      "loss": 0.7309,
      "step": 2487
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.202359778127206e-05,
      "loss": 0.7146,
      "step": 2488
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.201630219789619e-05,
      "loss": 0.8635,
      "step": 2489
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.200900356907862e-05,
      "loss": 0.6473,
      "step": 2490
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.200170189534836e-05,
      "loss": 0.7131,
      "step": 2491
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.199439717723466e-05,
      "loss": 0.6339,
      "step": 2492
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.198708941526699e-05,
      "loss": 0.6381,
      "step": 2493
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.197977860997502e-05,
      "loss": 0.5891,
      "step": 2494
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.197246476188865e-05,
      "loss": 0.8898,
      "step": 2495
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.196514787153803e-05,
      "loss": 0.6094,
      "step": 2496
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.195782793945346e-05,
      "loss": 0.8948,
      "step": 2497
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.195050496616557e-05,
      "loss": 0.6892,
      "step": 2498
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.19431789522051e-05,
      "loss": 0.5584,
      "step": 2499
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.193584989810306e-05,
      "loss": 0.8132,
      "step": 2500
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.192851780439072e-05,
      "loss": 1.1046,
      "step": 2501
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.192118267159946e-05,
      "loss": 0.5444,
      "step": 2502
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.1913844500261e-05,
      "loss": 0.886,
      "step": 2503
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.19065032909072e-05,
      "loss": 0.9214,
      "step": 2504
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.189915904407019e-05,
      "loss": 0.839,
      "step": 2505
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.189181176028228e-05,
      "loss": 0.3958,
      "step": 2506
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.188446144007601e-05,
      "loss": 0.8672,
      "step": 2507
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.187710808398419e-05,
      "loss": 0.6025,
      "step": 2508
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.186975169253977e-05,
      "loss": 0.6416,
      "step": 2509
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.186239226627598e-05,
      "loss": 0.771,
      "step": 2510
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.185502980572624e-05,
      "loss": 0.4333,
      "step": 2511
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.184766431142419e-05,
      "loss": 0.4765,
      "step": 2512
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.184029578390372e-05,
      "loss": 0.9606,
      "step": 2513
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.183292422369891e-05,
      "loss": 0.832,
      "step": 2514
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.182554963134406e-05,
      "loss": 0.5711,
      "step": 2515
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.18181720073737e-05,
      "loss": 0.5969,
      "step": 2516
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.181079135232258e-05,
      "loss": 0.7818,
      "step": 2517
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.180340766672568e-05,
      "loss": 0.7688,
      "step": 2518
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.179602095111817e-05,
      "loss": 0.8002,
      "step": 2519
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.178863120603545e-05,
      "loss": 0.4214,
      "step": 2520
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.178123843201315e-05,
      "loss": 0.4901,
      "step": 2521
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.177384262958718e-05,
      "loss": 0.7928,
      "step": 2522
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.176644379929349e-05,
      "loss": 0.3993,
      "step": 2523
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.175904194166844e-05,
      "loss": 0.6772,
      "step": 2524
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.175163705724854e-05,
      "loss": 0.7462,
      "step": 2525
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.174422914657049e-05,
      "loss": 0.766,
      "step": 2526
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.173681821017121e-05,
      "loss": 0.5428,
      "step": 2527
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.172940424858791e-05,
      "loss": 0.3808,
      "step": 2528
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.172198726235794e-05,
      "loss": 0.9379,
      "step": 2529
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.17145672520189e-05,
      "loss": 0.4787,
      "step": 2530
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.170714421810864e-05,
      "loss": 0.3134,
      "step": 2531
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.169971816116518e-05,
      "loss": 0.3715,
      "step": 2532
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.169228908172675e-05,
      "loss": 0.3897,
      "step": 2533
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.168485698033188e-05,
      "loss": 0.9403,
      "step": 2534
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.167742185751924e-05,
      "loss": 0.7018,
      "step": 2535
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.166998371382775e-05,
      "loss": 0.6294,
      "step": 2536
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.166254254979654e-05,
      "loss": 0.5487,
      "step": 2537
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.165509836596495e-05,
      "loss": 0.4948,
      "step": 2538
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.164765116287259e-05,
      "loss": 0.528,
      "step": 2539
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.164020094105921e-05,
      "loss": 0.4147,
      "step": 2540
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.163274770106485e-05,
      "loss": 0.3834,
      "step": 2541
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.162529144342971e-05,
      "loss": 0.7078,
      "step": 2542
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.161783216869428e-05,
      "loss": 0.5461,
      "step": 2543
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.161036987739919e-05,
      "loss": 1.1072,
      "step": 2544
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.160290457008531e-05,
      "loss": 0.4638,
      "step": 2545
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.159543624729379e-05,
      "loss": 0.9907,
      "step": 2546
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.158796490956593e-05,
      "loss": 0.5599,
      "step": 2547
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.158049055744326e-05,
      "loss": 0.5704,
      "step": 2548
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.157301319146756e-05,
      "loss": 0.7754,
      "step": 2549
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.156553281218079e-05,
      "loss": 0.9563,
      "step": 2550
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.155804942012516e-05,
      "loss": 0.5335,
      "step": 2551
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.155056301584306e-05,
      "loss": 0.7228,
      "step": 2552
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.154307359987714e-05,
      "loss": 0.4326,
      "step": 2553
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.153558117277027e-05,
      "loss": 0.4191,
      "step": 2554
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.152808573506549e-05,
      "loss": 0.6721,
      "step": 2555
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.152058728730608e-05,
      "loss": 1.2385,
      "step": 2556
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.151308583003557e-05,
      "loss": 0.4626,
      "step": 2557
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.150558136379768e-05,
      "loss": 0.4959,
      "step": 2558
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.149807388913634e-05,
      "loss": 0.572,
      "step": 2559
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.149056340659572e-05,
      "loss": 0.8054,
      "step": 2560
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.148304991672019e-05,
      "loss": 0.8096,
      "step": 2561
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.147553342005434e-05,
      "loss": 1.0112,
      "step": 2562
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.146801391714299e-05,
      "loss": 0.4479,
      "step": 2563
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.146049140853119e-05,
      "loss": 0.3235,
      "step": 2564
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.145296589476417e-05,
      "loss": 0.7407,
      "step": 2565
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.14454373763874e-05,
      "loss": 0.8478,
      "step": 2566
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.143790585394655e-05,
      "loss": 0.597,
      "step": 2567
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.143037132798755e-05,
      "loss": 0.6113,
      "step": 2568
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.142283379905651e-05,
      "loss": 0.8974,
      "step": 2569
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.141529326769976e-05,
      "loss": 0.4278,
      "step": 2570
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.140774973446387e-05,
      "loss": 1.3067,
      "step": 2571
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.140020319989559e-05,
      "loss": 0.643,
      "step": 2572
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.139265366454195e-05,
      "loss": 0.877,
      "step": 2573
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.138510112895011e-05,
      "loss": 0.5493,
      "step": 2574
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.137754559366754e-05,
      "loss": 0.7189,
      "step": 2575
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.136998705924185e-05,
      "loss": 0.912,
      "step": 2576
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.136242552622092e-05,
      "loss": 0.7453,
      "step": 2577
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.135486099515283e-05,
      "loss": 0.6665,
      "step": 2578
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.134729346658586e-05,
      "loss": 0.5843,
      "step": 2579
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.133972294106851e-05,
      "loss": 0.4556,
      "step": 2580
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.133214941914956e-05,
      "loss": 0.4825,
      "step": 2581
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.132457290137792e-05,
      "loss": 0.7299,
      "step": 2582
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.131699338830275e-05,
      "loss": 0.7095,
      "step": 2583
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.130941088047345e-05,
      "loss": 0.8716,
      "step": 2584
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.13018253784396e-05,
      "loss": 0.6251,
      "step": 2585
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.129423688275103e-05,
      "loss": 0.9428,
      "step": 2586
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.128664539395777e-05,
      "loss": 0.7273,
      "step": 2587
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.127905091261007e-05,
      "loss": 0.7214,
      "step": 2588
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.127145343925838e-05,
      "loss": 0.6213,
      "step": 2589
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.12638529744534e-05,
      "loss": 1.1145,
      "step": 2590
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.125624951874603e-05,
      "loss": 0.5008,
      "step": 2591
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.124864307268736e-05,
      "loss": 0.7038,
      "step": 2592
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.124103363682879e-05,
      "loss": 0.5995,
      "step": 2593
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.123342121172179e-05,
      "loss": 0.9316,
      "step": 2594
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.122580579791817e-05,
      "loss": 0.5173,
      "step": 2595
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.12181873959699e-05,
      "loss": 0.6394,
      "step": 2596
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.12105660064292e-05,
      "loss": 1.194,
      "step": 2597
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.120294162984846e-05,
      "loss": 0.7748,
      "step": 2598
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.119531426678033e-05,
      "loss": 0.5739,
      "step": 2599
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.118768391777766e-05,
      "loss": 0.5178,
      "step": 2600
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.118005058339349e-05,
      "loss": 0.3744,
      "step": 2601
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.117241426418115e-05,
      "loss": 0.6395,
      "step": 2602
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.116477496069408e-05,
      "loss": 0.7701,
      "step": 2603
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.115713267348606e-05,
      "loss": 0.4036,
      "step": 2604
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.114948740311097e-05,
      "loss": 0.4265,
      "step": 2605
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.114183915012298e-05,
      "loss": 0.6588,
      "step": 2606
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.113418791507644e-05,
      "loss": 0.6259,
      "step": 2607
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.112653369852593e-05,
      "loss": 0.7213,
      "step": 2608
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.111887650102626e-05,
      "loss": 0.6366,
      "step": 2609
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.111121632313244e-05,
      "loss": 0.8743,
      "step": 2610
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.11035531653997e-05,
      "loss": 0.6354,
      "step": 2611
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.109588702838348e-05,
      "loss": 0.3231,
      "step": 2612
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.108821791263942e-05,
      "loss": 0.5648,
      "step": 2613
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.108054581872343e-05,
      "loss": 1.0458,
      "step": 2614
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.107287074719158e-05,
      "loss": 0.9555,
      "step": 2615
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.106519269860017e-05,
      "loss": 0.4559,
      "step": 2616
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.105751167350574e-05,
      "loss": 0.6922,
      "step": 2617
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.104982767246503e-05,
      "loss": 0.7112,
      "step": 2618
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.104214069603498e-05,
      "loss": 0.5866,
      "step": 2619
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.103445074477279e-05,
      "loss": 0.6457,
      "step": 2620
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.102675781923581e-05,
      "loss": 0.787,
      "step": 2621
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.101906191998166e-05,
      "loss": 0.6289,
      "step": 2622
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.101136304756815e-05,
      "loss": 1.0767,
      "step": 2623
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.100366120255333e-05,
      "loss": 0.9206,
      "step": 2624
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.099595638549542e-05,
      "loss": 0.8106,
      "step": 2625
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.098824859695291e-05,
      "loss": 0.6052,
      "step": 2626
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.098053783748446e-05,
      "loss": 0.9496,
      "step": 2627
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.097282410764898e-05,
      "loss": 1.1007,
      "step": 2628
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.096510740800555e-05,
      "loss": 0.6803,
      "step": 2629
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.095738773911353e-05,
      "loss": 0.5939,
      "step": 2630
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.094966510153246e-05,
      "loss": 0.5568,
      "step": 2631
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.094193949582206e-05,
      "loss": 0.4663,
      "step": 2632
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.093421092254232e-05,
      "loss": 0.5703,
      "step": 2633
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.092647938225345e-05,
      "loss": 0.5993,
      "step": 2634
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.091874487551581e-05,
      "loss": 0.8606,
      "step": 2635
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.091100740289005e-05,
      "loss": 0.4954,
      "step": 2636
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.090326696493697e-05,
      "loss": 0.6305,
      "step": 2637
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.089552356221762e-05,
      "loss": 0.8093,
      "step": 2638
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.088777719529328e-05,
      "loss": 0.8843,
      "step": 2639
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.08800278647254e-05,
      "loss": 0.9596,
      "step": 2640
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.08722755710757e-05,
      "loss": 0.3091,
      "step": 2641
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.086452031490606e-05,
      "loss": 0.4866,
      "step": 2642
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.085676209677863e-05,
      "loss": 0.4815,
      "step": 2643
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.08490009172557e-05,
      "loss": 0.6179,
      "step": 2644
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.084123677689984e-05,
      "loss": 0.6822,
      "step": 2645
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.083346967627381e-05,
      "loss": 0.7411,
      "step": 2646
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.08256996159406e-05,
      "loss": 0.6681,
      "step": 2647
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.081792659646341e-05,
      "loss": 0.6385,
      "step": 2648
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.081015061840561e-05,
      "loss": 1.0552,
      "step": 2649
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.080237168233086e-05,
      "loss": 0.5548,
      "step": 2650
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.079458978880297e-05,
      "loss": 0.866,
      "step": 2651
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.0786804938386e-05,
      "loss": 1.4014,
      "step": 2652
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.07790171316442e-05,
      "loss": 0.4053,
      "step": 2653
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.077122636914209e-05,
      "loss": 0.9634,
      "step": 2654
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.076343265144432e-05,
      "loss": 0.647,
      "step": 2655
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.075563597911583e-05,
      "loss": 0.4087,
      "step": 2656
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.07478363527217e-05,
      "loss": 0.6818,
      "step": 2657
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.07400337728273e-05,
      "loss": 0.4912,
      "step": 2658
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.073222823999816e-05,
      "loss": 0.5853,
      "step": 2659
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.072441975480006e-05,
      "loss": 0.2619,
      "step": 2660
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.071660831779898e-05,
      "loss": 0.9038,
      "step": 2661
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.07087939295611e-05,
      "loss": 0.7197,
      "step": 2662
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.070097659065282e-05,
      "loss": 0.2644,
      "step": 2663
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.069315630164077e-05,
      "loss": 0.7726,
      "step": 2664
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.068533306309177e-05,
      "loss": 0.5682,
      "step": 2665
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.06775068755729e-05,
      "loss": 0.4745,
      "step": 2666
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.066967773965138e-05,
      "loss": 0.5426,
      "step": 2667
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.066184565589471e-05,
      "loss": 0.3863,
      "step": 2668
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.065401062487056e-05,
      "loss": 0.6736,
      "step": 2669
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.064617264714685e-05,
      "loss": 0.7772,
      "step": 2670
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.063833172329168e-05,
      "loss": 0.7175,
      "step": 2671
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.06304878538734e-05,
      "loss": 0.5554,
      "step": 2672
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.062264103946053e-05,
      "loss": 0.9886,
      "step": 2673
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.061479128062183e-05,
      "loss": 0.9234,
      "step": 2674
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.060693857792628e-05,
      "loss": 0.6471,
      "step": 2675
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.059908293194306e-05,
      "loss": 0.7875,
      "step": 2676
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.059122434324156e-05,
      "loss": 0.7133,
      "step": 2677
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.058336281239139e-05,
      "loss": 0.5599,
      "step": 2678
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.057549833996237e-05,
      "loss": 0.5127,
      "step": 2679
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.056763092652454e-05,
      "loss": 0.6813,
      "step": 2680
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.055976057264815e-05,
      "loss": 0.811,
      "step": 2681
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.055188727890366e-05,
      "loss": 0.4623,
      "step": 2682
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.054401104586175e-05,
      "loss": 0.4913,
      "step": 2683
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.053613187409331e-05,
      "loss": 0.7578,
      "step": 2684
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.052824976416941e-05,
      "loss": 1.057,
      "step": 2685
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.05203647166614e-05,
      "loss": 0.7079,
      "step": 2686
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.051247673214081e-05,
      "loss": 0.7492,
      "step": 2687
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.050458581117935e-05,
      "loss": 0.5016,
      "step": 2688
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.0496691954349e-05,
      "loss": 0.5525,
      "step": 2689
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.048879516222192e-05,
      "loss": 0.3531,
      "step": 2690
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.048089543537047e-05,
      "loss": 0.6307,
      "step": 2691
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.047299277436726e-05,
      "loss": 1.1069,
      "step": 2692
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.046508717978507e-05,
      "loss": 0.775,
      "step": 2693
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.045717865219696e-05,
      "loss": 0.7284,
      "step": 2694
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.044926719217611e-05,
      "loss": 0.7227,
      "step": 2695
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.0441352800296e-05,
      "loss": 0.5411,
      "step": 2696
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.043343547713026e-05,
      "loss": 0.6813,
      "step": 2697
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.042551522325276e-05,
      "loss": 0.6365,
      "step": 2698
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.041759203923759e-05,
      "loss": 0.5782,
      "step": 2699
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.040966592565902e-05,
      "loss": 0.7053,
      "step": 2700
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.040173688309159e-05,
      "loss": 0.7523,
      "step": 2701
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.039380491210999e-05,
      "loss": 0.7215,
      "step": 2702
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.038587001328915e-05,
      "loss": 0.5992,
      "step": 2703
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.037793218720421e-05,
      "loss": 0.439,
      "step": 2704
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.036999143443053e-05,
      "loss": 0.5182,
      "step": 2705
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.036204775554366e-05,
      "loss": 0.927,
      "step": 2706
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.035410115111939e-05,
      "loss": 0.5341,
      "step": 2707
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.034615162173371e-05,
      "loss": 0.3065,
      "step": 2708
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.033819916796282e-05,
      "loss": 0.9418,
      "step": 2709
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.033024379038311e-05,
      "loss": 0.3779,
      "step": 2710
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.032228548957124e-05,
      "loss": 0.9694,
      "step": 2711
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.031432426610402e-05,
      "loss": 0.5858,
      "step": 2712
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.030636012055854e-05,
      "loss": 0.7584,
      "step": 2713
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.029839305351199e-05,
      "loss": 0.4046,
      "step": 2714
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.02904230655419e-05,
      "loss": 0.6566,
      "step": 2715
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.028245015722592e-05,
      "loss": 0.8548,
      "step": 2716
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.027447432914197e-05,
      "loss": 0.5344,
      "step": 2717
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.026649558186815e-05,
      "loss": 0.5242,
      "step": 2718
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.025851391598278e-05,
      "loss": 0.9932,
      "step": 2719
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.025052933206438e-05,
      "loss": 0.7686,
      "step": 2720
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.024254183069169e-05,
      "loss": 0.7766,
      "step": 2721
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.023455141244369e-05,
      "loss": 0.5755,
      "step": 2722
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.022655807789951e-05,
      "loss": 0.2665,
      "step": 2723
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.021856182763855e-05,
      "loss": 0.73,
      "step": 2724
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.021056266224039e-05,
      "loss": 0.7065,
      "step": 2725
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.020256058228482e-05,
      "loss": 0.6212,
      "step": 2726
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.019455558835185e-05,
      "loss": 0.5688,
      "step": 2727
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.018654768102172e-05,
      "loss": 0.5874,
      "step": 2728
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.017853686087485e-05,
      "loss": 0.6924,
      "step": 2729
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.017052312849189e-05,
      "loss": 0.6646,
      "step": 2730
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.016250648445366e-05,
      "loss": 0.4761,
      "step": 2731
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.015448692934128e-05,
      "loss": 1.1007,
      "step": 2732
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.014646446373599e-05,
      "loss": 0.5256,
      "step": 2733
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.013843908821929e-05,
      "loss": 0.3643,
      "step": 2734
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.013041080337287e-05,
      "loss": 0.4737,
      "step": 2735
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.012237960977864e-05,
      "loss": 0.5003,
      "step": 2736
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.011434550801873e-05,
      "loss": 0.6448,
      "step": 2737
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.010630849867548e-05,
      "loss": 0.8185,
      "step": 2738
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.009826858233141e-05,
      "loss": 0.9148,
      "step": 2739
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.009022575956927e-05,
      "loss": 0.8062,
      "step": 2740
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.008218003097205e-05,
      "loss": 0.9915,
      "step": 2741
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.007413139712288e-05,
      "loss": 0.6854,
      "step": 2742
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.00660798586052e-05,
      "loss": 0.574,
      "step": 2743
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.005802541600257e-05,
      "loss": 1.1088,
      "step": 2744
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.004996806989878e-05,
      "loss": 1.0425,
      "step": 2745
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.004190782087787e-05,
      "loss": 0.3426,
      "step": 2746
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.003384466952407e-05,
      "loss": 0.7291,
      "step": 2747
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.00257786164218e-05,
      "loss": 0.4972,
      "step": 2748
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.001770966215573e-05,
      "loss": 0.4456,
      "step": 2749
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.000963780731071e-05,
      "loss": 0.8312,
      "step": 2750
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.000156305247179e-05,
      "loss": 0.7314,
      "step": 2751
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.999348539822424e-05,
      "loss": 0.4167,
      "step": 2752
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.998540484515356e-05,
      "loss": 0.7542,
      "step": 2753
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.997732139384548e-05,
      "loss": 0.8444,
      "step": 2754
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.996923504488586e-05,
      "loss": 0.2621,
      "step": 2755
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.996114579886086e-05,
      "loss": 0.3644,
      "step": 2756
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.995305365635677e-05,
      "loss": 0.7423,
      "step": 2757
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.994495861796015e-05,
      "loss": 0.6053,
      "step": 2758
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.993686068425772e-05,
      "loss": 0.95,
      "step": 2759
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.992875985583649e-05,
      "loss": 0.8906,
      "step": 2760
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.992065613328358e-05,
      "loss": 0.377,
      "step": 2761
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.991254951718637e-05,
      "loss": 0.6171,
      "step": 2762
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.99044400081325e-05,
      "loss": 0.6113,
      "step": 2763
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.98963276067097e-05,
      "loss": 0.4831,
      "step": 2764
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.988821231350599e-05,
      "loss": 0.6718,
      "step": 2765
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.988009412910962e-05,
      "loss": 0.4841,
      "step": 2766
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.987197305410899e-05,
      "loss": 0.7649,
      "step": 2767
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.986384908909273e-05,
      "loss": 0.9387,
      "step": 2768
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.98557222346497e-05,
      "loss": 0.7079,
      "step": 2769
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.984759249136894e-05,
      "loss": 0.6944,
      "step": 2770
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.983945985983972e-05,
      "loss": 0.4283,
      "step": 2771
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.983132434065153e-05,
      "loss": 0.7112,
      "step": 2772
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.982318593439402e-05,
      "loss": 0.7744,
      "step": 2773
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.98150446416571e-05,
      "loss": 0.8457,
      "step": 2774
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.980690046303087e-05,
      "loss": 0.278,
      "step": 2775
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.979875339910562e-05,
      "loss": 0.6957,
      "step": 2776
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.97906034504719e-05,
      "loss": 0.6288,
      "step": 2777
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.978245061772041e-05,
      "loss": 0.4109,
      "step": 2778
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.977429490144212e-05,
      "loss": 0.6532,
      "step": 2779
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.976613630222812e-05,
      "loss": 0.619,
      "step": 2780
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.975797482066981e-05,
      "loss": 0.4861,
      "step": 2781
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.974981045735875e-05,
      "loss": 0.801,
      "step": 2782
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.974164321288671e-05,
      "loss": 0.8523,
      "step": 2783
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.973347308784565e-05,
      "loss": 0.6692,
      "step": 2784
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.972530008282781e-05,
      "loss": 0.6204,
      "step": 2785
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.971712419842552e-05,
      "loss": 0.5818,
      "step": 2786
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.970894543523143e-05,
      "loss": 0.8203,
      "step": 2787
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.970076379383836e-05,
      "loss": 0.421,
      "step": 2788
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.969257927483931e-05,
      "loss": 0.78,
      "step": 2789
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.968439187882753e-05,
      "loss": 0.9,
      "step": 2790
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.967620160639647e-05,
      "loss": 0.4039,
      "step": 2791
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.966800845813976e-05,
      "loss": 0.6994,
      "step": 2792
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.965981243465128e-05,
      "loss": 0.7828,
      "step": 2793
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.965161353652506e-05,
      "loss": 0.5483,
      "step": 2794
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.964341176435545e-05,
      "loss": 0.4474,
      "step": 2795
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.963520711873685e-05,
      "loss": 0.6896,
      "step": 2796
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.9626999600264e-05,
      "loss": 1.015,
      "step": 2797
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.961878920953177e-05,
      "loss": 0.4503,
      "step": 2798
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.961057594713531e-05,
      "loss": 0.9483,
      "step": 2799
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.96023598136699e-05,
      "loss": 1.0205,
      "step": 2800
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.959414080973107e-05,
      "loss": 0.5675,
      "step": 2801
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.958591893591459e-05,
      "loss": 0.6257,
      "step": 2802
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.957769419281635e-05,
      "loss": 0.7793,
      "step": 2803
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.956946658103254e-05,
      "loss": 0.7098,
      "step": 2804
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.956123610115949e-05,
      "loss": 0.7322,
      "step": 2805
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.955300275379376e-05,
      "loss": 0.5874,
      "step": 2806
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.954476653953216e-05,
      "loss": 0.6125,
      "step": 2807
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.953652745897163e-05,
      "loss": 0.5347,
      "step": 2808
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.952828551270938e-05,
      "loss": 0.6045,
      "step": 2809
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.95200407013428e-05,
      "loss": 0.626,
      "step": 2810
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.95117930254695e-05,
      "loss": 0.4753,
      "step": 2811
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.950354248568729e-05,
      "loss": 0.9173,
      "step": 2812
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.949528908259418e-05,
      "loss": 0.5619,
      "step": 2813
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.948703281678838e-05,
      "loss": 0.981,
      "step": 2814
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.947877368886837e-05,
      "loss": 0.7344,
      "step": 2815
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.947051169943274e-05,
      "loss": 0.8142,
      "step": 2816
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.946224684908039e-05,
      "loss": 0.3685,
      "step": 2817
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.945397913841034e-05,
      "loss": 0.9367,
      "step": 2818
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.944570856802186e-05,
      "loss": 0.6845,
      "step": 2819
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.943743513851443e-05,
      "loss": 0.7567,
      "step": 2820
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.942915885048772e-05,
      "loss": 0.6767,
      "step": 2821
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.942087970454163e-05,
      "loss": 0.5273,
      "step": 2822
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.941259770127624e-05,
      "loss": 0.6218,
      "step": 2823
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.940431284129183e-05,
      "loss": 0.84,
      "step": 2824
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.939602512518894e-05,
      "loss": 0.5086,
      "step": 2825
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.938773455356826e-05,
      "loss": 1.0119,
      "step": 2826
    },
    {
      "epoch": 1.17,
      "learning_rate": 8.937944112703072e-05,
      "loss": 0.9274,
      "step": 2827
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.937114484617746e-05,
      "loss": 0.7977,
      "step": 2828
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.936284571160978e-05,
      "loss": 0.4898,
      "step": 2829
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.935454372392926e-05,
      "loss": 0.8878,
      "step": 2830
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.934623888373761e-05,
      "loss": 0.5816,
      "step": 2831
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.933793119163682e-05,
      "loss": 0.4111,
      "step": 2832
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.932962064822903e-05,
      "loss": 0.4037,
      "step": 2833
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.932130725411662e-05,
      "loss": 0.7963,
      "step": 2834
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.931299100990215e-05,
      "loss": 0.7668,
      "step": 2835
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.93046719161884e-05,
      "loss": 0.7895,
      "step": 2836
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.929634997357839e-05,
      "loss": 0.6357,
      "step": 2837
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.928802518267527e-05,
      "loss": 1.2345,
      "step": 2838
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.927969754408248e-05,
      "loss": 0.5328,
      "step": 2839
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.92713670584036e-05,
      "loss": 0.5694,
      "step": 2840
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.926303372624246e-05,
      "loss": 0.7542,
      "step": 2841
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.925469754820306e-05,
      "loss": 0.5215,
      "step": 2842
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.924635852488965e-05,
      "loss": 0.4358,
      "step": 2843
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.923801665690665e-05,
      "loss": 0.8333,
      "step": 2844
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.922967194485869e-05,
      "loss": 0.3554,
      "step": 2845
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.922132438935063e-05,
      "loss": 0.3215,
      "step": 2846
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.921297399098751e-05,
      "loss": 0.8204,
      "step": 2847
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.92046207503746e-05,
      "loss": 0.6582,
      "step": 2848
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.919626466811734e-05,
      "loss": 1.1372,
      "step": 2849
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.918790574482143e-05,
      "loss": 0.8285,
      "step": 2850
    },
    {
      "epoch": 1.18,
      "learning_rate": 8.917954398109272e-05,
      "loss": 0.963,
      "step": 2851
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.91711793775373e-05,
      "loss": 1.0224,
      "step": 2852
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.916281193476146e-05,
      "loss": 0.7778,
      "step": 2853
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.915444165337168e-05,
      "loss": 0.8512,
      "step": 2854
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.914606853397465e-05,
      "loss": 0.7218,
      "step": 2855
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.91376925771773e-05,
      "loss": 0.5491,
      "step": 2856
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.912931378358673e-05,
      "loss": 0.8199,
      "step": 2857
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.912093215381025e-05,
      "loss": 0.4714,
      "step": 2858
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.911254768845539e-05,
      "loss": 0.8003,
      "step": 2859
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.910416038812988e-05,
      "loss": 0.5619,
      "step": 2860
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.909577025344161e-05,
      "loss": 1.0215,
      "step": 2861
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.908737728499876e-05,
      "loss": 1.2036,
      "step": 2862
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.907898148340966e-05,
      "loss": 0.6333,
      "step": 2863
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.907058284928287e-05,
      "loss": 0.6758,
      "step": 2864
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.906218138322712e-05,
      "loss": 0.8092,
      "step": 2865
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.905377708585137e-05,
      "loss": 0.5927,
      "step": 2866
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.904536995776481e-05,
      "loss": 1.0726,
      "step": 2867
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.903695999957677e-05,
      "loss": 0.6121,
      "step": 2868
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.902854721189686e-05,
      "loss": 0.5065,
      "step": 2869
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.902013159533485e-05,
      "loss": 1.1294,
      "step": 2870
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.90117131505007e-05,
      "loss": 0.5852,
      "step": 2871
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.900329187800462e-05,
      "loss": 0.5844,
      "step": 2872
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.8994867778457e-05,
      "loss": 0.7194,
      "step": 2873
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.898644085246845e-05,
      "loss": 0.3972,
      "step": 2874
    },
    {
      "epoch": 1.19,
      "learning_rate": 8.897801110064976e-05,
      "loss": 0.9756,
      "step": 2875
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.896957852361195e-05,
      "loss": 0.5839,
      "step": 2876
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.89611431219662e-05,
      "loss": 1.0196,
      "step": 2877
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.895270489632397e-05,
      "loss": 0.8864,
      "step": 2878
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.894426384729686e-05,
      "loss": 0.5899,
      "step": 2879
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.893581997549671e-05,
      "loss": 0.926,
      "step": 2880
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.892737328153553e-05,
      "loss": 0.4891,
      "step": 2881
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.891892376602559e-05,
      "loss": 1.1357,
      "step": 2882
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.89104714295793e-05,
      "loss": 0.4738,
      "step": 2883
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.890201627280932e-05,
      "loss": 0.5114,
      "step": 2884
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.889355829632851e-05,
      "loss": 1.017,
      "step": 2885
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.888509750074989e-05,
      "loss": 0.639,
      "step": 2886
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.887663388668676e-05,
      "loss": 0.5864,
      "step": 2887
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.886816745475256e-05,
      "loss": 0.5564,
      "step": 2888
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.885969820556095e-05,
      "loss": 0.44,
      "step": 2889
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.885122613972581e-05,
      "loss": 0.9119,
      "step": 2890
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.884275125786123e-05,
      "loss": 1.0371,
      "step": 2891
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.883427356058147e-05,
      "loss": 0.4596,
      "step": 2892
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.882579304850103e-05,
      "loss": 0.5156,
      "step": 2893
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.881730972223458e-05,
      "loss": 0.6899,
      "step": 2894
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.880882358239702e-05,
      "loss": 0.7623,
      "step": 2895
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.880033462960345e-05,
      "loss": 0.9478,
      "step": 2896
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.879184286446915e-05,
      "loss": 0.3854,
      "step": 2897
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.878334828760966e-05,
      "loss": 0.6042,
      "step": 2898
    },
    {
      "epoch": 1.2,
      "learning_rate": 8.877485089964067e-05,
      "loss": 0.6191,
      "step": 2899
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.876635070117807e-05,
      "loss": 0.8412,
      "step": 2900
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.875784769283799e-05,
      "loss": 0.7563,
      "step": 2901
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.874934187523675e-05,
      "loss": 0.6673,
      "step": 2902
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.874083324899089e-05,
      "loss": 0.6148,
      "step": 2903
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.873232181471712e-05,
      "loss": 0.5948,
      "step": 2904
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.872380757303234e-05,
      "loss": 0.4266,
      "step": 2905
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.871529052455372e-05,
      "loss": 0.6017,
      "step": 2906
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.870677066989859e-05,
      "loss": 0.6477,
      "step": 2907
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.869824800968448e-05,
      "loss": 0.956,
      "step": 2908
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.868972254452914e-05,
      "loss": 0.4565,
      "step": 2909
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.86811942750505e-05,
      "loss": 0.4543,
      "step": 2910
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.867266320186675e-05,
      "loss": 0.7422,
      "step": 2911
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.86641293255962e-05,
      "loss": 0.6151,
      "step": 2912
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.865559264685741e-05,
      "loss": 0.7921,
      "step": 2913
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.864705316626915e-05,
      "loss": 0.6665,
      "step": 2914
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.86385108844504e-05,
      "loss": 0.9938,
      "step": 2915
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.86299658020203e-05,
      "loss": 0.8226,
      "step": 2916
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.862141791959821e-05,
      "loss": 0.7661,
      "step": 2917
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.861286723780373e-05,
      "loss": 0.6166,
      "step": 2918
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.860431375725662e-05,
      "loss": 0.6952,
      "step": 2919
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.859575747857685e-05,
      "loss": 0.3073,
      "step": 2920
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.858719840238459e-05,
      "loss": 0.7054,
      "step": 2921
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.857863652930025e-05,
      "loss": 0.5724,
      "step": 2922
    },
    {
      "epoch": 1.21,
      "learning_rate": 8.857007185994441e-05,
      "loss": 0.7354,
      "step": 2923
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.856150439493784e-05,
      "loss": 0.5584,
      "step": 2924
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.855293413490155e-05,
      "loss": 0.5097,
      "step": 2925
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.854436108045671e-05,
      "loss": 0.6964,
      "step": 2926
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.853578523222475e-05,
      "loss": 0.5111,
      "step": 2927
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.852720659082721e-05,
      "loss": 0.4212,
      "step": 2928
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.851862515688596e-05,
      "loss": 0.6866,
      "step": 2929
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.851004093102295e-05,
      "loss": 0.7763,
      "step": 2930
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.850145391386041e-05,
      "loss": 0.8702,
      "step": 2931
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.849286410602075e-05,
      "loss": 1.4094,
      "step": 2932
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.848427150812656e-05,
      "loss": 0.8811,
      "step": 2933
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.847567612080067e-05,
      "loss": 0.5907,
      "step": 2934
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.84670779446661e-05,
      "loss": 0.4002,
      "step": 2935
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.845847698034603e-05,
      "loss": 0.4945,
      "step": 2936
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.844987322846392e-05,
      "loss": 0.9381,
      "step": 2937
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.844126668964337e-05,
      "loss": 0.6273,
      "step": 2938
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.843265736450819e-05,
      "loss": 0.5999,
      "step": 2939
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.842404525368243e-05,
      "loss": 0.8411,
      "step": 2940
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.841543035779031e-05,
      "loss": 0.4307,
      "step": 2941
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.840681267745625e-05,
      "loss": 0.6089,
      "step": 2942
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.839819221330489e-05,
      "loss": 0.7718,
      "step": 2943
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.838956896596104e-05,
      "loss": 0.447,
      "step": 2944
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.838094293604978e-05,
      "loss": 0.6314,
      "step": 2945
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.837231412419629e-05,
      "loss": 0.4644,
      "step": 2946
    },
    {
      "epoch": 1.22,
      "learning_rate": 8.836368253102604e-05,
      "loss": 0.7286,
      "step": 2947
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.835504815716466e-05,
      "loss": 0.6173,
      "step": 2948
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.834641100323799e-05,
      "loss": 0.7405,
      "step": 2949
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.833777106987206e-05,
      "loss": 0.9246,
      "step": 2950
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.832912835769314e-05,
      "loss": 0.9727,
      "step": 2951
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.832048286732766e-05,
      "loss": 0.4466,
      "step": 2952
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.831183459940226e-05,
      "loss": 0.3337,
      "step": 2953
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.83031835545438e-05,
      "loss": 0.9062,
      "step": 2954
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.82945297333793e-05,
      "loss": 1.0026,
      "step": 2955
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.828587313653604e-05,
      "loss": 0.7001,
      "step": 2956
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.827721376464147e-05,
      "loss": 0.403,
      "step": 2957
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.826855161832322e-05,
      "loss": 0.5817,
      "step": 2958
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.825988669820914e-05,
      "loss": 0.7905,
      "step": 2959
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.825121900492733e-05,
      "loss": 0.596,
      "step": 2960
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.824254853910601e-05,
      "loss": 0.6252,
      "step": 2961
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.823387530137364e-05,
      "loss": 0.8115,
      "step": 2962
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.822519929235886e-05,
      "loss": 0.63,
      "step": 2963
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.821652051269056e-05,
      "loss": 0.7723,
      "step": 2964
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.820783896299778e-05,
      "loss": 0.5079,
      "step": 2965
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.819915464390978e-05,
      "loss": 0.5896,
      "step": 2966
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.819046755605603e-05,
      "loss": 0.7116,
      "step": 2967
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.818177770006619e-05,
      "loss": 0.6059,
      "step": 2968
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.81730850765701e-05,
      "loss": 0.7932,
      "step": 2969
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.816438968619785e-05,
      "loss": 1.0503,
      "step": 2970
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.815569152957968e-05,
      "loss": 0.6752,
      "step": 2971
    },
    {
      "epoch": 1.23,
      "learning_rate": 8.814699060734606e-05,
      "loss": 0.7673,
      "step": 2972
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.813828692012766e-05,
      "loss": 0.4474,
      "step": 2973
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.812958046855537e-05,
      "loss": 0.5291,
      "step": 2974
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.81208712532602e-05,
      "loss": 0.6942,
      "step": 2975
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.811215927487344e-05,
      "loss": 0.4569,
      "step": 2976
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.810344453402655e-05,
      "loss": 0.3954,
      "step": 2977
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.809472703135122e-05,
      "loss": 0.7766,
      "step": 2978
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.808600676747927e-05,
      "loss": 0.9646,
      "step": 2979
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.807728374304282e-05,
      "loss": 0.5747,
      "step": 2980
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.806855795867408e-05,
      "loss": 0.8765,
      "step": 2981
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.805982941500557e-05,
      "loss": 1.0583,
      "step": 2982
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.80510981126699e-05,
      "loss": 0.5799,
      "step": 2983
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.804236405229998e-05,
      "loss": 0.4849,
      "step": 2984
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.803362723452886e-05,
      "loss": 1.0672,
      "step": 2985
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.802488765998979e-05,
      "loss": 0.4497,
      "step": 2986
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.801614532931627e-05,
      "loss": 0.8591,
      "step": 2987
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.800740024314193e-05,
      "loss": 0.6402,
      "step": 2988
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.799865240210065e-05,
      "loss": 1.09,
      "step": 2989
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.798990180682651e-05,
      "loss": 0.6028,
      "step": 2990
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.798114845795373e-05,
      "loss": 0.9984,
      "step": 2991
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.797239235611682e-05,
      "loss": 0.5452,
      "step": 2992
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.796363350195043e-05,
      "loss": 0.7859,
      "step": 2993
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.79548718960894e-05,
      "loss": 0.6143,
      "step": 2994
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.794610753916883e-05,
      "loss": 1.1083,
      "step": 2995
    },
    {
      "epoch": 1.24,
      "learning_rate": 8.793734043182393e-05,
      "loss": 0.5595,
      "step": 2996
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.792857057469021e-05,
      "loss": 1.0029,
      "step": 2997
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.791979796840333e-05,
      "loss": 0.444,
      "step": 2998
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.791102261359911e-05,
      "loss": 1.1149,
      "step": 2999
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.790224451091362e-05,
      "loss": 0.7239,
      "step": 3000
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.789346366098315e-05,
      "loss": 0.9066,
      "step": 3001
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.788468006444413e-05,
      "loss": 0.5556,
      "step": 3002
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.78758937219332e-05,
      "loss": 0.5986,
      "step": 3003
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.786710463408726e-05,
      "loss": 0.3391,
      "step": 3004
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.785831280154333e-05,
      "loss": 0.7261,
      "step": 3005
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.784951822493866e-05,
      "loss": 0.7104,
      "step": 3006
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.784072090491073e-05,
      "loss": 0.8635,
      "step": 3007
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.783192084209717e-05,
      "loss": 0.4697,
      "step": 3008
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.782311803713583e-05,
      "loss": 0.7681,
      "step": 3009
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.781431249066478e-05,
      "loss": 1.075,
      "step": 3010
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.780550420332223e-05,
      "loss": 0.4832,
      "step": 3011
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.779669317574666e-05,
      "loss": 0.4796,
      "step": 3012
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.778787940857669e-05,
      "loss": 0.5876,
      "step": 3013
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.777906290245117e-05,
      "loss": 0.5351,
      "step": 3014
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.777024365800916e-05,
      "loss": 0.9089,
      "step": 3015
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.776142167588987e-05,
      "loss": 0.9349,
      "step": 3016
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.775259695673275e-05,
      "loss": 0.5321,
      "step": 3017
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.774376950117745e-05,
      "loss": 0.6695,
      "step": 3018
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.773493930986379e-05,
      "loss": 0.593,
      "step": 3019
    },
    {
      "epoch": 1.25,
      "learning_rate": 8.77261063834318e-05,
      "loss": 0.5263,
      "step": 3020
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.771727072252174e-05,
      "loss": 0.8924,
      "step": 3021
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.7708432327774e-05,
      "loss": 0.7914,
      "step": 3022
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.769959119982922e-05,
      "loss": 0.8208,
      "step": 3023
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.769074733932825e-05,
      "loss": 0.509,
      "step": 3024
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.768190074691208e-05,
      "loss": 0.7717,
      "step": 3025
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.767305142322195e-05,
      "loss": 0.6714,
      "step": 3026
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.766419936889928e-05,
      "loss": 0.6186,
      "step": 3027
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.76553445845857e-05,
      "loss": 0.4032,
      "step": 3028
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.7646487070923e-05,
      "loss": 0.658,
      "step": 3029
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.763762682855322e-05,
      "loss": 0.6327,
      "step": 3030
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.762876385811854e-05,
      "loss": 0.5828,
      "step": 3031
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.761989816026139e-05,
      "loss": 0.3798,
      "step": 3032
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.761102973562439e-05,
      "loss": 0.5674,
      "step": 3033
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.760215858485032e-05,
      "loss": 0.5174,
      "step": 3034
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.759328470858219e-05,
      "loss": 0.527,
      "step": 3035
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.75844081074632e-05,
      "loss": 0.4946,
      "step": 3036
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.757552878213673e-05,
      "loss": 0.3921,
      "step": 3037
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.756664673324642e-05,
      "loss": 0.6489,
      "step": 3038
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.755776196143602e-05,
      "loss": 0.581,
      "step": 3039
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.754887446734953e-05,
      "loss": 0.4955,
      "step": 3040
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.753998425163114e-05,
      "loss": 0.6722,
      "step": 3041
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.753109131492524e-05,
      "loss": 0.5216,
      "step": 3042
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.752219565787641e-05,
      "loss": 0.7912,
      "step": 3043
    },
    {
      "epoch": 1.26,
      "learning_rate": 8.751329728112941e-05,
      "loss": 0.575,
      "step": 3044
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.750439618532923e-05,
      "loss": 0.6503,
      "step": 3045
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.749549237112106e-05,
      "loss": 0.5048,
      "step": 3046
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.748658583915024e-05,
      "loss": 0.6812,
      "step": 3047
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.747767659006235e-05,
      "loss": 0.8109,
      "step": 3048
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.746876462450316e-05,
      "loss": 0.5487,
      "step": 3049
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.745984994311862e-05,
      "loss": 0.7464,
      "step": 3050
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.745093254655488e-05,
      "loss": 0.5101,
      "step": 3051
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.744201243545831e-05,
      "loss": 0.8789,
      "step": 3052
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.743308961047545e-05,
      "loss": 1.2142,
      "step": 3053
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.742416407225305e-05,
      "loss": 0.6557,
      "step": 3054
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.741523582143806e-05,
      "loss": 0.659,
      "step": 3055
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.740630485867763e-05,
      "loss": 0.8096,
      "step": 3056
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.739737118461908e-05,
      "loss": 0.4538,
      "step": 3057
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.738843479990994e-05,
      "loss": 0.7288,
      "step": 3058
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.737949570519796e-05,
      "loss": 0.6427,
      "step": 3059
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.737055390113106e-05,
      "loss": 0.7906,
      "step": 3060
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.736160938835734e-05,
      "loss": 0.4316,
      "step": 3061
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.735266216752516e-05,
      "loss": 1.3685,
      "step": 3062
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.734371223928301e-05,
      "loss": 0.7678,
      "step": 3063
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.73347596042796e-05,
      "loss": 0.817,
      "step": 3064
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.732580426316386e-05,
      "loss": 0.4438,
      "step": 3065
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.731684621658487e-05,
      "loss": 0.7142,
      "step": 3066
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.730788546519193e-05,
      "loss": 0.52,
      "step": 3067
    },
    {
      "epoch": 1.27,
      "learning_rate": 8.729892200963456e-05,
      "loss": 0.777,
      "step": 3068
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.728995585056243e-05,
      "loss": 0.7515,
      "step": 3069
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.728098698862546e-05,
      "loss": 0.7926,
      "step": 3070
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.72720154244737e-05,
      "loss": 0.5979,
      "step": 3071
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.726304115875745e-05,
      "loss": 0.7124,
      "step": 3072
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.725406419212716e-05,
      "loss": 0.5995,
      "step": 3073
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.724508452523354e-05,
      "loss": 0.6536,
      "step": 3074
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.723610215872744e-05,
      "loss": 0.7219,
      "step": 3075
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.722711709325992e-05,
      "loss": 0.7654,
      "step": 3076
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.721812932948224e-05,
      "loss": 0.8529,
      "step": 3077
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.720913886804584e-05,
      "loss": 0.5199,
      "step": 3078
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.72001457096024e-05,
      "loss": 0.4423,
      "step": 3079
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.719114985480376e-05,
      "loss": 0.6297,
      "step": 3080
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.718215130430194e-05,
      "loss": 0.9434,
      "step": 3081
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.71731500587492e-05,
      "loss": 0.7362,
      "step": 3082
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.716414611879795e-05,
      "loss": 0.5666,
      "step": 3083
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.715513948510084e-05,
      "loss": 0.6117,
      "step": 3084
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.714613015831067e-05,
      "loss": 0.8323,
      "step": 3085
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.713711813908047e-05,
      "loss": 0.4333,
      "step": 3086
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.712810342806345e-05,
      "loss": 0.5416,
      "step": 3087
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.711908602591303e-05,
      "loss": 0.9154,
      "step": 3088
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.711006593328277e-05,
      "loss": 0.5595,
      "step": 3089
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.710104315082652e-05,
      "loss": 0.3815,
      "step": 3090
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.709201767919826e-05,
      "loss": 0.8766,
      "step": 3091
    },
    {
      "epoch": 1.28,
      "learning_rate": 8.708298951905215e-05,
      "loss": 0.9601,
      "step": 3092
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.70739586710426e-05,
      "loss": 0.3754,
      "step": 3093
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.706492513582418e-05,
      "loss": 0.5338,
      "step": 3094
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.705588891405166e-05,
      "loss": 0.4864,
      "step": 3095
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.704685000638e-05,
      "loss": 0.7482,
      "step": 3096
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.703780841346439e-05,
      "loss": 0.6639,
      "step": 3097
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.702876413596014e-05,
      "loss": 0.8719,
      "step": 3098
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.701971717452284e-05,
      "loss": 0.7545,
      "step": 3099
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.701066752980823e-05,
      "loss": 0.4333,
      "step": 3100
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.700161520247223e-05,
      "loss": 0.7285,
      "step": 3101
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.6992560193171e-05,
      "loss": 0.5429,
      "step": 3102
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.698350250256082e-05,
      "loss": 0.8412,
      "step": 3103
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.697444213129827e-05,
      "loss": 1.0245,
      "step": 3104
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.696537908004005e-05,
      "loss": 0.4632,
      "step": 3105
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.695631334944306e-05,
      "loss": 0.8918,
      "step": 3106
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.694724494016442e-05,
      "loss": 0.6657,
      "step": 3107
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.69381738528614e-05,
      "loss": 0.3902,
      "step": 3108
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.692910008819155e-05,
      "loss": 0.6609,
      "step": 3109
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.692002364681251e-05,
      "loss": 0.7824,
      "step": 3110
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.691094452938215e-05,
      "loss": 0.6739,
      "step": 3111
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.69018627365586e-05,
      "loss": 0.7141,
      "step": 3112
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.68927782690001e-05,
      "loss": 0.8709,
      "step": 3113
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.688369112736512e-05,
      "loss": 0.6223,
      "step": 3114
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.687460131231232e-05,
      "loss": 0.5446,
      "step": 3115
    },
    {
      "epoch": 1.29,
      "learning_rate": 8.686550882450052e-05,
      "loss": 1.0432,
      "step": 3116
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.685641366458881e-05,
      "loss": 0.691,
      "step": 3117
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.68473158332364e-05,
      "loss": 0.6929,
      "step": 3118
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.683821533110273e-05,
      "loss": 0.6757,
      "step": 3119
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.682911215884744e-05,
      "loss": 0.6454,
      "step": 3120
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.682000631713032e-05,
      "loss": 0.7194,
      "step": 3121
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.681089780661141e-05,
      "loss": 0.4771,
      "step": 3122
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.680178662795091e-05,
      "loss": 0.4445,
      "step": 3123
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.679267278180922e-05,
      "loss": 0.6641,
      "step": 3124
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.678355626884692e-05,
      "loss": 0.5751,
      "step": 3125
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.677443708972481e-05,
      "loss": 0.5542,
      "step": 3126
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.676531524510388e-05,
      "loss": 0.6455,
      "step": 3127
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.675619073564528e-05,
      "loss": 0.7267,
      "step": 3128
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.674706356201038e-05,
      "loss": 0.4535,
      "step": 3129
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.673793372486076e-05,
      "loss": 0.5799,
      "step": 3130
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.672880122485813e-05,
      "loss": 0.4153,
      "step": 3131
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.671966606266449e-05,
      "loss": 0.5272,
      "step": 3132
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.671052823894194e-05,
      "loss": 0.5891,
      "step": 3133
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.670138775435283e-05,
      "loss": 0.7306,
      "step": 3134
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.669224460955967e-05,
      "loss": 0.831,
      "step": 3135
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.668309880522519e-05,
      "loss": 0.3181,
      "step": 3136
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.66739503420123e-05,
      "loss": 0.6292,
      "step": 3137
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.666479922058408e-05,
      "loss": 0.7323,
      "step": 3138
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.665564544160385e-05,
      "loss": 0.8772,
      "step": 3139
    },
    {
      "epoch": 1.3,
      "learning_rate": 8.664648900573509e-05,
      "loss": 0.8734,
      "step": 3140
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.663732991364147e-05,
      "loss": 0.9699,
      "step": 3141
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.662816816598687e-05,
      "loss": 0.7368,
      "step": 3142
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.661900376343537e-05,
      "loss": 0.6269,
      "step": 3143
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.66098367066512e-05,
      "loss": 0.378,
      "step": 3144
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.660066699629883e-05,
      "loss": 0.8149,
      "step": 3145
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.65914946330429e-05,
      "loss": 0.5537,
      "step": 3146
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.658231961754824e-05,
      "loss": 0.8892,
      "step": 3147
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.657314195047988e-05,
      "loss": 0.3387,
      "step": 3148
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.656396163250304e-05,
      "loss": 1.0327,
      "step": 3149
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.655477866428313e-05,
      "loss": 0.3688,
      "step": 3150
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.654559304648575e-05,
      "loss": 0.57,
      "step": 3151
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.65364047797767e-05,
      "loss": 0.4711,
      "step": 3152
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.652721386482196e-05,
      "loss": 0.8423,
      "step": 3153
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.651802030228772e-05,
      "loss": 0.7499,
      "step": 3154
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.650882409284036e-05,
      "loss": 0.9547,
      "step": 3155
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.649962523714643e-05,
      "loss": 0.5235,
      "step": 3156
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.649042373587266e-05,
      "loss": 0.7042,
      "step": 3157
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.648121958968604e-05,
      "loss": 0.7885,
      "step": 3158
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.647201279925368e-05,
      "loss": 0.637,
      "step": 3159
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.646280336524293e-05,
      "loss": 0.6835,
      "step": 3160
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.645359128832129e-05,
      "loss": 0.3959,
      "step": 3161
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.644437656915649e-05,
      "loss": 0.5405,
      "step": 3162
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.643515920841644e-05,
      "loss": 0.4817,
      "step": 3163
    },
    {
      "epoch": 1.31,
      "learning_rate": 8.642593920676923e-05,
      "loss": 0.9184,
      "step": 3164
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.641671656488312e-05,
      "loss": 0.6399,
      "step": 3165
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.640749128342662e-05,
      "loss": 0.6984,
      "step": 3166
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.63982633630684e-05,
      "loss": 0.8532,
      "step": 3167
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.638903280447732e-05,
      "loss": 0.6302,
      "step": 3168
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.63797996083224e-05,
      "loss": 0.5344,
      "step": 3169
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.637056377527295e-05,
      "loss": 0.7058,
      "step": 3170
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.636132530599835e-05,
      "loss": 0.4271,
      "step": 3171
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.635208420116824e-05,
      "loss": 0.6101,
      "step": 3172
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.634284046145245e-05,
      "loss": 0.7125,
      "step": 3173
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.633359408752097e-05,
      "loss": 0.4196,
      "step": 3174
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.632434508004402e-05,
      "loss": 0.7926,
      "step": 3175
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.631509343969197e-05,
      "loss": 0.7598,
      "step": 3176
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.630583916713539e-05,
      "loss": 0.4885,
      "step": 3177
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.62965822630451e-05,
      "loss": 0.7265,
      "step": 3178
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.628732272809202e-05,
      "loss": 1.2404,
      "step": 3179
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.627806056294732e-05,
      "loss": 0.6619,
      "step": 3180
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.626879576828235e-05,
      "loss": 0.8284,
      "step": 3181
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.625952834476863e-05,
      "loss": 0.4846,
      "step": 3182
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.625025829307787e-05,
      "loss": 0.804,
      "step": 3183
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.624098561388203e-05,
      "loss": 0.5511,
      "step": 3184
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.623171030785318e-05,
      "loss": 0.5252,
      "step": 3185
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.622243237566364e-05,
      "loss": 0.8441,
      "step": 3186
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.621315181798586e-05,
      "loss": 0.5413,
      "step": 3187
    },
    {
      "epoch": 1.32,
      "learning_rate": 8.620386863549254e-05,
      "loss": 0.7538,
      "step": 3188
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.619458282885657e-05,
      "loss": 0.8334,
      "step": 3189
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.618529439875099e-05,
      "loss": 0.8376,
      "step": 3190
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.617600334584901e-05,
      "loss": 0.7032,
      "step": 3191
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.616670967082411e-05,
      "loss": 0.413,
      "step": 3192
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.615741337434994e-05,
      "loss": 0.8189,
      "step": 3193
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.614811445710027e-05,
      "loss": 0.6728,
      "step": 3194
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.613881291974912e-05,
      "loss": 0.2318,
      "step": 3195
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.612950876297068e-05,
      "loss": 0.8117,
      "step": 3196
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.612020198743937e-05,
      "loss": 0.4731,
      "step": 3197
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.611089259382973e-05,
      "loss": 0.8467,
      "step": 3198
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.610158058281654e-05,
      "loss": 0.6716,
      "step": 3199
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.609226595507478e-05,
      "loss": 0.6205,
      "step": 3200
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.608294871127956e-05,
      "loss": 0.698,
      "step": 3201
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.607362885210624e-05,
      "loss": 0.3832,
      "step": 3202
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.606430637823033e-05,
      "loss": 0.5112,
      "step": 3203
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.605498129032756e-05,
      "loss": 0.711,
      "step": 3204
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.604565358907383e-05,
      "loss": 0.5574,
      "step": 3205
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.603632327514522e-05,
      "loss": 0.5474,
      "step": 3206
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.602699034921804e-05,
      "loss": 0.6378,
      "step": 3207
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.601765481196874e-05,
      "loss": 0.4306,
      "step": 3208
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.600831666407398e-05,
      "loss": 0.796,
      "step": 3209
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.599897590621065e-05,
      "loss": 0.6821,
      "step": 3210
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.598963253905573e-05,
      "loss": 1.1261,
      "step": 3211
    },
    {
      "epoch": 1.33,
      "learning_rate": 8.598028656328649e-05,
      "loss": 0.7088,
      "step": 3212
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.597093797958035e-05,
      "loss": 0.5355,
      "step": 3213
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.59615867886149e-05,
      "loss": 0.7942,
      "step": 3214
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.595223299106793e-05,
      "loss": 0.7948,
      "step": 3215
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.594287658761746e-05,
      "loss": 0.7393,
      "step": 3216
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.593351757894164e-05,
      "loss": 0.5271,
      "step": 3217
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.592415596571883e-05,
      "loss": 0.6637,
      "step": 3218
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.59147917486276e-05,
      "loss": 0.6895,
      "step": 3219
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.590542492834667e-05,
      "loss": 0.8303,
      "step": 3220
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.589605550555498e-05,
      "loss": 0.6415,
      "step": 3221
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.588668348093165e-05,
      "loss": 0.9732,
      "step": 3222
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.587730885515599e-05,
      "loss": 0.9301,
      "step": 3223
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.586793162890747e-05,
      "loss": 0.6007,
      "step": 3224
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.585855180286581e-05,
      "loss": 1.0513,
      "step": 3225
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.584916937771086e-05,
      "loss": 0.6972,
      "step": 3226
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.583978435412268e-05,
      "loss": 0.8218,
      "step": 3227
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.583039673278154e-05,
      "loss": 0.7059,
      "step": 3228
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.582100651436785e-05,
      "loss": 0.7329,
      "step": 3229
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.581161369956226e-05,
      "loss": 0.7132,
      "step": 3230
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.580221828904556e-05,
      "loss": 0.4245,
      "step": 3231
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.579282028349875e-05,
      "loss": 0.6987,
      "step": 3232
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.578341968360304e-05,
      "loss": 0.8729,
      "step": 3233
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.57740164900398e-05,
      "loss": 0.6457,
      "step": 3234
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.57646107034906e-05,
      "loss": 0.7282,
      "step": 3235
    },
    {
      "epoch": 1.34,
      "learning_rate": 8.57552023246372e-05,
      "loss": 1.1711,
      "step": 3236
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.574579135416152e-05,
      "loss": 0.9391,
      "step": 3237
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.57363777927457e-05,
      "loss": 0.7911,
      "step": 3238
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.572696164107206e-05,
      "loss": 0.4701,
      "step": 3239
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.57175428998231e-05,
      "loss": 0.5416,
      "step": 3240
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.570812156968153e-05,
      "loss": 0.6082,
      "step": 3241
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.569869765133021e-05,
      "loss": 0.5984,
      "step": 3242
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.56892711454522e-05,
      "loss": 0.3592,
      "step": 3243
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.567984205273078e-05,
      "loss": 0.8579,
      "step": 3244
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.56704103738494e-05,
      "loss": 0.7254,
      "step": 3245
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.566097610949165e-05,
      "loss": 1.1932,
      "step": 3246
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.565153926034137e-05,
      "loss": 0.5824,
      "step": 3247
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.564209982708259e-05,
      "loss": 0.5477,
      "step": 3248
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.563265781039946e-05,
      "loss": 0.7158,
      "step": 3249
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.56232132109764e-05,
      "loss": 0.5776,
      "step": 3250
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.561376602949794e-05,
      "loss": 1.0258,
      "step": 3251
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.560431626664886e-05,
      "loss": 0.5765,
      "step": 3252
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.559486392311409e-05,
      "loss": 0.9327,
      "step": 3253
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.558540899957877e-05,
      "loss": 0.6299,
      "step": 3254
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.557595149672819e-05,
      "loss": 1.1034,
      "step": 3255
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.556649141524788e-05,
      "loss": 0.3365,
      "step": 3256
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.555702875582351e-05,
      "loss": 0.9142,
      "step": 3257
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.554756351914098e-05,
      "loss": 0.4949,
      "step": 3258
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.553809570588631e-05,
      "loss": 0.5657,
      "step": 3259
    },
    {
      "epoch": 1.35,
      "learning_rate": 8.55286253167458e-05,
      "loss": 0.5354,
      "step": 3260
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.551915235240584e-05,
      "loss": 0.3674,
      "step": 3261
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.55096768135531e-05,
      "loss": 0.6373,
      "step": 3262
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.550019870087436e-05,
      "loss": 0.7998,
      "step": 3263
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.549071801505661e-05,
      "loss": 0.4506,
      "step": 3264
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.548123475678705e-05,
      "loss": 0.545,
      "step": 3265
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.547174892675305e-05,
      "loss": 0.6554,
      "step": 3266
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.546226052564215e-05,
      "loss": 1.0177,
      "step": 3267
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.545276955414211e-05,
      "loss": 0.9394,
      "step": 3268
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.544327601294084e-05,
      "loss": 0.7056,
      "step": 3269
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.543377990272647e-05,
      "loss": 0.4724,
      "step": 3270
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.54242812241873e-05,
      "loss": 0.5119,
      "step": 3271
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.54147799780118e-05,
      "loss": 0.8033,
      "step": 3272
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.540527616488867e-05,
      "loss": 0.5182,
      "step": 3273
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.539576978550675e-05,
      "loss": 0.5828,
      "step": 3274
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.538626084055506e-05,
      "loss": 0.6372,
      "step": 3275
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.537674933072288e-05,
      "loss": 0.9408,
      "step": 3276
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.536723525669962e-05,
      "loss": 0.527,
      "step": 3277
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.535771861917486e-05,
      "loss": 0.6993,
      "step": 3278
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.534819941883839e-05,
      "loss": 0.7318,
      "step": 3279
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.533867765638019e-05,
      "loss": 0.4166,
      "step": 3280
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.532915333249043e-05,
      "loss": 0.5529,
      "step": 3281
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.531962644785943e-05,
      "loss": 0.7034,
      "step": 3282
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.531009700317777e-05,
      "loss": 0.4977,
      "step": 3283
    },
    {
      "epoch": 1.36,
      "learning_rate": 8.530056499913613e-05,
      "loss": 0.8854,
      "step": 3284
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.52910304364254e-05,
      "loss": 0.4703,
      "step": 3285
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.528149331573669e-05,
      "loss": 0.8083,
      "step": 3286
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.527195363776128e-05,
      "loss": 0.7601,
      "step": 3287
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.52624114031906e-05,
      "loss": 0.2043,
      "step": 3288
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.525286661271632e-05,
      "loss": 1.4866,
      "step": 3289
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.524331926703028e-05,
      "loss": 0.5736,
      "step": 3290
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.523376936682446e-05,
      "loss": 0.7401,
      "step": 3291
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.522421691279108e-05,
      "loss": 0.8569,
      "step": 3292
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.52146619056225e-05,
      "loss": 0.6008,
      "step": 3293
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.520510434601132e-05,
      "loss": 0.5542,
      "step": 3294
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.519554423465029e-05,
      "loss": 0.8091,
      "step": 3295
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.518598157223234e-05,
      "loss": 0.4126,
      "step": 3296
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.517641635945059e-05,
      "loss": 0.7922,
      "step": 3297
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.516684859699837e-05,
      "loss": 0.4583,
      "step": 3298
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.515727828556916e-05,
      "loss": 0.7778,
      "step": 3299
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.514770542585665e-05,
      "loss": 0.9578,
      "step": 3300
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.513813001855469e-05,
      "loss": 0.3112,
      "step": 3301
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.512855206435731e-05,
      "loss": 1.0323,
      "step": 3302
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.511897156395878e-05,
      "loss": 0.5038,
      "step": 3303
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.510938851805352e-05,
      "loss": 0.5269,
      "step": 3304
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.509980292733608e-05,
      "loss": 0.8422,
      "step": 3305
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.50902147925013e-05,
      "loss": 0.5995,
      "step": 3306
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.508062411424414e-05,
      "loss": 0.6204,
      "step": 3307
    },
    {
      "epoch": 1.37,
      "learning_rate": 8.507103089325974e-05,
      "loss": 0.6534,
      "step": 3308
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.506143513024345e-05,
      "loss": 0.6025,
      "step": 3309
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.505183682589079e-05,
      "loss": 0.727,
      "step": 3310
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.504223598089746e-05,
      "loss": 0.6446,
      "step": 3311
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.503263259595939e-05,
      "loss": 0.4854,
      "step": 3312
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.50230266717726e-05,
      "loss": 0.4838,
      "step": 3313
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.501341820903338e-05,
      "loss": 0.8857,
      "step": 3314
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.500380720843817e-05,
      "loss": 0.7091,
      "step": 3315
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.499419367068362e-05,
      "loss": 0.6177,
      "step": 3316
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.49845775964665e-05,
      "loss": 0.3092,
      "step": 3317
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.497495898648384e-05,
      "loss": 0.9649,
      "step": 3318
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.49653378414328e-05,
      "loss": 0.5278,
      "step": 3319
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.495571416201075e-05,
      "loss": 0.7892,
      "step": 3320
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.494608794891525e-05,
      "loss": 0.4504,
      "step": 3321
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.4936459202844e-05,
      "loss": 0.6698,
      "step": 3322
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.492682792449496e-05,
      "loss": 0.5824,
      "step": 3323
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.491719411456619e-05,
      "loss": 0.5409,
      "step": 3324
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.490755777375598e-05,
      "loss": 0.3993,
      "step": 3325
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.489791890276282e-05,
      "loss": 0.525,
      "step": 3326
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.48882775022853e-05,
      "loss": 0.7152,
      "step": 3327
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.487863357302232e-05,
      "loss": 0.8437,
      "step": 3328
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.486898711567284e-05,
      "loss": 0.6184,
      "step": 3329
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.48593381309361e-05,
      "loss": 0.4218,
      "step": 3330
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.484968661951145e-05,
      "loss": 0.5189,
      "step": 3331
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.484003258209847e-05,
      "loss": 0.5923,
      "step": 3332
    },
    {
      "epoch": 1.38,
      "learning_rate": 8.483037601939692e-05,
      "loss": 0.4641,
      "step": 3333
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.48207169321067e-05,
      "loss": 0.7935,
      "step": 3334
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.481105532092794e-05,
      "loss": 0.5581,
      "step": 3335
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.480139118656094e-05,
      "loss": 0.6255,
      "step": 3336
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.479172452970618e-05,
      "loss": 0.7138,
      "step": 3337
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.478205535106431e-05,
      "loss": 0.5001,
      "step": 3338
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.477238365133618e-05,
      "loss": 0.8063,
      "step": 3339
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.476270943122283e-05,
      "loss": 0.4529,
      "step": 3340
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.475303269142546e-05,
      "loss": 1.0186,
      "step": 3341
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.474335343264548e-05,
      "loss": 0.8087,
      "step": 3342
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.473367165558442e-05,
      "loss": 0.7793,
      "step": 3343
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.472398736094408e-05,
      "loss": 0.6461,
      "step": 3344
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.47143005494264e-05,
      "loss": 0.7504,
      "step": 3345
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.470461122173347e-05,
      "loss": 0.6337,
      "step": 3346
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.469491937856763e-05,
      "loss": 0.7789,
      "step": 3347
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.468522502063137e-05,
      "loss": 0.8998,
      "step": 3348
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.467552814862734e-05,
      "loss": 0.5412,
      "step": 3349
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.46658287632584e-05,
      "loss": 0.8744,
      "step": 3350
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.465612686522758e-05,
      "loss": 0.7485,
      "step": 3351
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.464642245523808e-05,
      "loss": 0.6818,
      "step": 3352
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.463671553399335e-05,
      "loss": 0.6623,
      "step": 3353
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.462700610219691e-05,
      "loss": 0.7271,
      "step": 3354
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.461729416055257e-05,
      "loss": 0.7352,
      "step": 3355
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.460757970976425e-05,
      "loss": 0.5916,
      "step": 3356
    },
    {
      "epoch": 1.39,
      "learning_rate": 8.45978627505361e-05,
      "loss": 0.6484,
      "step": 3357
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.45881432835724e-05,
      "loss": 0.7816,
      "step": 3358
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.457842130957765e-05,
      "loss": 0.507,
      "step": 3359
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.456869682925652e-05,
      "loss": 0.3365,
      "step": 3360
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.455896984331388e-05,
      "loss": 0.9297,
      "step": 3361
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.454924035245473e-05,
      "loss": 0.3947,
      "step": 3362
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.453950835738433e-05,
      "loss": 0.6977,
      "step": 3363
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.452977385880805e-05,
      "loss": 0.8943,
      "step": 3364
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.452003685743148e-05,
      "loss": 0.8351,
      "step": 3365
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.451029735396037e-05,
      "loss": 0.6965,
      "step": 3366
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.450055534910068e-05,
      "loss": 0.4865,
      "step": 3367
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.449081084355851e-05,
      "loss": 0.425,
      "step": 3368
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.44810638380402e-05,
      "loss": 0.6807,
      "step": 3369
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.447131433325219e-05,
      "loss": 0.6943,
      "step": 3370
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.44615623299012e-05,
      "loss": 0.7837,
      "step": 3371
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.445180782869404e-05,
      "loss": 1.0621,
      "step": 3372
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.444205083033774e-05,
      "loss": 0.6474,
      "step": 3373
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.443229133553952e-05,
      "loss": 1.0001,
      "step": 3374
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.442252934500679e-05,
      "loss": 0.494,
      "step": 3375
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.441276485944709e-05,
      "loss": 0.555,
      "step": 3376
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.44029978795682e-05,
      "loss": 0.5881,
      "step": 3377
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.439322840607804e-05,
      "loss": 0.4488,
      "step": 3378
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.438345643968472e-05,
      "loss": 1.1612,
      "step": 3379
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.437368198109654e-05,
      "loss": 0.5667,
      "step": 3380
    },
    {
      "epoch": 1.4,
      "learning_rate": 8.436390503102199e-05,
      "loss": 0.8699,
      "step": 3381
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.43541255901697e-05,
      "loss": 0.6364,
      "step": 3382
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.434434365924852e-05,
      "loss": 0.5824,
      "step": 3383
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.43345592389675e-05,
      "loss": 0.5004,
      "step": 3384
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.432477233003577e-05,
      "loss": 0.616,
      "step": 3385
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.431498293316278e-05,
      "loss": 0.4417,
      "step": 3386
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.430519104905803e-05,
      "loss": 0.456,
      "step": 3387
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.429539667843129e-05,
      "loss": 0.53,
      "step": 3388
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.428559982199247e-05,
      "loss": 0.7485,
      "step": 3389
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.427580048045169e-05,
      "loss": 0.4375,
      "step": 3390
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.426599865451918e-05,
      "loss": 0.5376,
      "step": 3391
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.425619434490547e-05,
      "loss": 0.555,
      "step": 3392
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.424638755232115e-05,
      "loss": 1.1132,
      "step": 3393
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.423657827747704e-05,
      "loss": 0.9192,
      "step": 3394
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.422676652108417e-05,
      "loss": 0.4118,
      "step": 3395
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.421695228385368e-05,
      "loss": 0.6047,
      "step": 3396
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.420713556649695e-05,
      "loss": 0.7945,
      "step": 3397
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.419731636972554e-05,
      "loss": 0.8111,
      "step": 3398
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.418749469425114e-05,
      "loss": 0.8151,
      "step": 3399
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.417767054078565e-05,
      "loss": 0.5931,
      "step": 3400
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.416784391004115e-05,
      "loss": 0.6196,
      "step": 3401
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.415801480272991e-05,
      "loss": 0.9351,
      "step": 3402
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.414818321956435e-05,
      "loss": 1.1179,
      "step": 3403
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.413834916125712e-05,
      "loss": 0.6753,
      "step": 3404
    },
    {
      "epoch": 1.41,
      "learning_rate": 8.412851262852096e-05,
      "loss": 0.6409,
      "step": 3405
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.411867362206889e-05,
      "loss": 0.8207,
      "step": 3406
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.410883214261405e-05,
      "loss": 0.3948,
      "step": 3407
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.409898819086978e-05,
      "loss": 0.7451,
      "step": 3408
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.40891417675496e-05,
      "loss": 0.7522,
      "step": 3409
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.407929287336718e-05,
      "loss": 0.7335,
      "step": 3410
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.40694415090364e-05,
      "loss": 0.5865,
      "step": 3411
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.405958767527131e-05,
      "loss": 0.7683,
      "step": 3412
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.404973137278614e-05,
      "loss": 0.8183,
      "step": 3413
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.403987260229531e-05,
      "loss": 0.4979,
      "step": 3414
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.40300113645134e-05,
      "loss": 0.5802,
      "step": 3415
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.402014766015515e-05,
      "loss": 1.1656,
      "step": 3416
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.401028148993554e-05,
      "loss": 0.799,
      "step": 3417
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.40004128545697e-05,
      "loss": 0.5876,
      "step": 3418
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.39905417547729e-05,
      "loss": 1.0019,
      "step": 3419
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.398066819126064e-05,
      "loss": 0.6199,
      "step": 3420
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.397079216474856e-05,
      "loss": 0.7565,
      "step": 3421
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.396091367595253e-05,
      "loss": 0.6357,
      "step": 3422
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.395103272558853e-05,
      "loss": 0.5086,
      "step": 3423
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.394114931437278e-05,
      "loss": 0.4122,
      "step": 3424
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.393126344302166e-05,
      "loss": 0.6457,
      "step": 3425
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.392137511225168e-05,
      "loss": 0.5004,
      "step": 3426
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.391148432277963e-05,
      "loss": 0.8608,
      "step": 3427
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.390159107532238e-05,
      "loss": 0.55,
      "step": 3428
    },
    {
      "epoch": 1.42,
      "learning_rate": 8.3891695370597e-05,
      "loss": 1.0148,
      "step": 3429
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.38817972093208e-05,
      "loss": 0.5648,
      "step": 3430
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.387189659221117e-05,
      "loss": 0.7243,
      "step": 3431
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.386199351998578e-05,
      "loss": 0.7206,
      "step": 3432
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.385208799336242e-05,
      "loss": 0.9803,
      "step": 3433
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.384218001305904e-05,
      "loss": 0.9024,
      "step": 3434
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.38322695797938e-05,
      "loss": 0.7406,
      "step": 3435
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.382235669428504e-05,
      "loss": 0.8102,
      "step": 3436
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.381244135725126e-05,
      "loss": 0.8169,
      "step": 3437
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.380252356941117e-05,
      "loss": 0.7977,
      "step": 3438
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.379260333148362e-05,
      "loss": 0.5951,
      "step": 3439
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.378268064418765e-05,
      "loss": 0.3993,
      "step": 3440
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.377275550824248e-05,
      "loss": 0.8144,
      "step": 3441
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.376282792436752e-05,
      "loss": 0.5509,
      "step": 3442
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.375289789328234e-05,
      "loss": 0.5477,
      "step": 3443
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.374296541570667e-05,
      "loss": 0.71,
      "step": 3444
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.373303049236047e-05,
      "loss": 0.7324,
      "step": 3445
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.372309312396385e-05,
      "loss": 0.7424,
      "step": 3446
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.371315331123705e-05,
      "loss": 0.7957,
      "step": 3447
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.370321105490057e-05,
      "loss": 0.6886,
      "step": 3448
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.369326635567505e-05,
      "loss": 0.6086,
      "step": 3449
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.368331921428127e-05,
      "loss": 0.8546,
      "step": 3450
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.367336963144028e-05,
      "loss": 0.5839,
      "step": 3451
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.36634176078732e-05,
      "loss": 0.6077,
      "step": 3452
    },
    {
      "epoch": 1.43,
      "learning_rate": 8.365346314430139e-05,
      "loss": 0.6468,
      "step": 3453
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.364350624144637e-05,
      "loss": 0.4964,
      "step": 3454
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.363354690002985e-05,
      "loss": 0.8373,
      "step": 3455
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.36235851207737e-05,
      "loss": 0.641,
      "step": 3456
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.361362090439997e-05,
      "loss": 1.2086,
      "step": 3457
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.36036542516309e-05,
      "loss": 0.6599,
      "step": 3458
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.35936851631889e-05,
      "loss": 1.0264,
      "step": 3459
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.358371363979655e-05,
      "loss": 0.6491,
      "step": 3460
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.357373968217661e-05,
      "loss": 0.6773,
      "step": 3461
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.3563763291052e-05,
      "loss": 0.4908,
      "step": 3462
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.355378446714584e-05,
      "loss": 0.8404,
      "step": 3463
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.354380321118144e-05,
      "loss": 1.1353,
      "step": 3464
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.353381952388224e-05,
      "loss": 1.1552,
      "step": 3465
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.352383340597189e-05,
      "loss": 0.3125,
      "step": 3466
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.351384485817422e-05,
      "loss": 1.027,
      "step": 3467
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.350385388121319e-05,
      "loss": 0.6694,
      "step": 3468
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.349386047581301e-05,
      "loss": 0.8558,
      "step": 3469
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.348386464269799e-05,
      "loss": 0.5351,
      "step": 3470
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.347386638259268e-05,
      "loss": 0.8754,
      "step": 3471
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.346386569622175e-05,
      "loss": 0.6928,
      "step": 3472
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.345386258431012e-05,
      "loss": 0.7926,
      "step": 3473
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.34438570475828e-05,
      "loss": 0.4311,
      "step": 3474
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.343384908676501e-05,
      "loss": 0.9837,
      "step": 3475
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.342383870258219e-05,
      "loss": 0.5365,
      "step": 3476
    },
    {
      "epoch": 1.44,
      "learning_rate": 8.341382589575987e-05,
      "loss": 0.7899,
      "step": 3477
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.340381066702384e-05,
      "loss": 0.607,
      "step": 3478
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.339379301709999e-05,
      "loss": 0.5541,
      "step": 3479
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.338377294671447e-05,
      "loss": 0.4157,
      "step": 3480
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.337375045659352e-05,
      "loss": 0.2861,
      "step": 3481
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.336372554746362e-05,
      "loss": 0.3616,
      "step": 3482
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.335369822005138e-05,
      "loss": 0.3769,
      "step": 3483
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.334366847508361e-05,
      "loss": 0.6422,
      "step": 3484
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.33336363132873e-05,
      "loss": 0.7425,
      "step": 3485
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.332360173538959e-05,
      "loss": 0.7231,
      "step": 3486
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.331356474211782e-05,
      "loss": 0.8173,
      "step": 3487
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.330352533419948e-05,
      "loss": 0.6442,
      "step": 3488
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.329348351236229e-05,
      "loss": 0.5305,
      "step": 3489
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.328343927733404e-05,
      "loss": 0.5881,
      "step": 3490
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.327339262984282e-05,
      "loss": 0.7013,
      "step": 3491
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.32633435706168e-05,
      "loss": 0.682,
      "step": 3492
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.325329210038439e-05,
      "loss": 0.5206,
      "step": 3493
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.324323821987411e-05,
      "loss": 0.4576,
      "step": 3494
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.323318192981471e-05,
      "loss": 0.6986,
      "step": 3495
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.32231232309351e-05,
      "loss": 0.513,
      "step": 3496
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.321306212396433e-05,
      "loss": 0.8516,
      "step": 3497
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.320299860963168e-05,
      "loss": 0.7741,
      "step": 3498
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.319293268866657e-05,
      "loss": 0.5296,
      "step": 3499
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.31828643617986e-05,
      "loss": 1.0368,
      "step": 3500
    },
    {
      "epoch": 1.45,
      "learning_rate": 8.317279362975756e-05,
      "loss": 0.7653,
      "step": 3501
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.316272049327336e-05,
      "loss": 0.5245,
      "step": 3502
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.315264495307618e-05,
      "loss": 0.4743,
      "step": 3503
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.314256700989627e-05,
      "loss": 0.5957,
      "step": 3504
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.313248666446413e-05,
      "loss": 0.8004,
      "step": 3505
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.312240391751041e-05,
      "loss": 0.8453,
      "step": 3506
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.31123187697659e-05,
      "loss": 0.8675,
      "step": 3507
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.310223122196164e-05,
      "loss": 0.9443,
      "step": 3508
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.30921412748288e-05,
      "loss": 0.3937,
      "step": 3509
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.308204892909868e-05,
      "loss": 0.4222,
      "step": 3510
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.307195418550282e-05,
      "loss": 0.707,
      "step": 3511
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.306185704477289e-05,
      "loss": 0.6985,
      "step": 3512
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.30517575076408e-05,
      "loss": 0.3821,
      "step": 3513
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.304165557483856e-05,
      "loss": 0.6401,
      "step": 3514
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.303155124709838e-05,
      "loss": 0.6602,
      "step": 3515
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.302144452515267e-05,
      "loss": 0.6674,
      "step": 3516
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.301133540973396e-05,
      "loss": 0.6896,
      "step": 3517
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.3001223901575e-05,
      "loss": 0.4712,
      "step": 3518
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.299111000140869e-05,
      "loss": 0.7878,
      "step": 3519
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.29809937099681e-05,
      "loss": 0.8113,
      "step": 3520
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.297087502798648e-05,
      "loss": 0.6724,
      "step": 3521
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.296075395619728e-05,
      "loss": 0.8658,
      "step": 3522
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.295063049533412e-05,
      "loss": 0.6356,
      "step": 3523
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.29405046461307e-05,
      "loss": 0.5291,
      "step": 3524
    },
    {
      "epoch": 1.46,
      "learning_rate": 8.293037640932102e-05,
      "loss": 0.6159,
      "step": 3525
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.29202457856392e-05,
      "loss": 0.5535,
      "step": 3526
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.29101127758195e-05,
      "loss": 0.9069,
      "step": 3527
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.28999773805964e-05,
      "loss": 0.726,
      "step": 3528
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.288983960070455e-05,
      "loss": 0.5929,
      "step": 3529
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.287969943687874e-05,
      "loss": 0.454,
      "step": 3530
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.286955688985397e-05,
      "loss": 0.4216,
      "step": 3531
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.285941196036538e-05,
      "loss": 0.387,
      "step": 3532
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.284926464914832e-05,
      "loss": 0.6499,
      "step": 3533
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.283911495693828e-05,
      "loss": 0.5085,
      "step": 3534
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.282896288447093e-05,
      "loss": 0.7026,
      "step": 3535
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.281880843248211e-05,
      "loss": 0.7761,
      "step": 3536
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.280865160170785e-05,
      "loss": 0.4374,
      "step": 3537
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.279849239288435e-05,
      "loss": 0.4709,
      "step": 3538
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.278833080674796e-05,
      "loss": 0.5332,
      "step": 3539
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.277816684403521e-05,
      "loss": 0.6808,
      "step": 3540
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.276800050548281e-05,
      "loss": 0.6845,
      "step": 3541
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.275783179182767e-05,
      "loss": 0.8018,
      "step": 3542
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.274766070380682e-05,
      "loss": 0.2865,
      "step": 3543
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.273748724215746e-05,
      "loss": 0.6969,
      "step": 3544
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.272731140761704e-05,
      "loss": 0.5297,
      "step": 3545
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.27171332009231e-05,
      "loss": 0.6246,
      "step": 3546
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.270695262281336e-05,
      "loss": 0.8058,
      "step": 3547
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.269676967402575e-05,
      "loss": 0.5121,
      "step": 3548
    },
    {
      "epoch": 1.47,
      "learning_rate": 8.268658435529836e-05,
      "loss": 0.4855,
      "step": 3549
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.267639666736948e-05,
      "loss": 0.7905,
      "step": 3550
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.266620661097748e-05,
      "loss": 0.4905,
      "step": 3551
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.265601418686097e-05,
      "loss": 0.6343,
      "step": 3552
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.264581939575876e-05,
      "loss": 0.4733,
      "step": 3553
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.263562223840973e-05,
      "loss": 0.5423,
      "step": 3554
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.262542271555306e-05,
      "loss": 0.4961,
      "step": 3555
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.261522082792799e-05,
      "loss": 0.5841,
      "step": 3556
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.2605016576274e-05,
      "loss": 0.6177,
      "step": 3557
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.259480996133071e-05,
      "loss": 0.5387,
      "step": 3558
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.258460098383791e-05,
      "loss": 0.6445,
      "step": 3559
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.257438964453562e-05,
      "loss": 0.8644,
      "step": 3560
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.25641759441639e-05,
      "loss": 0.7498,
      "step": 3561
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.255395988346313e-05,
      "loss": 0.9152,
      "step": 3562
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.254374146317376e-05,
      "loss": 0.5154,
      "step": 3563
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.253352068403646e-05,
      "loss": 0.5738,
      "step": 3564
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.252329754679206e-05,
      "loss": 0.5227,
      "step": 3565
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.251307205218155e-05,
      "loss": 0.669,
      "step": 3566
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.250284420094609e-05,
      "loss": 0.9176,
      "step": 3567
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.249261399382705e-05,
      "loss": 0.7917,
      "step": 3568
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.24823814315659e-05,
      "loss": 0.5218,
      "step": 3569
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.247214651490434e-05,
      "loss": 0.8681,
      "step": 3570
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.246190924458423e-05,
      "loss": 0.7686,
      "step": 3571
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.245166962134759e-05,
      "loss": 0.7706,
      "step": 3572
    },
    {
      "epoch": 1.48,
      "learning_rate": 8.244142764593657e-05,
      "loss": 0.6201,
      "step": 3573
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.243118331909359e-05,
      "loss": 0.6049,
      "step": 3574
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.242093664156116e-05,
      "loss": 0.7236,
      "step": 3575
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.2410687614082e-05,
      "loss": 0.4723,
      "step": 3576
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.240043623739893e-05,
      "loss": 1.12,
      "step": 3577
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.239018251225505e-05,
      "loss": 0.3479,
      "step": 3578
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.237992643939356e-05,
      "loss": 0.71,
      "step": 3579
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.236966801955784e-05,
      "loss": 0.5712,
      "step": 3580
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.235940725349144e-05,
      "loss": 0.2618,
      "step": 3581
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.23491441419381e-05,
      "loss": 0.4522,
      "step": 3582
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.23388786856417e-05,
      "loss": 0.622,
      "step": 3583
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.23286108853463e-05,
      "loss": 0.3591,
      "step": 3584
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.231834074179617e-05,
      "loss": 0.7564,
      "step": 3585
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.230806825573567e-05,
      "loss": 0.9832,
      "step": 3586
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.229779342790938e-05,
      "loss": 0.5808,
      "step": 3587
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.228751625906208e-05,
      "loss": 0.4874,
      "step": 3588
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.227723674993866e-05,
      "loss": 0.5502,
      "step": 3589
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.226695490128419e-05,
      "loss": 0.4861,
      "step": 3590
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.225667071384395e-05,
      "loss": 0.5848,
      "step": 3591
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.224638418836334e-05,
      "loss": 0.7452,
      "step": 3592
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.223609532558798e-05,
      "loss": 0.5923,
      "step": 3593
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.222580412626359e-05,
      "loss": 0.825,
      "step": 3594
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.221551059113613e-05,
      "loss": 0.49,
      "step": 3595
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.22052147209517e-05,
      "loss": 0.245,
      "step": 3596
    },
    {
      "epoch": 1.49,
      "learning_rate": 8.219491651645654e-05,
      "loss": 0.7347,
      "step": 3597
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.218461597839714e-05,
      "loss": 0.6687,
      "step": 3598
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.217431310752006e-05,
      "loss": 0.303,
      "step": 3599
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.21640079045721e-05,
      "loss": 0.6912,
      "step": 3600
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.215370037030022e-05,
      "loss": 0.6513,
      "step": 3601
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.21433905054515e-05,
      "loss": 0.5663,
      "step": 3602
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.213307831077324e-05,
      "loss": 0.5891,
      "step": 3603
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.212276378701288e-05,
      "loss": 0.3617,
      "step": 3604
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.211244693491806e-05,
      "loss": 0.5751,
      "step": 3605
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.210212775523657e-05,
      "loss": 0.3709,
      "step": 3606
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.209180624871635e-05,
      "loss": 0.7106,
      "step": 3607
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.208148241610555e-05,
      "loss": 0.5401,
      "step": 3608
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.207115625815245e-05,
      "loss": 0.682,
      "step": 3609
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.206082777560553e-05,
      "loss": 0.5245,
      "step": 3610
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.205049696921338e-05,
      "loss": 0.8522,
      "step": 3611
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.204016383972486e-05,
      "loss": 0.8787,
      "step": 3612
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.202982838788892e-05,
      "loss": 0.6212,
      "step": 3613
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.201949061445469e-05,
      "loss": 0.6707,
      "step": 3614
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.200915052017146e-05,
      "loss": 0.9006,
      "step": 3615
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.199880810578873e-05,
      "loss": 0.7573,
      "step": 3616
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.198846337205614e-05,
      "loss": 0.4842,
      "step": 3617
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.19781163197235e-05,
      "loss": 0.6061,
      "step": 3618
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.196776694954078e-05,
      "loss": 0.7012,
      "step": 3619
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.195741526225813e-05,
      "loss": 0.4958,
      "step": 3620
    },
    {
      "epoch": 1.5,
      "learning_rate": 8.194706125862588e-05,
      "loss": 0.5677,
      "step": 3621
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.193670493939449e-05,
      "loss": 0.5495,
      "step": 3622
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.192634630531461e-05,
      "loss": 0.5425,
      "step": 3623
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.191598535713709e-05,
      "loss": 0.6193,
      "step": 3624
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.19056220956129e-05,
      "loss": 0.7809,
      "step": 3625
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.189525652149316e-05,
      "loss": 0.3507,
      "step": 3626
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.188488863552926e-05,
      "loss": 0.9637,
      "step": 3627
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.187451843847262e-05,
      "loss": 0.6286,
      "step": 3628
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.186414593107495e-05,
      "loss": 0.8659,
      "step": 3629
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.185377111408805e-05,
      "loss": 0.9795,
      "step": 3630
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.184339398826388e-05,
      "loss": 0.339,
      "step": 3631
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.183301455435466e-05,
      "loss": 0.7954,
      "step": 3632
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.182263281311269e-05,
      "loss": 0.6457,
      "step": 3633
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.181224876529045e-05,
      "loss": 0.7075,
      "step": 3634
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.180186241164063e-05,
      "loss": 0.7838,
      "step": 3635
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.179147375291601e-05,
      "loss": 0.3329,
      "step": 3636
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.178108278986963e-05,
      "loss": 0.8434,
      "step": 3637
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.177068952325464e-05,
      "loss": 0.6921,
      "step": 3638
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.176029395382434e-05,
      "loss": 1.0928,
      "step": 3639
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.174989608233228e-05,
      "loss": 0.4386,
      "step": 3640
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.173949590953206e-05,
      "loss": 0.7706,
      "step": 3641
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.172909343617758e-05,
      "loss": 0.7544,
      "step": 3642
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.171868866302279e-05,
      "loss": 0.5264,
      "step": 3643
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.170828159082184e-05,
      "loss": 0.5586,
      "step": 3644
    },
    {
      "epoch": 1.51,
      "learning_rate": 8.16978722203291e-05,
      "loss": 0.7558,
      "step": 3645
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.168746055229902e-05,
      "loss": 0.8063,
      "step": 3646
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.167704658748632e-05,
      "loss": 0.6458,
      "step": 3647
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.166663032664576e-05,
      "loss": 0.635,
      "step": 3648
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.16562117705324e-05,
      "loss": 0.6988,
      "step": 3649
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.164579091990137e-05,
      "loss": 0.8767,
      "step": 3650
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.163536777550799e-05,
      "loss": 0.4677,
      "step": 3651
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.16249423381078e-05,
      "loss": 0.3605,
      "step": 3652
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.16145146084564e-05,
      "loss": 0.6672,
      "step": 3653
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.160408458730966e-05,
      "loss": 0.8763,
      "step": 3654
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.159365227542355e-05,
      "loss": 0.5775,
      "step": 3655
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.158321767355424e-05,
      "loss": 0.5708,
      "step": 3656
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.157278078245804e-05,
      "loss": 0.6873,
      "step": 3657
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.156234160289144e-05,
      "loss": 0.5237,
      "step": 3658
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.155190013561114e-05,
      "loss": 0.7178,
      "step": 3659
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.154145638137392e-05,
      "loss": 0.4802,
      "step": 3660
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.153101034093678e-05,
      "loss": 0.5003,
      "step": 3661
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.152056201505685e-05,
      "loss": 0.5581,
      "step": 3662
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.151011140449149e-05,
      "loss": 0.5781,
      "step": 3663
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.149965850999817e-05,
      "loss": 0.6547,
      "step": 3664
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.148920333233454e-05,
      "loss": 0.4867,
      "step": 3665
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.14787458722584e-05,
      "loss": 0.7489,
      "step": 3666
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.146828613052776e-05,
      "loss": 0.6506,
      "step": 3667
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.145782410790075e-05,
      "loss": 0.5104,
      "step": 3668
    },
    {
      "epoch": 1.52,
      "learning_rate": 8.144735980513569e-05,
      "loss": 0.6142,
      "step": 3669
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.143689322299104e-05,
      "loss": 0.496,
      "step": 3670
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.142642436222546e-05,
      "loss": 0.6737,
      "step": 3671
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.141595322359776e-05,
      "loss": 0.7788,
      "step": 3672
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.14054798078669e-05,
      "loss": 0.5895,
      "step": 3673
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.139500411579202e-05,
      "loss": 0.6682,
      "step": 3674
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.138452614813243e-05,
      "loss": 0.5527,
      "step": 3675
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.13740459056476e-05,
      "loss": 0.8437,
      "step": 3676
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.136356338909716e-05,
      "loss": 0.6901,
      "step": 3677
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.13530785992409e-05,
      "loss": 0.8891,
      "step": 3678
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.134259153683881e-05,
      "loss": 1.094,
      "step": 3679
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.133210220265098e-05,
      "loss": 0.6165,
      "step": 3680
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.132161059743771e-05,
      "loss": 0.6511,
      "step": 3681
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.131111672195948e-05,
      "loss": 0.3794,
      "step": 3682
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.130062057697689e-05,
      "loss": 0.6819,
      "step": 3683
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.129012216325074e-05,
      "loss": 0.543,
      "step": 3684
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.127962148154196e-05,
      "loss": 0.602,
      "step": 3685
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.126911853261167e-05,
      "loss": 0.852,
      "step": 3686
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.125861331722117e-05,
      "loss": 0.4625,
      "step": 3687
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.12481058361319e-05,
      "loss": 1.0846,
      "step": 3688
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.123759609010543e-05,
      "loss": 0.9511,
      "step": 3689
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.122708407990356e-05,
      "loss": 1.1614,
      "step": 3690
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.121656980628823e-05,
      "loss": 0.5518,
      "step": 3691
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.120605327002153e-05,
      "loss": 0.5514,
      "step": 3692
    },
    {
      "epoch": 1.53,
      "learning_rate": 8.119553447186572e-05,
      "loss": 0.5156,
      "step": 3693
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.118501341258325e-05,
      "loss": 0.6271,
      "step": 3694
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.117449009293668e-05,
      "loss": 0.6075,
      "step": 3695
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.11639645136888e-05,
      "loss": 0.5932,
      "step": 3696
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.115343667560248e-05,
      "loss": 0.5982,
      "step": 3697
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.114290657944083e-05,
      "loss": 0.638,
      "step": 3698
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.113237422596712e-05,
      "loss": 0.9257,
      "step": 3699
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.112183961594473e-05,
      "loss": 0.4827,
      "step": 3700
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.111130275013725e-05,
      "loss": 0.5441,
      "step": 3701
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.11007636293084e-05,
      "loss": 0.4346,
      "step": 3702
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.109022225422209e-05,
      "loss": 0.3504,
      "step": 3703
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.107967862564239e-05,
      "loss": 0.5198,
      "step": 3704
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.10691327443335e-05,
      "loss": 0.7811,
      "step": 3705
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.105858461105983e-05,
      "loss": 0.6264,
      "step": 3706
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.104803422658596e-05,
      "loss": 0.5685,
      "step": 3707
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.103748159167659e-05,
      "loss": 0.8867,
      "step": 3708
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.102692670709658e-05,
      "loss": 0.5458,
      "step": 3709
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.101636957361098e-05,
      "loss": 0.6171,
      "step": 3710
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.1005810191985e-05,
      "loss": 0.5116,
      "step": 3711
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.099524856298402e-05,
      "loss": 0.9722,
      "step": 3712
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.098468468737355e-05,
      "loss": 1.0813,
      "step": 3713
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.09741185659193e-05,
      "loss": 0.5357,
      "step": 3714
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.096355019938714e-05,
      "loss": 0.5791,
      "step": 3715
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.095297958854306e-05,
      "loss": 0.5149,
      "step": 3716
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.094240673415326e-05,
      "loss": 0.6161,
      "step": 3717
    },
    {
      "epoch": 1.54,
      "learning_rate": 8.09318316369841e-05,
      "loss": 0.5919,
      "step": 3718
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.092125429780206e-05,
      "loss": 0.7233,
      "step": 3719
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.091067471737383e-05,
      "loss": 0.7391,
      "step": 3720
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.090009289646622e-05,
      "loss": 1.1962,
      "step": 3721
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.088950883584628e-05,
      "loss": 1.4093,
      "step": 3722
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.08789225362811e-05,
      "loss": 0.6219,
      "step": 3723
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.086833399853804e-05,
      "loss": 0.7507,
      "step": 3724
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.085774322338456e-05,
      "loss": 0.7198,
      "step": 3725
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.084715021158833e-05,
      "loss": 0.8201,
      "step": 3726
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.083655496391716e-05,
      "loss": 0.4755,
      "step": 3727
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.082595748113897e-05,
      "loss": 0.6964,
      "step": 3728
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.081535776402193e-05,
      "loss": 0.5118,
      "step": 3729
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.080475581333436e-05,
      "loss": 0.5049,
      "step": 3730
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.079415162984465e-05,
      "loss": 0.5036,
      "step": 3731
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.078354521432146e-05,
      "loss": 0.4378,
      "step": 3732
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.077293656753355e-05,
      "loss": 0.8012,
      "step": 3733
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.076232569024986e-05,
      "loss": 0.6259,
      "step": 3734
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.075171258323951e-05,
      "loss": 0.6522,
      "step": 3735
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.074109724727175e-05,
      "loss": 0.7483,
      "step": 3736
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.073047968311602e-05,
      "loss": 0.7925,
      "step": 3737
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.071985989154187e-05,
      "loss": 0.6246,
      "step": 3738
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.070923787331908e-05,
      "loss": 0.4962,
      "step": 3739
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.069861362921754e-05,
      "loss": 0.5508,
      "step": 3740
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.068798716000732e-05,
      "loss": 0.5382,
      "step": 3741
    },
    {
      "epoch": 1.55,
      "learning_rate": 8.067735846645869e-05,
      "loss": 0.7472,
      "step": 3742
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.0666727549342e-05,
      "loss": 0.4889,
      "step": 3743
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.065609440942783e-05,
      "loss": 0.5694,
      "step": 3744
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.064545904748687e-05,
      "loss": 0.4598,
      "step": 3745
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.063482146429002e-05,
      "loss": 1.0707,
      "step": 3746
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.06241816606083e-05,
      "loss": 0.7081,
      "step": 3747
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.061353963721292e-05,
      "loss": 0.736,
      "step": 3748
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.060289539487523e-05,
      "loss": 0.8675,
      "step": 3749
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.059224893436676e-05,
      "loss": 0.6119,
      "step": 3750
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.058160025645918e-05,
      "loss": 0.95,
      "step": 3751
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.057094936192437e-05,
      "loss": 0.4953,
      "step": 3752
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.056029625153428e-05,
      "loss": 1.232,
      "step": 3753
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.054964092606108e-05,
      "loss": 1.3282,
      "step": 3754
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.05389833862771e-05,
      "loss": 0.7732,
      "step": 3755
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.052832363295485e-05,
      "loss": 0.5457,
      "step": 3756
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.051766166686696e-05,
      "loss": 0.491,
      "step": 3757
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.050699748878621e-05,
      "loss": 0.6393,
      "step": 3758
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.04963310994856e-05,
      "loss": 0.3901,
      "step": 3759
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.048566249973824e-05,
      "loss": 0.7229,
      "step": 3760
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.04749916903174e-05,
      "loss": 0.891,
      "step": 3761
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.046431867199655e-05,
      "loss": 0.4879,
      "step": 3762
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.045364344554928e-05,
      "loss": 0.6749,
      "step": 3763
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.044296601174938e-05,
      "loss": 1.3163,
      "step": 3764
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.043228637137074e-05,
      "loss": 0.7265,
      "step": 3765
    },
    {
      "epoch": 1.56,
      "learning_rate": 8.042160452518747e-05,
      "loss": 0.6516,
      "step": 3766
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.04109204739738e-05,
      "loss": 0.48,
      "step": 3767
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.040023421850417e-05,
      "loss": 0.6124,
      "step": 3768
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.03895457595531e-05,
      "loss": 0.6125,
      "step": 3769
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.037885509789536e-05,
      "loss": 0.4499,
      "step": 3770
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.036816223430578e-05,
      "loss": 0.4927,
      "step": 3771
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.035746716955946e-05,
      "loss": 0.7469,
      "step": 3772
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.034676990443157e-05,
      "loss": 0.4237,
      "step": 3773
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.033607043969746e-05,
      "loss": 0.4935,
      "step": 3774
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.03253687761327e-05,
      "loss": 0.9916,
      "step": 3775
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.031466491451294e-05,
      "loss": 0.3938,
      "step": 3776
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.030395885561402e-05,
      "loss": 0.5185,
      "step": 3777
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.029325060021195e-05,
      "loss": 0.3877,
      "step": 3778
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.028254014908289e-05,
      "loss": 0.8204,
      "step": 3779
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.027182750300318e-05,
      "loss": 0.5934,
      "step": 3780
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.026111266274924e-05,
      "loss": 0.9171,
      "step": 3781
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.025039562909775e-05,
      "loss": 0.4393,
      "step": 3782
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.023967640282553e-05,
      "loss": 0.7933,
      "step": 3783
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.022895498470945e-05,
      "loss": 0.6148,
      "step": 3784
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.021823137552673e-05,
      "loss": 0.6839,
      "step": 3785
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.020750557605456e-05,
      "loss": 0.6091,
      "step": 3786
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.019677758707042e-05,
      "loss": 0.5404,
      "step": 3787
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.018604740935186e-05,
      "loss": 0.7432,
      "step": 3788
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.017531504367666e-05,
      "loss": 0.6422,
      "step": 3789
    },
    {
      "epoch": 1.57,
      "learning_rate": 8.016458049082271e-05,
      "loss": 0.506,
      "step": 3790
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.015384375156812e-05,
      "loss": 0.6151,
      "step": 3791
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.014310482669104e-05,
      "loss": 0.4447,
      "step": 3792
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.013236371696991e-05,
      "loss": 0.7061,
      "step": 3793
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.012162042318325e-05,
      "loss": 0.6098,
      "step": 3794
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.011087494610976e-05,
      "loss": 0.8992,
      "step": 3795
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.010012728652829e-05,
      "loss": 0.8416,
      "step": 3796
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.008937744521789e-05,
      "loss": 0.5948,
      "step": 3797
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.00786254229577e-05,
      "loss": 0.4529,
      "step": 3798
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.006787122052705e-05,
      "loss": 0.8282,
      "step": 3799
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.005711483870546e-05,
      "loss": 0.4125,
      "step": 3800
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.004635627827255e-05,
      "loss": 0.4412,
      "step": 3801
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.003559554000814e-05,
      "loss": 0.4608,
      "step": 3802
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.002483262469221e-05,
      "loss": 0.6871,
      "step": 3803
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.001406753310484e-05,
      "loss": 0.7676,
      "step": 3804
    },
    {
      "epoch": 1.58,
      "learning_rate": 8.000330026602633e-05,
      "loss": 0.5188,
      "step": 3805
    },
    {
      "epoch": 1.58,
      "learning_rate": 7.999253082423712e-05,
      "loss": 0.6437,
      "step": 3806
    },
    {
      "epoch": 1.58,
      "learning_rate": 7.998175920851781e-05,
      "loss": 0.6257,
      "step": 3807
    },
    {
      "epoch": 1.58,
      "learning_rate": 7.997098541964914e-05,
      "loss": 1.2157,
      "step": 3808
    },
    {
      "epoch": 1.58,
      "learning_rate": 7.996020945841204e-05,
      "loss": 0.6628,
      "step": 3809
    },
    {
      "epoch": 1.58,
      "learning_rate": 7.994943132558755e-05,
      "loss": 0.3606,
      "step": 3810
    },
    {
      "epoch": 1.58,
      "learning_rate": 7.993865102195692e-05,
      "loss": 0.6319,
      "step": 3811
    },
    {
      "epoch": 1.58,
      "learning_rate": 7.992786854830153e-05,
      "loss": 0.6536,
      "step": 3812
    },
    {
      "epoch": 1.58,
      "learning_rate": 7.991708390540289e-05,
      "loss": 0.5354,
      "step": 3813
    },
    {
      "epoch": 1.58,
      "learning_rate": 7.990629709404272e-05,
      "loss": 0.7063,
      "step": 3814
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.989550811500287e-05,
      "loss": 0.4806,
      "step": 3815
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.988471696906536e-05,
      "loss": 0.5504,
      "step": 3816
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.987392365701234e-05,
      "loss": 0.5454,
      "step": 3817
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.986312817962615e-05,
      "loss": 0.6618,
      "step": 3818
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.985233053768925e-05,
      "loss": 0.8394,
      "step": 3819
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.984153073198433e-05,
      "loss": 0.6478,
      "step": 3820
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.983072876329411e-05,
      "loss": 0.8256,
      "step": 3821
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.981992463240163e-05,
      "loss": 0.6168,
      "step": 3822
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.98091183400899e-05,
      "loss": 0.4113,
      "step": 3823
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.979830988714227e-05,
      "loss": 0.6319,
      "step": 3824
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.978749927434213e-05,
      "loss": 0.5714,
      "step": 3825
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.977668650247305e-05,
      "loss": 0.7231,
      "step": 3826
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.976587157231877e-05,
      "loss": 0.4326,
      "step": 3827
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.975505448466321e-05,
      "loss": 0.9593,
      "step": 3828
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.974423524029038e-05,
      "loss": 0.5327,
      "step": 3829
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.973341383998449e-05,
      "loss": 0.721,
      "step": 3830
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.972259028452993e-05,
      "loss": 0.7212,
      "step": 3831
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.971176457471118e-05,
      "loss": 0.9718,
      "step": 3832
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.970093671131294e-05,
      "loss": 0.7762,
      "step": 3833
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.969010669512003e-05,
      "loss": 0.7498,
      "step": 3834
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.967927452691744e-05,
      "loss": 0.4987,
      "step": 3835
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.966844020749032e-05,
      "loss": 0.6931,
      "step": 3836
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.965760373762393e-05,
      "loss": 0.7584,
      "step": 3837
    },
    {
      "epoch": 1.59,
      "learning_rate": 7.964676511810374e-05,
      "loss": 0.8458,
      "step": 3838
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.963592434971539e-05,
      "loss": 0.7474,
      "step": 3839
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.962508143324461e-05,
      "loss": 0.9817,
      "step": 3840
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.961423636947734e-05,
      "loss": 0.9682,
      "step": 3841
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.960338915919962e-05,
      "loss": 0.4354,
      "step": 3842
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.959253980319774e-05,
      "loss": 0.5571,
      "step": 3843
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.958168830225805e-05,
      "loss": 0.6542,
      "step": 3844
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.95708346571671e-05,
      "loss": 0.5491,
      "step": 3845
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.955997886871158e-05,
      "loss": 1.1503,
      "step": 3846
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.954912093767836e-05,
      "loss": 0.6549,
      "step": 3847
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.953826086485443e-05,
      "loss": 0.7093,
      "step": 3848
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.952739865102697e-05,
      "loss": 1.0346,
      "step": 3849
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.951653429698329e-05,
      "loss": 0.8163,
      "step": 3850
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.950566780351086e-05,
      "loss": 0.5951,
      "step": 3851
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.949479917139732e-05,
      "loss": 0.5047,
      "step": 3852
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.948392840143047e-05,
      "loss": 0.6023,
      "step": 3853
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.947305549439821e-05,
      "loss": 1.1775,
      "step": 3854
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.946218045108866e-05,
      "loss": 0.4759,
      "step": 3855
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.945130327229006e-05,
      "loss": 0.6631,
      "step": 3856
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.944042395879082e-05,
      "loss": 0.6942,
      "step": 3857
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.942954251137949e-05,
      "loss": 0.5952,
      "step": 3858
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.941865893084479e-05,
      "loss": 0.5448,
      "step": 3859
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.940777321797559e-05,
      "loss": 0.7106,
      "step": 3860
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.939688537356091e-05,
      "loss": 0.8162,
      "step": 3861
    },
    {
      "epoch": 1.6,
      "learning_rate": 7.938599539838991e-05,
      "loss": 0.6814,
      "step": 3862
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.937510329325195e-05,
      "loss": 0.8837,
      "step": 3863
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.936420905893651e-05,
      "loss": 0.4059,
      "step": 3864
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.93533126962332e-05,
      "loss": 0.8622,
      "step": 3865
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.934241420593187e-05,
      "loss": 0.7128,
      "step": 3866
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.933151358882242e-05,
      "loss": 0.6478,
      "step": 3867
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.932061084569496e-05,
      "loss": 0.3858,
      "step": 3868
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.930970597733976e-05,
      "loss": 1.041,
      "step": 3869
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.929879898454724e-05,
      "loss": 0.5625,
      "step": 3870
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.928788986810794e-05,
      "loss": 0.6197,
      "step": 3871
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.927697862881258e-05,
      "loss": 0.454,
      "step": 3872
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.926606526745206e-05,
      "loss": 0.5058,
      "step": 3873
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.925514978481739e-05,
      "loss": 1.1117,
      "step": 3874
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.924423218169973e-05,
      "loss": 0.4728,
      "step": 3875
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.923331245889045e-05,
      "loss": 0.6029,
      "step": 3876
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.9222390617181e-05,
      "loss": 0.421,
      "step": 3877
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.921146665736306e-05,
      "loss": 1.2189,
      "step": 3878
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.920054058022841e-05,
      "loss": 0.4755,
      "step": 3879
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.918961238656898e-05,
      "loss": 0.6786,
      "step": 3880
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.917868207717689e-05,
      "loss": 0.6165,
      "step": 3881
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.91677496528444e-05,
      "loss": 0.4454,
      "step": 3882
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.91568151143639e-05,
      "loss": 0.8377,
      "step": 3883
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.914587846252796e-05,
      "loss": 0.287,
      "step": 3884
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.913493969812928e-05,
      "loss": 0.5605,
      "step": 3885
    },
    {
      "epoch": 1.61,
      "learning_rate": 7.912399882196078e-05,
      "loss": 0.6962,
      "step": 3886
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.911305583481542e-05,
      "loss": 0.9269,
      "step": 3887
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.91021107374864e-05,
      "loss": 0.6548,
      "step": 3888
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.909116353076704e-05,
      "loss": 0.9184,
      "step": 3889
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.908021421545083e-05,
      "loss": 0.6238,
      "step": 3890
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.90692627923314e-05,
      "loss": 0.6681,
      "step": 3891
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.905830926220252e-05,
      "loss": 0.6527,
      "step": 3892
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.904735362585815e-05,
      "loss": 0.6204,
      "step": 3893
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.903639588409235e-05,
      "loss": 0.6946,
      "step": 3894
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.90254360376994e-05,
      "loss": 1.0269,
      "step": 3895
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.901447408747369e-05,
      "loss": 0.5473,
      "step": 3896
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.900351003420973e-05,
      "loss": 0.6873,
      "step": 3897
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.899254387870227e-05,
      "loss": 0.6104,
      "step": 3898
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.898157562174613e-05,
      "loss": 1.0757,
      "step": 3899
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.897060526413633e-05,
      "loss": 0.2846,
      "step": 3900
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.895963280666802e-05,
      "loss": 0.8391,
      "step": 3901
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.89486582501365e-05,
      "loss": 0.6947,
      "step": 3902
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.893768159533728e-05,
      "loss": 0.3704,
      "step": 3903
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.892670284306592e-05,
      "loss": 0.7338,
      "step": 3904
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.89157219941182e-05,
      "loss": 0.5817,
      "step": 3905
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.890473904929006e-05,
      "loss": 0.4969,
      "step": 3906
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.889375400937755e-05,
      "loss": 0.2576,
      "step": 3907
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.888276687517688e-05,
      "loss": 0.728,
      "step": 3908
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.887177764748446e-05,
      "loss": 0.4254,
      "step": 3909
    },
    {
      "epoch": 1.62,
      "learning_rate": 7.886078632709678e-05,
      "loss": 0.7231,
      "step": 3910
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.884979291481054e-05,
      "loss": 0.6693,
      "step": 3911
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.883879741142257e-05,
      "loss": 0.9232,
      "step": 3912
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.882779981772982e-05,
      "loss": 0.7424,
      "step": 3913
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.881680013452946e-05,
      "loss": 0.8984,
      "step": 3914
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.880579836261876e-05,
      "loss": 0.5102,
      "step": 3915
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.879479450279514e-05,
      "loss": 0.5892,
      "step": 3916
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.87837885558562e-05,
      "loss": 0.8653,
      "step": 3917
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.877278052259966e-05,
      "loss": 0.4546,
      "step": 3918
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.876177040382345e-05,
      "loss": 0.486,
      "step": 3919
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.875075820032555e-05,
      "loss": 0.6542,
      "step": 3920
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.873974391290421e-05,
      "loss": 0.4044,
      "step": 3921
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.872872754235773e-05,
      "loss": 0.7317,
      "step": 3922
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.871770908948464e-05,
      "loss": 0.4913,
      "step": 3923
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.870668855508354e-05,
      "loss": 0.6801,
      "step": 3924
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.869566593995325e-05,
      "loss": 0.5979,
      "step": 3925
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.868464124489271e-05,
      "loss": 0.7983,
      "step": 3926
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.867361447070103e-05,
      "loss": 0.6342,
      "step": 3927
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.866258561817743e-05,
      "loss": 0.6842,
      "step": 3928
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.865155468812134e-05,
      "loss": 0.9112,
      "step": 3929
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.864052168133227e-05,
      "loss": 0.5971,
      "step": 3930
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.862948659860994e-05,
      "loss": 0.9699,
      "step": 3931
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.861844944075421e-05,
      "loss": 0.5534,
      "step": 3932
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.860741020856506e-05,
      "loss": 0.6565,
      "step": 3933
    },
    {
      "epoch": 1.63,
      "learning_rate": 7.859636890284263e-05,
      "loss": 0.6467,
      "step": 3934
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.858532552438727e-05,
      "loss": 0.4555,
      "step": 3935
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.857428007399938e-05,
      "loss": 0.6454,
      "step": 3936
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.856323255247958e-05,
      "loss": 0.8258,
      "step": 3937
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.855218296062861e-05,
      "loss": 0.9237,
      "step": 3938
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.85411312992474e-05,
      "loss": 0.5212,
      "step": 3939
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.853007756913697e-05,
      "loss": 0.5506,
      "step": 3940
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.851902177109853e-05,
      "loss": 0.5314,
      "step": 3941
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.850796390593342e-05,
      "loss": 0.708,
      "step": 3942
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.849690397444317e-05,
      "loss": 0.4193,
      "step": 3943
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.84858419774294e-05,
      "loss": 0.7407,
      "step": 3944
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.847477791569395e-05,
      "loss": 0.393,
      "step": 3945
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.84637117900387e-05,
      "loss": 0.6119,
      "step": 3946
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.845264360126583e-05,
      "loss": 0.841,
      "step": 3947
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.844157335017752e-05,
      "loss": 0.3554,
      "step": 3948
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.843050103757621e-05,
      "loss": 0.8644,
      "step": 3949
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.841942666426443e-05,
      "loss": 0.8821,
      "step": 3950
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.840835023104489e-05,
      "loss": 0.7007,
      "step": 3951
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.839727173872042e-05,
      "loss": 0.4453,
      "step": 3952
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.838619118809402e-05,
      "loss": 0.2931,
      "step": 3953
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.837510857996888e-05,
      "loss": 0.6966,
      "step": 3954
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.83640239151482e-05,
      "loss": 0.6334,
      "step": 3955
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.83529371944355e-05,
      "loss": 0.7421,
      "step": 3956
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.834184841863434e-05,
      "loss": 0.5813,
      "step": 3957
    },
    {
      "epoch": 1.64,
      "learning_rate": 7.833075758854848e-05,
      "loss": 0.4765,
      "step": 3958
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.83196647049818e-05,
      "loss": 0.5439,
      "step": 3959
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.830856976873832e-05,
      "loss": 0.7128,
      "step": 3960
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.829747278062225e-05,
      "loss": 0.51,
      "step": 3961
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.828637374143792e-05,
      "loss": 0.3618,
      "step": 3962
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.827527265198983e-05,
      "loss": 1.0191,
      "step": 3963
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.826416951308259e-05,
      "loss": 0.8645,
      "step": 3964
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.825306432552098e-05,
      "loss": 0.7119,
      "step": 3965
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.824195709010995e-05,
      "loss": 0.4883,
      "step": 3966
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.823084780765457e-05,
      "loss": 0.6486,
      "step": 3967
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.821973647896006e-05,
      "loss": 0.5101,
      "step": 3968
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.820862310483182e-05,
      "loss": 0.6982,
      "step": 3969
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.819750768607535e-05,
      "loss": 0.7841,
      "step": 3970
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.818639022349632e-05,
      "loss": 0.4733,
      "step": 3971
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.817527071790056e-05,
      "loss": 0.391,
      "step": 3972
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.816414917009403e-05,
      "loss": 1.0383,
      "step": 3973
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.815302558088287e-05,
      "loss": 0.5425,
      "step": 3974
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.814189995107332e-05,
      "loss": 0.508,
      "step": 3975
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.81307722814718e-05,
      "loss": 0.6051,
      "step": 3976
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.811964257288486e-05,
      "loss": 0.7448,
      "step": 3977
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.810851082611924e-05,
      "loss": 0.4623,
      "step": 3978
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.809737704198176e-05,
      "loss": 0.8626,
      "step": 3979
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.808624122127945e-05,
      "loss": 0.9498,
      "step": 3980
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.807510336481943e-05,
      "loss": 1.0833,
      "step": 3981
    },
    {
      "epoch": 1.65,
      "learning_rate": 7.806396347340902e-05,
      "loss": 0.7623,
      "step": 3982
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.805282154785567e-05,
      "loss": 0.4852,
      "step": 3983
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.804167758896696e-05,
      "loss": 0.8625,
      "step": 3984
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.803053159755063e-05,
      "loss": 0.5522,
      "step": 3985
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.801938357441457e-05,
      "loss": 0.7145,
      "step": 3986
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.800823352036684e-05,
      "loss": 0.5402,
      "step": 3987
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.799708143621558e-05,
      "loss": 0.5631,
      "step": 3988
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.798592732276914e-05,
      "loss": 0.8066,
      "step": 3989
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.797477118083601e-05,
      "loss": 1.0343,
      "step": 3990
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.79636130112248e-05,
      "loss": 0.7298,
      "step": 3991
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.795245281474428e-05,
      "loss": 0.3282,
      "step": 3992
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.794129059220337e-05,
      "loss": 0.4319,
      "step": 3993
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.793012634441112e-05,
      "loss": 0.3662,
      "step": 3994
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.791896007217678e-05,
      "loss": 0.4645,
      "step": 3995
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.790779177630966e-05,
      "loss": 0.7042,
      "step": 3996
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.789662145761931e-05,
      "loss": 0.4471,
      "step": 3997
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.788544911691536e-05,
      "loss": 0.5804,
      "step": 3998
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.78742747550076e-05,
      "loss": 0.5215,
      "step": 3999
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.786309837270597e-05,
      "loss": 0.848,
      "step": 4000
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.785191997082058e-05,
      "loss": 0.5346,
      "step": 4001
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.784073955016167e-05,
      "loss": 0.7834,
      "step": 4002
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.78295571115396e-05,
      "loss": 1.0765,
      "step": 4003
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.781837265576491e-05,
      "loss": 0.4474,
      "step": 4004
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.78071861836483e-05,
      "loss": 0.6426,
      "step": 4005
    },
    {
      "epoch": 1.66,
      "learning_rate": 7.779599769600056e-05,
      "loss": 0.5349,
      "step": 4006
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.778480719363265e-05,
      "loss": 0.532,
      "step": 4007
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.777361467735573e-05,
      "loss": 0.5595,
      "step": 4008
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.776242014798103e-05,
      "loss": 0.7281,
      "step": 4009
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.775122360631995e-05,
      "loss": 0.3701,
      "step": 4010
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.774002505318408e-05,
      "loss": 0.421,
      "step": 4011
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.772882448938506e-05,
      "loss": 0.4556,
      "step": 4012
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.771762191573477e-05,
      "loss": 0.4742,
      "step": 4013
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.77064173330452e-05,
      "loss": 0.4403,
      "step": 4014
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.769521074212847e-05,
      "loss": 0.5263,
      "step": 4015
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.768400214379687e-05,
      "loss": 0.6636,
      "step": 4016
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.767279153886283e-05,
      "loss": 0.5143,
      "step": 4017
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.766157892813891e-05,
      "loss": 0.8166,
      "step": 4018
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.765036431243785e-05,
      "loss": 0.9826,
      "step": 4019
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.763914769257245e-05,
      "loss": 0.7672,
      "step": 4020
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.762792906935581e-05,
      "loss": 0.9862,
      "step": 4021
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.761670844360101e-05,
      "loss": 0.6459,
      "step": 4022
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.760548581612137e-05,
      "loss": 0.4343,
      "step": 4023
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.759426118773034e-05,
      "loss": 0.9257,
      "step": 4024
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.75830345592415e-05,
      "loss": 0.5613,
      "step": 4025
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.75718059314686e-05,
      "loss": 0.5382,
      "step": 4026
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.756057530522549e-05,
      "loss": 0.7301,
      "step": 4027
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.75493426813262e-05,
      "loss": 0.9615,
      "step": 4028
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.753810806058492e-05,
      "loss": 0.9854,
      "step": 4029
    },
    {
      "epoch": 1.67,
      "learning_rate": 7.752687144381593e-05,
      "loss": 0.773,
      "step": 4030
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.751563283183371e-05,
      "loss": 0.8324,
      "step": 4031
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.750439222545285e-05,
      "loss": 0.8035,
      "step": 4032
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.749314962548809e-05,
      "loss": 0.4395,
      "step": 4033
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.748190503275435e-05,
      "loss": 0.4217,
      "step": 4034
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.747065844806662e-05,
      "loss": 0.6041,
      "step": 4035
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.745940987224012e-05,
      "loss": 0.7204,
      "step": 4036
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.744815930609015e-05,
      "loss": 0.5988,
      "step": 4037
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.74369067504322e-05,
      "loss": 0.8782,
      "step": 4038
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.742565220608185e-05,
      "loss": 0.6141,
      "step": 4039
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.741439567385486e-05,
      "loss": 0.2674,
      "step": 4040
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.740313715456717e-05,
      "loss": 0.7031,
      "step": 4041
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.73918766490348e-05,
      "loss": 0.7795,
      "step": 4042
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.738061415807392e-05,
      "loss": 0.6053,
      "step": 4043
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.736934968250088e-05,
      "loss": 0.384,
      "step": 4044
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.735808322313216e-05,
      "loss": 0.7871,
      "step": 4045
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.734681478078439e-05,
      "loss": 0.8076,
      "step": 4046
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.733554435627428e-05,
      "loss": 0.7011,
      "step": 4047
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.73242719504188e-05,
      "loss": 0.466,
      "step": 4048
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.731299756403497e-05,
      "loss": 0.8231,
      "step": 4049
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.730172119793999e-05,
      "loss": 0.836,
      "step": 4050
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.72904428529512e-05,
      "loss": 0.6846,
      "step": 4051
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.727916252988608e-05,
      "loss": 0.6422,
      "step": 4052
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.726788022956224e-05,
      "loss": 0.4723,
      "step": 4053
    },
    {
      "epoch": 1.68,
      "learning_rate": 7.725659595279748e-05,
      "loss": 0.8659,
      "step": 4054
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.72453097004097e-05,
      "loss": 0.8781,
      "step": 4055
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.723402147321692e-05,
      "loss": 0.6225,
      "step": 4056
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.722273127203739e-05,
      "loss": 0.5459,
      "step": 4057
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.721143909768942e-05,
      "loss": 0.7273,
      "step": 4058
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.72001449509915e-05,
      "loss": 0.4674,
      "step": 4059
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.718884883276225e-05,
      "loss": 0.4203,
      "step": 4060
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.717755074382045e-05,
      "loss": 0.59,
      "step": 4061
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.7166250684985e-05,
      "loss": 0.3845,
      "step": 4062
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.715494865707497e-05,
      "loss": 0.8299,
      "step": 4063
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.714364466090955e-05,
      "loss": 0.8185,
      "step": 4064
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.71323386973081e-05,
      "loss": 0.8178,
      "step": 4065
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.712103076709006e-05,
      "loss": 1.1757,
      "step": 4066
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.71097208710751e-05,
      "loss": 0.7064,
      "step": 4067
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.709840901008296e-05,
      "loss": 0.7043,
      "step": 4068
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.708709518493356e-05,
      "loss": 0.6686,
      "step": 4069
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.707577939644695e-05,
      "loss": 0.963,
      "step": 4070
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.706446164544334e-05,
      "loss": 0.717,
      "step": 4071
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.705314193274305e-05,
      "loss": 0.5051,
      "step": 4072
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.704182025916655e-05,
      "loss": 0.3368,
      "step": 4073
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.703049662553449e-05,
      "loss": 0.8081,
      "step": 4074
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.701917103266763e-05,
      "loss": 0.7129,
      "step": 4075
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.700784348138686e-05,
      "loss": 0.613,
      "step": 4076
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.699651397251322e-05,
      "loss": 0.8017,
      "step": 4077
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.698518250686793e-05,
      "loss": 0.3457,
      "step": 4078
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.697384908527232e-05,
      "loss": 0.5799,
      "step": 4079
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.696251370854784e-05,
      "loss": 0.9073,
      "step": 4080
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.69511763775161e-05,
      "loss": 0.5608,
      "step": 4081
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.693983709299888e-05,
      "loss": 0.7778,
      "step": 4082
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.692849585581808e-05,
      "loss": 0.6859,
      "step": 4083
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.691715266679572e-05,
      "loss": 0.5637,
      "step": 4084
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.6905807526754e-05,
      "loss": 0.7242,
      "step": 4085
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.689446043651522e-05,
      "loss": 0.5074,
      "step": 4086
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.688311139690188e-05,
      "loss": 0.5294,
      "step": 4087
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.687176040873656e-05,
      "loss": 0.7829,
      "step": 4088
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.6860407472842e-05,
      "loss": 0.756,
      "step": 4089
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.684905259004111e-05,
      "loss": 0.4614,
      "step": 4090
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.68376957611569e-05,
      "loss": 0.4822,
      "step": 4091
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.682633698701254e-05,
      "loss": 1.002,
      "step": 4092
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.681497626843134e-05,
      "loss": 0.4874,
      "step": 4093
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.680361360623677e-05,
      "loss": 0.3736,
      "step": 4094
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.679224900125243e-05,
      "loss": 0.5715,
      "step": 4095
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.678088245430203e-05,
      "loss": 0.3616,
      "step": 4096
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.676951396620945e-05,
      "loss": 0.5809,
      "step": 4097
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.67581435377987e-05,
      "loss": 0.4565,
      "step": 4098
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.674677116989395e-05,
      "loss": 0.7876,
      "step": 4099
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.673539686331947e-05,
      "loss": 0.5132,
      "step": 4100
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.672402061889974e-05,
      "loss": 0.4527,
      "step": 4101
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.67126424374593e-05,
      "loss": 0.9248,
      "step": 4102
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.670126231982287e-05,
      "loss": 0.5274,
      "step": 4103
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.668988026681536e-05,
      "loss": 0.805,
      "step": 4104
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.667849627926168e-05,
      "loss": 0.5453,
      "step": 4105
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.666711035798703e-05,
      "loss": 0.5788,
      "step": 4106
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.665572250381668e-05,
      "loss": 0.6936,
      "step": 4107
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.664433271757604e-05,
      "loss": 0.5775,
      "step": 4108
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.663294100009067e-05,
      "loss": 0.7323,
      "step": 4109
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.662154735218628e-05,
      "loss": 0.818,
      "step": 4110
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.661015177468868e-05,
      "loss": 0.5654,
      "step": 4111
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.659875426842388e-05,
      "loss": 0.8292,
      "step": 4112
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.658735483421799e-05,
      "loss": 0.8148,
      "step": 4113
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.657595347289725e-05,
      "loss": 0.4867,
      "step": 4114
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.65645501852881e-05,
      "loss": 0.8091,
      "step": 4115
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.655314497221702e-05,
      "loss": 0.8719,
      "step": 4116
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.654173783451073e-05,
      "loss": 0.854,
      "step": 4117
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.653032877299604e-05,
      "loss": 0.7299,
      "step": 4118
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.651891778849988e-05,
      "loss": 0.7105,
      "step": 4119
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.650750488184937e-05,
      "loss": 0.8234,
      "step": 4120
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.649609005387174e-05,
      "loss": 0.8318,
      "step": 4121
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.648467330539438e-05,
      "loss": 0.8887,
      "step": 4122
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.647325463724477e-05,
      "loss": 0.6086,
      "step": 4123
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.646183405025058e-05,
      "loss": 0.5634,
      "step": 4124
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.64504115452396e-05,
      "loss": 0.7012,
      "step": 4125
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.643898712303976e-05,
      "loss": 0.4097,
      "step": 4126
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.642756078447915e-05,
      "loss": 0.7458,
      "step": 4127
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.641613253038593e-05,
      "loss": 0.8045,
      "step": 4128
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.640470236158848e-05,
      "loss": 0.6282,
      "step": 4129
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.63932702789153e-05,
      "loss": 0.4631,
      "step": 4130
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.638183628319497e-05,
      "loss": 1.0655,
      "step": 4131
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.637040037525631e-05,
      "loss": 0.6647,
      "step": 4132
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.635896255592816e-05,
      "loss": 0.4214,
      "step": 4133
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.63475228260396e-05,
      "loss": 0.8499,
      "step": 4134
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.633608118641979e-05,
      "loss": 0.7018,
      "step": 4135
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.632463763789807e-05,
      "loss": 0.7654,
      "step": 4136
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.63131921813039e-05,
      "loss": 0.5818,
      "step": 4137
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.630174481746683e-05,
      "loss": 0.8737,
      "step": 4138
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.629029554721663e-05,
      "loss": 0.5329,
      "step": 4139
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.627884437138315e-05,
      "loss": 0.3745,
      "step": 4140
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.626739129079643e-05,
      "loss": 0.6711,
      "step": 4141
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.625593630628657e-05,
      "loss": 0.607,
      "step": 4142
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.62444794186839e-05,
      "loss": 0.4783,
      "step": 4143
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.62330206288188e-05,
      "loss": 0.7979,
      "step": 4144
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.622155993752186e-05,
      "loss": 0.6538,
      "step": 4145
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.62100973456238e-05,
      "loss": 0.4548,
      "step": 4146
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.619863285395537e-05,
      "loss": 1.0285,
      "step": 4147
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.618716646334763e-05,
      "loss": 0.8194,
      "step": 4148
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.617569817463166e-05,
      "loss": 0.5227,
      "step": 4149
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.616422798863871e-05,
      "loss": 0.3071,
      "step": 4150
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.615275590620017e-05,
      "loss": 0.7185,
      "step": 4151
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.614128192814755e-05,
      "loss": 0.6494,
      "step": 4152
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.612980605531252e-05,
      "loss": 1.0007,
      "step": 4153
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.61183282885269e-05,
      "loss": 0.3899,
      "step": 4154
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.610684862862261e-05,
      "loss": 0.3609,
      "step": 4155
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.609536707643169e-05,
      "loss": 0.7647,
      "step": 4156
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.60838836327864e-05,
      "loss": 0.4422,
      "step": 4157
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.607239829851907e-05,
      "loss": 0.6437,
      "step": 4158
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.606091107446216e-05,
      "loss": 0.5676,
      "step": 4159
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.604942196144832e-05,
      "loss": 0.6039,
      "step": 4160
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.60379309603103e-05,
      "loss": 0.4023,
      "step": 4161
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.6026438071881e-05,
      "loss": 0.8222,
      "step": 4162
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.601494329699346e-05,
      "loss": 0.6886,
      "step": 4163
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.600344663648082e-05,
      "loss": 0.7705,
      "step": 4164
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.59919480911764e-05,
      "loss": 0.5705,
      "step": 4165
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.598044766191366e-05,
      "loss": 0.5284,
      "step": 4166
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.596894534952616e-05,
      "loss": 0.8941,
      "step": 4167
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.59574411548476e-05,
      "loss": 0.6436,
      "step": 4168
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.594593507871185e-05,
      "loss": 0.6586,
      "step": 4169
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.593442712195293e-05,
      "loss": 0.5878,
      "step": 4170
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.59229172854049e-05,
      "loss": 0.5886,
      "step": 4171
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.591140556990206e-05,
      "loss": 0.898,
      "step": 4172
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.589989197627881e-05,
      "loss": 0.3902,
      "step": 4173
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.588837650536965e-05,
      "loss": 0.84,
      "step": 4174
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.587685915800927e-05,
      "loss": 0.8024,
      "step": 4175
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.586533993503248e-05,
      "loss": 0.5665,
      "step": 4176
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.585381883727422e-05,
      "loss": 0.6918,
      "step": 4177
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.584229586556956e-05,
      "loss": 0.2792,
      "step": 4178
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.583077102075371e-05,
      "loss": 0.6518,
      "step": 4179
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.581924430366203e-05,
      "loss": 0.4657,
      "step": 4180
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.580771571512999e-05,
      "loss": 0.5023,
      "step": 4181
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.579618525599321e-05,
      "loss": 0.7857,
      "step": 4182
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.578465292708746e-05,
      "loss": 0.6541,
      "step": 4183
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.577311872924862e-05,
      "loss": 0.6572,
      "step": 4184
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.576158266331271e-05,
      "loss": 0.7792,
      "step": 4185
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.57500447301159e-05,
      "loss": 0.9837,
      "step": 4186
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.573850493049448e-05,
      "loss": 0.81,
      "step": 4187
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.572696326528492e-05,
      "loss": 0.5367,
      "step": 4188
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.571541973532372e-05,
      "loss": 0.7848,
      "step": 4189
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.570387434144765e-05,
      "loss": 0.6326,
      "step": 4190
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.569232708449349e-05,
      "loss": 0.806,
      "step": 4191
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.568077796529824e-05,
      "loss": 0.565,
      "step": 4192
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.566922698469903e-05,
      "loss": 0.6192,
      "step": 4193
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.565767414353304e-05,
      "loss": 0.3421,
      "step": 4194
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.56461194426377e-05,
      "loss": 0.6373,
      "step": 4195
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.563456288285053e-05,
      "loss": 0.3421,
      "step": 4196
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.562300446500915e-05,
      "loss": 0.6867,
      "step": 4197
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.561144418995135e-05,
      "loss": 0.5346,
      "step": 4198
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.559988205851502e-05,
      "loss": 0.4468,
      "step": 4199
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.558831807153826e-05,
      "loss": 0.5917,
      "step": 4200
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.557675222985921e-05,
      "loss": 0.4734,
      "step": 4201
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.55651845343162e-05,
      "loss": 0.4541,
      "step": 4202
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.555361498574771e-05,
      "loss": 0.7061,
      "step": 4203
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.554204358499232e-05,
      "loss": 0.6389,
      "step": 4204
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.553047033288874e-05,
      "loss": 1.0822,
      "step": 4205
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.551889523027582e-05,
      "loss": 0.6102,
      "step": 4206
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.550731827799258e-05,
      "loss": 0.6391,
      "step": 4207
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.549573947687811e-05,
      "loss": 0.4691,
      "step": 4208
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.54841588277717e-05,
      "loss": 0.6499,
      "step": 4209
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.547257633151272e-05,
      "loss": 0.6717,
      "step": 4210
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.546099198894072e-05,
      "loss": 0.5986,
      "step": 4211
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.544940580089534e-05,
      "loss": 0.5157,
      "step": 4212
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.543781776821639e-05,
      "loss": 0.6611,
      "step": 4213
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.542622789174378e-05,
      "loss": 0.7755,
      "step": 4214
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.541463617231758e-05,
      "loss": 0.7357,
      "step": 4215
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.5403042610778e-05,
      "loss": 0.6821,
      "step": 4216
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.539144720796533e-05,
      "loss": 0.6168,
      "step": 4217
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.537984996472008e-05,
      "loss": 0.9607,
      "step": 4218
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.536825088188283e-05,
      "loss": 0.4276,
      "step": 4219
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.535664996029427e-05,
      "loss": 0.4824,
      "step": 4220
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.534504720079533e-05,
      "loss": 0.9308,
      "step": 4221
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.533344260422698e-05,
      "loss": 0.6807,
      "step": 4222
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.532183617143032e-05,
      "loss": 0.4974,
      "step": 4223
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.531022790324664e-05,
      "loss": 0.6674,
      "step": 4224
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.529861780051732e-05,
      "loss": 0.7442,
      "step": 4225
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.52870058640839e-05,
      "loss": 0.6432,
      "step": 4226
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.527539209478805e-05,
      "loss": 0.5398,
      "step": 4227
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.526377649347153e-05,
      "loss": 0.703,
      "step": 4228
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.52521590609763e-05,
      "loss": 0.5533,
      "step": 4229
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.524053979814441e-05,
      "loss": 0.7707,
      "step": 4230
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.522891870581804e-05,
      "loss": 1.223,
      "step": 4231
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.521729578483951e-05,
      "loss": 0.8252,
      "step": 4232
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.520567103605129e-05,
      "loss": 0.5,
      "step": 4233
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.519404446029598e-05,
      "loss": 0.7441,
      "step": 4234
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.518241605841629e-05,
      "loss": 1.0944,
      "step": 4235
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.517078583125508e-05,
      "loss": 0.6777,
      "step": 4236
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.515915377965533e-05,
      "loss": 0.771,
      "step": 4237
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.514751990446014e-05,
      "loss": 0.3473,
      "step": 4238
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.513588420651281e-05,
      "loss": 0.8434,
      "step": 4239
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.512424668665668e-05,
      "loss": 0.8256,
      "step": 4240
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.511260734573528e-05,
      "loss": 0.4007,
      "step": 4241
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.510096618459226e-05,
      "loss": 0.7368,
      "step": 4242
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.508932320407139e-05,
      "loss": 0.5116,
      "step": 4243
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.507767840501658e-05,
      "loss": 0.5382,
      "step": 4244
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.506603178827189e-05,
      "loss": 1.0135,
      "step": 4245
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.505438335468149e-05,
      "loss": 0.5487,
      "step": 4246
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.504273310508967e-05,
      "loss": 0.4892,
      "step": 4247
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.503108104034088e-05,
      "loss": 0.6708,
      "step": 4248
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.50194271612797e-05,
      "loss": 0.8283,
      "step": 4249
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.50077714687508e-05,
      "loss": 0.7322,
      "step": 4250
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.499611396359904e-05,
      "loss": 0.7266,
      "step": 4251
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.498445464666936e-05,
      "loss": 0.5551,
      "step": 4252
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.497279351880688e-05,
      "loss": 0.6426,
      "step": 4253
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.496113058085683e-05,
      "loss": 0.6842,
      "step": 4254
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.494946583366455e-05,
      "loss": 0.5643,
      "step": 4255
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.493779927807552e-05,
      "loss": 1.0324,
      "step": 4256
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.492613091493537e-05,
      "loss": 0.8504,
      "step": 4257
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.491446074508985e-05,
      "loss": 0.8137,
      "step": 4258
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.490278876938486e-05,
      "loss": 0.6397,
      "step": 4259
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.489111498866639e-05,
      "loss": 0.6012,
      "step": 4260
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.48794394037806e-05,
      "loss": 0.8649,
      "step": 4261
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.486776201557373e-05,
      "loss": 0.4032,
      "step": 4262
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.485608282489222e-05,
      "loss": 0.4733,
      "step": 4263
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.48444018325826e-05,
      "loss": 0.4544,
      "step": 4264
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.483271903949155e-05,
      "loss": 0.5844,
      "step": 4265
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.482103444646582e-05,
      "loss": 0.7,
      "step": 4266
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.480934805435238e-05,
      "loss": 0.6256,
      "step": 4267
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.479765986399828e-05,
      "loss": 0.444,
      "step": 4268
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.478596987625068e-05,
      "loss": 0.5802,
      "step": 4269
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.477427809195693e-05,
      "loss": 0.9644,
      "step": 4270
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.476258451196445e-05,
      "loss": 0.5084,
      "step": 4271
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.475088913712088e-05,
      "loss": 0.5369,
      "step": 4272
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.473919196827384e-05,
      "loss": 1.171,
      "step": 4273
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.472749300627121e-05,
      "loss": 0.5482,
      "step": 4274
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.471579225196101e-05,
      "loss": 0.551,
      "step": 4275
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.470408970619125e-05,
      "loss": 0.7444,
      "step": 4276
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.46923853698102e-05,
      "loss": 0.7036,
      "step": 4277
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.468067924366622e-05,
      "loss": 0.6056,
      "step": 4278
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.466897132860781e-05,
      "loss": 0.851,
      "step": 4279
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.465726162548357e-05,
      "loss": 0.7942,
      "step": 4280
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.464555013514224e-05,
      "loss": 0.586,
      "step": 4281
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.46338368584327e-05,
      "loss": 0.711,
      "step": 4282
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.462212179620397e-05,
      "loss": 0.9385,
      "step": 4283
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.461040494930517e-05,
      "loss": 1.1096,
      "step": 4284
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.45986863185856e-05,
      "loss": 0.6499,
      "step": 4285
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.45869659048946e-05,
      "loss": 0.7583,
      "step": 4286
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.457524370908173e-05,
      "loss": 0.4134,
      "step": 4287
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.456351973199663e-05,
      "loss": 1.0028,
      "step": 4288
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.45517939744891e-05,
      "loss": 0.846,
      "step": 4289
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.454006643740903e-05,
      "loss": 0.6401,
      "step": 4290
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.452833712160646e-05,
      "loss": 0.4007,
      "step": 4291
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.451660602793156e-05,
      "loss": 0.718,
      "step": 4292
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.450487315723464e-05,
      "loss": 0.7496,
      "step": 4293
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.449313851036612e-05,
      "loss": 0.9438,
      "step": 4294
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.448140208817655e-05,
      "loss": 0.4787,
      "step": 4295
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.446966389151661e-05,
      "loss": 0.6378,
      "step": 4296
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.445792392123715e-05,
      "loss": 0.5825,
      "step": 4297
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.444618217818906e-05,
      "loss": 0.7667,
      "step": 4298
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.443443866322342e-05,
      "loss": 0.5097,
      "step": 4299
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.442269337719145e-05,
      "loss": 0.8279,
      "step": 4300
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.441094632094447e-05,
      "loss": 0.6922,
      "step": 4301
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.439919749533393e-05,
      "loss": 0.3592,
      "step": 4302
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.438744690121141e-05,
      "loss": 0.3555,
      "step": 4303
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.437569453942862e-05,
      "loss": 0.5624,
      "step": 4304
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.436394041083741e-05,
      "loss": 0.4486,
      "step": 4305
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.435218451628973e-05,
      "loss": 0.7244,
      "step": 4306
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.43404268566377e-05,
      "loss": 0.5193,
      "step": 4307
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.432866743273351e-05,
      "loss": 0.6224,
      "step": 4308
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.431690624542954e-05,
      "loss": 0.5053,
      "step": 4309
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.430514329557826e-05,
      "loss": 0.4603,
      "step": 4310
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.429337858403228e-05,
      "loss": 0.4925,
      "step": 4311
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.428161211164432e-05,
      "loss": 0.7801,
      "step": 4312
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.426984387926724e-05,
      "loss": 0.4186,
      "step": 4313
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.425807388775408e-05,
      "loss": 0.6701,
      "step": 4314
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.424630213795788e-05,
      "loss": 1.1154,
      "step": 4315
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.423452863073194e-05,
      "loss": 0.5376,
      "step": 4316
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.422275336692961e-05,
      "loss": 0.7789,
      "step": 4317
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.42109763474044e-05,
      "loss": 0.542,
      "step": 4318
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.419919757300991e-05,
      "loss": 0.26,
      "step": 4319
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.418741704459993e-05,
      "loss": 0.7946,
      "step": 4320
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.417563476302833e-05,
      "loss": 0.3408,
      "step": 4321
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.41638507291491e-05,
      "loss": 0.365,
      "step": 4322
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.415206494381639e-05,
      "loss": 0.8884,
      "step": 4323
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.414027740788446e-05,
      "loss": 0.4564,
      "step": 4324
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.41284881222077e-05,
      "loss": 0.4008,
      "step": 4325
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.411669708764062e-05,
      "loss": 0.6991,
      "step": 4326
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.410490430503786e-05,
      "loss": 0.7375,
      "step": 4327
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.40931097752542e-05,
      "loss": 0.7079,
      "step": 4328
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.408131349914453e-05,
      "loss": 0.5292,
      "step": 4329
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.406951547756387e-05,
      "loss": 0.5716,
      "step": 4330
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.40577157113674e-05,
      "loss": 0.625,
      "step": 4331
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.404591420141032e-05,
      "loss": 0.4282,
      "step": 4332
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.403411094854811e-05,
      "loss": 0.7298,
      "step": 4333
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.402230595363624e-05,
      "loss": 0.4021,
      "step": 4334
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.401049921753041e-05,
      "loss": 0.7116,
      "step": 4335
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.399869074108638e-05,
      "loss": 0.585,
      "step": 4336
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.398688052516003e-05,
      "loss": 0.5037,
      "step": 4337
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.397506857060744e-05,
      "loss": 0.7721,
      "step": 4338
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.396325487828475e-05,
      "loss": 0.7141,
      "step": 4339
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.395143944904826e-05,
      "loss": 0.691,
      "step": 4340
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.393962228375432e-05,
      "loss": 0.5247,
      "step": 4341
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.392780338325953e-05,
      "loss": 0.6031,
      "step": 4342
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.391598274842054e-05,
      "loss": 0.4631,
      "step": 4343
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.390416038009411e-05,
      "loss": 0.9205,
      "step": 4344
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.38923362791372e-05,
      "loss": 0.6512,
      "step": 4345
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.388051044640679e-05,
      "loss": 0.6925,
      "step": 4346
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.38686828827601e-05,
      "loss": 0.6207,
      "step": 4347
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.385685358905438e-05,
      "loss": 0.4096,
      "step": 4348
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.384502256614708e-05,
      "loss": 0.6292,
      "step": 4349
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.38331898148957e-05,
      "loss": 0.7379,
      "step": 4350
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.382135533615796e-05,
      "loss": 0.8914,
      "step": 4351
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.380951913079159e-05,
      "loss": 0.7077,
      "step": 4352
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.379768119965453e-05,
      "loss": 1.1263,
      "step": 4353
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.378584154360485e-05,
      "loss": 0.7284,
      "step": 4354
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.377400016350068e-05,
      "loss": 0.5115,
      "step": 4355
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.376215706020034e-05,
      "loss": 0.9294,
      "step": 4356
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.375031223456222e-05,
      "loss": 0.5677,
      "step": 4357
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.373846568744487e-05,
      "loss": 0.4072,
      "step": 4358
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.372661741970697e-05,
      "loss": 0.6961,
      "step": 4359
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.371476743220728e-05,
      "loss": 0.8466,
      "step": 4360
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.370291572580475e-05,
      "loss": 0.546,
      "step": 4361
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.36910623013584e-05,
      "loss": 0.7744,
      "step": 4362
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.36792071597274e-05,
      "loss": 0.6367,
      "step": 4363
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.366735030177105e-05,
      "loss": 0.8928,
      "step": 4364
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.365549172834874e-05,
      "loss": 0.5792,
      "step": 4365
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.364363144032002e-05,
      "loss": 0.4487,
      "step": 4366
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.363176943854452e-05,
      "loss": 0.6282,
      "step": 4367
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.361990572388211e-05,
      "loss": 0.6479,
      "step": 4368
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.360804029719262e-05,
      "loss": 0.4143,
      "step": 4369
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.359617315933612e-05,
      "loss": 0.7528,
      "step": 4370
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.358430431117274e-05,
      "loss": 0.7953,
      "step": 4371
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.357243375356279e-05,
      "loss": 0.946,
      "step": 4372
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.35605614873667e-05,
      "loss": 0.4084,
      "step": 4373
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.354868751344493e-05,
      "loss": 0.6975,
      "step": 4374
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.35368118326582e-05,
      "loss": 0.7638,
      "step": 4375
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.352493444586724e-05,
      "loss": 0.6435,
      "step": 4376
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.351305535393298e-05,
      "loss": 0.7472,
      "step": 4377
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.350117455771644e-05,
      "loss": 0.6516,
      "step": 4378
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.348929205807876e-05,
      "loss": 0.4901,
      "step": 4379
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.347740785588122e-05,
      "loss": 0.7096,
      "step": 4380
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.346552195198522e-05,
      "loss": 0.3659,
      "step": 4381
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.345363434725226e-05,
      "loss": 0.5307,
      "step": 4382
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.3441745042544e-05,
      "loss": 0.5116,
      "step": 4383
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.342985403872219e-05,
      "loss": 0.5749,
      "step": 4384
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.341796133664873e-05,
      "loss": 0.7831,
      "step": 4385
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.340606693718562e-05,
      "loss": 0.3493,
      "step": 4386
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.339417084119501e-05,
      "loss": 0.7132,
      "step": 4387
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.338227304953915e-05,
      "loss": 0.9758,
      "step": 4388
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.337037356308044e-05,
      "loss": 0.7669,
      "step": 4389
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.335847238268134e-05,
      "loss": 0.5489,
      "step": 4390
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.334656950920448e-05,
      "loss": 0.7134,
      "step": 4391
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.333466494351266e-05,
      "loss": 0.5502,
      "step": 4392
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.332275868646871e-05,
      "loss": 0.6291,
      "step": 4393
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.331085073893564e-05,
      "loss": 0.6006,
      "step": 4394
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.329894110177653e-05,
      "loss": 0.5589,
      "step": 4395
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.328702977585468e-05,
      "loss": 1.0056,
      "step": 4396
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.32751167620334e-05,
      "loss": 0.5783,
      "step": 4397
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.32632020611762e-05,
      "loss": 0.6975,
      "step": 4398
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.325128567414669e-05,
      "loss": 0.5775,
      "step": 4399
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.323936760180857e-05,
      "loss": 0.7517,
      "step": 4400
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.322744784502571e-05,
      "loss": 1.1932,
      "step": 4401
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.321552640466207e-05,
      "loss": 0.9973,
      "step": 4402
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.320360328158178e-05,
      "loss": 0.4703,
      "step": 4403
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.319167847664902e-05,
      "loss": 0.7048,
      "step": 4404
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.317975199072814e-05,
      "loss": 0.8154,
      "step": 4405
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.316782382468359e-05,
      "loss": 0.7634,
      "step": 4406
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.315589397937996e-05,
      "loss": 0.681,
      "step": 4407
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.314396245568195e-05,
      "loss": 0.6376,
      "step": 4408
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.31320292544544e-05,
      "loss": 0.5498,
      "step": 4409
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.312009437656225e-05,
      "loss": 0.6666,
      "step": 4410
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.310815782287055e-05,
      "loss": 0.7624,
      "step": 4411
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.309621959424451e-05,
      "loss": 0.4392,
      "step": 4412
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.308427969154946e-05,
      "loss": 0.577,
      "step": 4413
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.307233811565077e-05,
      "loss": 0.7283,
      "step": 4414
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.306039486741407e-05,
      "loss": 1.306,
      "step": 4415
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.304844994770497e-05,
      "loss": 0.6005,
      "step": 4416
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.303650335738931e-05,
      "loss": 0.7882,
      "step": 4417
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.302455509733298e-05,
      "loss": 0.6918,
      "step": 4418
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.301260516840203e-05,
      "loss": 0.9641,
      "step": 4419
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.300065357146261e-05,
      "loss": 0.6619,
      "step": 4420
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.298870030738103e-05,
      "loss": 0.8748,
      "step": 4421
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.297674537702365e-05,
      "loss": 0.5957,
      "step": 4422
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.296478878125703e-05,
      "loss": 0.7699,
      "step": 4423
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.295283052094779e-05,
      "loss": 0.5551,
      "step": 4424
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.29408705969627e-05,
      "loss": 0.87,
      "step": 4425
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.292890901016863e-05,
      "loss": 0.4118,
      "step": 4426
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.291694576143261e-05,
      "loss": 0.7786,
      "step": 4427
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.290498085162173e-05,
      "loss": 0.774,
      "step": 4428
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.289301428160327e-05,
      "loss": 0.5763,
      "step": 4429
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.288104605224459e-05,
      "loss": 0.7902,
      "step": 4430
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.286907616441315e-05,
      "loss": 0.5534,
      "step": 4431
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.285710461897656e-05,
      "loss": 0.6897,
      "step": 4432
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.28451314168026e-05,
      "loss": 0.5725,
      "step": 4433
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.283315655875905e-05,
      "loss": 0.6043,
      "step": 4434
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.28211800457139e-05,
      "loss": 0.4726,
      "step": 4435
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.280920187853525e-05,
      "loss": 0.6284,
      "step": 4436
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.279722205809128e-05,
      "loss": 0.6775,
      "step": 4437
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.278524058525033e-05,
      "loss": 0.3785,
      "step": 4438
    },
    {
      "epoch": 1.84,
      "learning_rate": 7.277325746088086e-05,
      "loss": 0.4884,
      "step": 4439
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.276127268585143e-05,
      "loss": 0.845,
      "step": 4440
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.274928626103069e-05,
      "loss": 0.7374,
      "step": 4441
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.273729818728748e-05,
      "loss": 0.4656,
      "step": 4442
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.27253084654907e-05,
      "loss": 0.516,
      "step": 4443
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.271331709650943e-05,
      "loss": 0.8315,
      "step": 4444
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.27013240812128e-05,
      "loss": 0.7836,
      "step": 4445
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.26893294204701e-05,
      "loss": 0.3968,
      "step": 4446
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.267733311515075e-05,
      "loss": 0.506,
      "step": 4447
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.266533516612426e-05,
      "loss": 0.7958,
      "step": 4448
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.265333557426024e-05,
      "loss": 0.6215,
      "step": 4449
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.26413343404285e-05,
      "loss": 0.7183,
      "step": 4450
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.262933146549889e-05,
      "loss": 0.7729,
      "step": 4451
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.26173269503414e-05,
      "loss": 0.4154,
      "step": 4452
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.260532079582616e-05,
      "loss": 0.4733,
      "step": 4453
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.25933130028234e-05,
      "loss": 0.4589,
      "step": 4454
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.258130357220347e-05,
      "loss": 0.5489,
      "step": 4455
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.256929250483686e-05,
      "loss": 0.925,
      "step": 4456
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.255727980159414e-05,
      "loss": 0.5273,
      "step": 4457
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.254526546334601e-05,
      "loss": 0.3866,
      "step": 4458
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.253324949096333e-05,
      "loss": 0.474,
      "step": 4459
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.252123188531704e-05,
      "loss": 0.322,
      "step": 4460
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.250921264727818e-05,
      "loss": 0.4296,
      "step": 4461
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.249719177771795e-05,
      "loss": 0.5778,
      "step": 4462
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.248516927750764e-05,
      "loss": 0.7303,
      "step": 4463
    },
    {
      "epoch": 1.85,
      "learning_rate": 7.24731451475187e-05,
      "loss": 0.4476,
      "step": 4464
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.246111938862264e-05,
      "loss": 0.5498,
      "step": 4465
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.244909200169112e-05,
      "loss": 0.4513,
      "step": 4466
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.243706298759593e-05,
      "loss": 0.6455,
      "step": 4467
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.242503234720892e-05,
      "loss": 0.966,
      "step": 4468
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.241300008140214e-05,
      "loss": 0.7308,
      "step": 4469
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.24009661910477e-05,
      "loss": 0.7048,
      "step": 4470
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.238893067701786e-05,
      "loss": 1.0181,
      "step": 4471
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.237689354018496e-05,
      "loss": 0.615,
      "step": 4472
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.236485478142151e-05,
      "loss": 0.5867,
      "step": 4473
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.235281440160007e-05,
      "loss": 0.5505,
      "step": 4474
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.23407724015934e-05,
      "loss": 0.7096,
      "step": 4475
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.232872878227428e-05,
      "loss": 0.4825,
      "step": 4476
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.231668354451571e-05,
      "loss": 0.9756,
      "step": 4477
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.230463668919071e-05,
      "loss": 0.8976,
      "step": 4478
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.229258821717253e-05,
      "loss": 0.5956,
      "step": 4479
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.22805381293344e-05,
      "loss": 1.052,
      "step": 4480
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.226848642654979e-05,
      "loss": 0.7187,
      "step": 4481
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.225643310969222e-05,
      "loss": 0.477,
      "step": 4482
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.224437817963533e-05,
      "loss": 0.6383,
      "step": 4483
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.22323216372529e-05,
      "loss": 0.5903,
      "step": 4484
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.222026348341884e-05,
      "loss": 0.8,
      "step": 4485
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.220820371900711e-05,
      "loss": 0.9362,
      "step": 4486
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.219614234489185e-05,
      "loss": 0.5919,
      "step": 4487
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.218407936194732e-05,
      "loss": 0.4412,
      "step": 4488
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.217201477104784e-05,
      "loss": 0.4831,
      "step": 4489
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.21599485730679e-05,
      "loss": 0.4397,
      "step": 4490
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.214788076888208e-05,
      "loss": 0.7309,
      "step": 4491
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.213581135936508e-05,
      "loss": 0.7195,
      "step": 4492
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.212374034539172e-05,
      "loss": 0.5259,
      "step": 4493
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.211166772783695e-05,
      "loss": 0.6822,
      "step": 4494
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.20995935075758e-05,
      "loss": 0.641,
      "step": 4495
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.208751768548345e-05,
      "loss": 0.7128,
      "step": 4496
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.20754402624352e-05,
      "loss": 1.0982,
      "step": 4497
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.206336123930643e-05,
      "loss": 0.4828,
      "step": 4498
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.205128061697264e-05,
      "loss": 0.7935,
      "step": 4499
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.20391983963095e-05,
      "loss": 0.6998,
      "step": 4500
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.202711457819274e-05,
      "loss": 0.8754,
      "step": 4501
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.201502916349823e-05,
      "loss": 0.4343,
      "step": 4502
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.200294215310193e-05,
      "loss": 0.3604,
      "step": 4503
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.199085354787997e-05,
      "loss": 0.5358,
      "step": 4504
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.197876334870852e-05,
      "loss": 0.6833,
      "step": 4505
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.196667155646394e-05,
      "loss": 0.4056,
      "step": 4506
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.195457817202265e-05,
      "loss": 0.6444,
      "step": 4507
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.194248319626124e-05,
      "loss": 0.543,
      "step": 4508
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.193038663005634e-05,
      "loss": 0.5799,
      "step": 4509
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.191828847428475e-05,
      "loss": 0.8185,
      "step": 4510
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.19061887298234e-05,
      "loss": 0.7462,
      "step": 4511
    },
    {
      "epoch": 1.87,
      "learning_rate": 7.189408739754927e-05,
      "loss": 0.623,
      "step": 4512
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.188198447833952e-05,
      "loss": 0.9764,
      "step": 4513
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.186987997307139e-05,
      "loss": 0.6212,
      "step": 4514
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.185777388262226e-05,
      "loss": 0.4693,
      "step": 4515
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.184566620786958e-05,
      "loss": 0.4991,
      "step": 4516
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.183355694969094e-05,
      "loss": 0.7507,
      "step": 4517
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.182144610896409e-05,
      "loss": 0.7321,
      "step": 4518
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.18093336865668e-05,
      "loss": 0.6287,
      "step": 4519
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.179721968337704e-05,
      "loss": 0.9037,
      "step": 4520
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.178510410027288e-05,
      "loss": 0.7809,
      "step": 4521
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.177298693813243e-05,
      "loss": 1.0486,
      "step": 4522
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.176086819783402e-05,
      "loss": 0.9718,
      "step": 4523
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.174874788025603e-05,
      "loss": 0.6687,
      "step": 4524
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.173662598627695e-05,
      "loss": 0.8218,
      "step": 4525
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.172450251677543e-05,
      "loss": 0.7263,
      "step": 4526
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.17123774726302e-05,
      "loss": 0.774,
      "step": 4527
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.17002508547201e-05,
      "loss": 0.6356,
      "step": 4528
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.168812266392411e-05,
      "loss": 0.8561,
      "step": 4529
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.167599290112133e-05,
      "loss": 0.51,
      "step": 4530
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.166386156719089e-05,
      "loss": 0.6631,
      "step": 4531
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.165172866301218e-05,
      "loss": 0.8687,
      "step": 4532
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.163959418946455e-05,
      "loss": 0.6656,
      "step": 4533
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.16274581474276e-05,
      "loss": 0.5092,
      "step": 4534
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.161532053778091e-05,
      "loss": 0.6743,
      "step": 4535
    },
    {
      "epoch": 1.88,
      "learning_rate": 7.160318136140429e-05,
      "loss": 0.5712,
      "step": 4536
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.159104061917762e-05,
      "loss": 0.6731,
      "step": 4537
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.157889831198086e-05,
      "loss": 0.7018,
      "step": 4538
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.156675444069413e-05,
      "loss": 0.6102,
      "step": 4539
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.155460900619764e-05,
      "loss": 0.6064,
      "step": 4540
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.154246200937173e-05,
      "loss": 0.9532,
      "step": 4541
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.153031345109685e-05,
      "loss": 0.7006,
      "step": 4542
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.151816333225353e-05,
      "loss": 0.5373,
      "step": 4543
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.150601165372246e-05,
      "loss": 0.405,
      "step": 4544
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.14938584163844e-05,
      "loss": 0.5225,
      "step": 4545
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.148170362112029e-05,
      "loss": 0.4755,
      "step": 4546
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.14695472688111e-05,
      "loss": 0.4766,
      "step": 4547
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.145738936033796e-05,
      "loss": 0.7564,
      "step": 4548
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.144522989658211e-05,
      "loss": 0.8504,
      "step": 4549
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.143306887842489e-05,
      "loss": 0.7281,
      "step": 4550
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.142090630674775e-05,
      "loss": 0.6149,
      "step": 4551
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.14087421824323e-05,
      "loss": 0.6018,
      "step": 4552
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.13965765063602e-05,
      "loss": 0.8383,
      "step": 4553
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.138440927941324e-05,
      "loss": 0.9037,
      "step": 4554
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.13722405024733e-05,
      "loss": 0.5244,
      "step": 4555
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.136007017642249e-05,
      "loss": 0.9289,
      "step": 4556
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.134789830214288e-05,
      "loss": 0.8379,
      "step": 4557
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.133572488051672e-05,
      "loss": 0.4995,
      "step": 4558
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.132354991242638e-05,
      "loss": 0.8374,
      "step": 4559
    },
    {
      "epoch": 1.89,
      "learning_rate": 7.131137339875434e-05,
      "loss": 0.6508,
      "step": 4560
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.129919534038313e-05,
      "loss": 0.7863,
      "step": 4561
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.128701573819553e-05,
      "loss": 0.5582,
      "step": 4562
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.12748345930743e-05,
      "loss": 0.62,
      "step": 4563
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.126265190590234e-05,
      "loss": 1.0903,
      "step": 4564
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.125046767756273e-05,
      "loss": 0.4887,
      "step": 4565
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.123828190893856e-05,
      "loss": 0.6162,
      "step": 4566
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.122609460091312e-05,
      "loss": 0.7945,
      "step": 4567
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.121390575436975e-05,
      "loss": 0.6928,
      "step": 4568
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.120171537019194e-05,
      "loss": 1.1387,
      "step": 4569
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.118952344926329e-05,
      "loss": 0.9178,
      "step": 4570
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.117732999246747e-05,
      "loss": 0.5609,
      "step": 4571
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.116513500068831e-05,
      "loss": 0.8488,
      "step": 4572
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.115293847480973e-05,
      "loss": 0.7942,
      "step": 4573
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.114074041571577e-05,
      "loss": 0.4346,
      "step": 4574
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.112854082429057e-05,
      "loss": 0.8037,
      "step": 4575
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.111633970141836e-05,
      "loss": 0.5938,
      "step": 4576
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.110413704798356e-05,
      "loss": 0.4579,
      "step": 4577
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.10919328648706e-05,
      "loss": 0.4855,
      "step": 4578
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.107972715296406e-05,
      "loss": 0.8276,
      "step": 4579
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.106751991314868e-05,
      "loss": 0.6647,
      "step": 4580
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.105531114630928e-05,
      "loss": 0.8299,
      "step": 4581
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.104310085333072e-05,
      "loss": 0.4215,
      "step": 4582
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.103088903509809e-05,
      "loss": 0.6572,
      "step": 4583
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.101867569249649e-05,
      "loss": 0.8589,
      "step": 4584
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.100646082641119e-05,
      "loss": 0.6621,
      "step": 4585
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.099424443772755e-05,
      "loss": 0.7456,
      "step": 4586
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.098202652733105e-05,
      "loss": 0.5958,
      "step": 4587
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.096980709610726e-05,
      "loss": 0.3431,
      "step": 4588
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.09575861449419e-05,
      "loss": 0.957,
      "step": 4589
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.094536367472073e-05,
      "loss": 0.5453,
      "step": 4590
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.09331396863297e-05,
      "loss": 0.7378,
      "step": 4591
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.092091418065483e-05,
      "loss": 0.6571,
      "step": 4592
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.090868715858223e-05,
      "loss": 0.5578,
      "step": 4593
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.089645862099816e-05,
      "loss": 0.5529,
      "step": 4594
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.088422856878898e-05,
      "loss": 0.7516,
      "step": 4595
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.087199700284116e-05,
      "loss": 0.3069,
      "step": 4596
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.085976392404125e-05,
      "loss": 0.5978,
      "step": 4597
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.084752933327594e-05,
      "loss": 0.3891,
      "step": 4598
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.083529323143202e-05,
      "loss": 0.6157,
      "step": 4599
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.08230556193964e-05,
      "loss": 0.7193,
      "step": 4600
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.081081649805609e-05,
      "loss": 0.4675,
      "step": 4601
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.07985758682982e-05,
      "loss": 0.5762,
      "step": 4602
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.078633373100997e-05,
      "loss": 0.546,
      "step": 4603
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.077409008707874e-05,
      "loss": 0.6407,
      "step": 4604
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.076184493739198e-05,
      "loss": 0.6632,
      "step": 4605
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.074959828283718e-05,
      "loss": 0.7446,
      "step": 4606
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.073735012430209e-05,
      "loss": 0.4297,
      "step": 4607
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.072510046267442e-05,
      "loss": 0.6486,
      "step": 4608
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.071284929884209e-05,
      "loss": 0.7547,
      "step": 4609
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.070059663369309e-05,
      "loss": 0.6323,
      "step": 4610
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.068834246811551e-05,
      "loss": 0.6805,
      "step": 4611
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.067608680299758e-05,
      "loss": 0.4261,
      "step": 4612
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.066382963922758e-05,
      "loss": 0.8888,
      "step": 4613
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.0651570977694e-05,
      "loss": 0.4773,
      "step": 4614
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.063931081928532e-05,
      "loss": 0.4841,
      "step": 4615
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.062704916489022e-05,
      "loss": 0.8577,
      "step": 4616
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.061478601539743e-05,
      "loss": 0.8518,
      "step": 4617
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.060252137169583e-05,
      "loss": 0.7866,
      "step": 4618
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.059025523467439e-05,
      "loss": 0.4816,
      "step": 4619
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.057798760522218e-05,
      "loss": 0.5343,
      "step": 4620
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.05657184842284e-05,
      "loss": 0.3523,
      "step": 4621
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.055344787258232e-05,
      "loss": 0.5368,
      "step": 4622
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.054117577117337e-05,
      "loss": 0.4831,
      "step": 4623
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.052890218089106e-05,
      "loss": 0.5785,
      "step": 4624
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.051662710262498e-05,
      "loss": 0.8367,
      "step": 4625
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.05043505372649e-05,
      "loss": 0.7028,
      "step": 4626
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.049207248570062e-05,
      "loss": 0.8086,
      "step": 4627
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.04797929488221e-05,
      "loss": 0.6103,
      "step": 4628
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.046751192751937e-05,
      "loss": 0.7974,
      "step": 4629
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.045522942268261e-05,
      "loss": 0.3329,
      "step": 4630
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.044294543520206e-05,
      "loss": 0.7929,
      "step": 4631
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.043065996596813e-05,
      "loss": 0.4458,
      "step": 4632
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.041837301587129e-05,
      "loss": 0.7797,
      "step": 4633
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.04060845858021e-05,
      "loss": 0.9811,
      "step": 4634
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.039379467665128e-05,
      "loss": 0.7693,
      "step": 4635
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.03815032893096e-05,
      "loss": 0.7589,
      "step": 4636
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.036921042466802e-05,
      "loss": 0.8756,
      "step": 4637
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.035691608361752e-05,
      "loss": 0.6468,
      "step": 4638
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.034462026704923e-05,
      "loss": 0.7349,
      "step": 4639
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.03323229758544e-05,
      "loss": 1.0334,
      "step": 4640
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.032002421092431e-05,
      "loss": 0.5282,
      "step": 4641
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.030772397315049e-05,
      "loss": 0.5073,
      "step": 4642
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.02954222634244e-05,
      "loss": 0.4937,
      "step": 4643
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.028311908263777e-05,
      "loss": 0.567,
      "step": 4644
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.027081443168234e-05,
      "loss": 0.6478,
      "step": 4645
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.025850831144994e-05,
      "loss": 0.6246,
      "step": 4646
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.02462007228326e-05,
      "loss": 0.7827,
      "step": 4647
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.023389166672238e-05,
      "loss": 0.6923,
      "step": 4648
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.022158114401151e-05,
      "loss": 0.6771,
      "step": 4649
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.02092691555922e-05,
      "loss": 0.5611,
      "step": 4650
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.019695570235694e-05,
      "loss": 0.6302,
      "step": 4651
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.018464078519818e-05,
      "loss": 0.6459,
      "step": 4652
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.017232440500856e-05,
      "loss": 0.4423,
      "step": 4653
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.016000656268081e-05,
      "loss": 0.7872,
      "step": 4654
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.014768725910773e-05,
      "loss": 1.0465,
      "step": 4655
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.013536649518227e-05,
      "loss": 0.8676,
      "step": 4656
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.012304427179748e-05,
      "loss": 0.8073,
      "step": 4657
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.011072058984646e-05,
      "loss": 0.7506,
      "step": 4658
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.009839545022251e-05,
      "loss": 0.8334,
      "step": 4659
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.008606885381895e-05,
      "loss": 0.3277,
      "step": 4660
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.007374080152927e-05,
      "loss": 0.5196,
      "step": 4661
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.0061411294247e-05,
      "loss": 0.8483,
      "step": 4662
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.004908033286584e-05,
      "loss": 0.5679,
      "step": 4663
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.003674791827956e-05,
      "loss": 0.5887,
      "step": 4664
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.002441405138205e-05,
      "loss": 0.4217,
      "step": 4665
    },
    {
      "epoch": 1.94,
      "learning_rate": 7.001207873306731e-05,
      "loss": 0.7052,
      "step": 4666
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.999974196422939e-05,
      "loss": 0.4472,
      "step": 4667
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.998740374576251e-05,
      "loss": 0.5833,
      "step": 4668
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.997506407856099e-05,
      "loss": 0.869,
      "step": 4669
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.996272296351923e-05,
      "loss": 0.6854,
      "step": 4670
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.995038040153172e-05,
      "loss": 0.7923,
      "step": 4671
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.99380363934931e-05,
      "loss": 0.5769,
      "step": 4672
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.99256909402981e-05,
      "loss": 0.7475,
      "step": 4673
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.991334404284153e-05,
      "loss": 1.0942,
      "step": 4674
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.990099570201836e-05,
      "loss": 0.7865,
      "step": 4675
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.988864591872356e-05,
      "loss": 0.6924,
      "step": 4676
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.987629469385234e-05,
      "loss": 0.7648,
      "step": 4677
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.98639420282999e-05,
      "loss": 0.4494,
      "step": 4678
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.985158792296158e-05,
      "loss": 1.1058,
      "step": 4679
    },
    {
      "epoch": 1.94,
      "learning_rate": 6.98392323787329e-05,
      "loss": 0.6858,
      "step": 4680
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.982687539650936e-05,
      "loss": 0.8602,
      "step": 4681
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.981451697718665e-05,
      "loss": 0.4641,
      "step": 4682
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.980215712166054e-05,
      "loss": 0.2978,
      "step": 4683
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.978979583082689e-05,
      "loss": 0.8859,
      "step": 4684
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.977743310558166e-05,
      "loss": 0.7488,
      "step": 4685
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.976506894682098e-05,
      "loss": 0.4805,
      "step": 4686
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.975270335544098e-05,
      "loss": 0.7404,
      "step": 4687
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.974033633233798e-05,
      "loss": 0.5984,
      "step": 4688
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.972796787840837e-05,
      "loss": 0.5371,
      "step": 4689
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.97155979945486e-05,
      "loss": 0.9242,
      "step": 4690
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.970322668165536e-05,
      "loss": 0.826,
      "step": 4691
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.969085394062526e-05,
      "loss": 0.4646,
      "step": 4692
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.967847977235515e-05,
      "loss": 0.6485,
      "step": 4693
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.966610417774195e-05,
      "loss": 0.5444,
      "step": 4694
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.965372715768264e-05,
      "loss": 0.5824,
      "step": 4695
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.964134871307435e-05,
      "loss": 0.9863,
      "step": 4696
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.962896884481429e-05,
      "loss": 0.7471,
      "step": 4697
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.961658755379982e-05,
      "loss": 0.5209,
      "step": 4698
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.960420484092832e-05,
      "loss": 0.7357,
      "step": 4699
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.959182070709734e-05,
      "loss": 0.8121,
      "step": 4700
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.957943515320452e-05,
      "loss": 0.83,
      "step": 4701
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.956704818014757e-05,
      "loss": 0.8127,
      "step": 4702
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.955465978882434e-05,
      "loss": 0.4563,
      "step": 4703
    },
    {
      "epoch": 1.95,
      "learning_rate": 6.954226998013275e-05,
      "loss": 0.6175,
      "step": 4704
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.95298787549709e-05,
      "loss": 0.7133,
      "step": 4705
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.951748611423688e-05,
      "loss": 0.4028,
      "step": 4706
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.950509205882896e-05,
      "loss": 0.497,
      "step": 4707
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.94926965896455e-05,
      "loss": 0.6623,
      "step": 4708
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.948029970758491e-05,
      "loss": 0.489,
      "step": 4709
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.94679014135458e-05,
      "loss": 0.8331,
      "step": 4710
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.945550170842679e-05,
      "loss": 0.5945,
      "step": 4711
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.944310059312666e-05,
      "loss": 0.6508,
      "step": 4712
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.943069806854427e-05,
      "loss": 0.664,
      "step": 4713
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.941829413557858e-05,
      "loss": 0.7292,
      "step": 4714
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.940588879512866e-05,
      "loss": 0.9757,
      "step": 4715
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.939348204809367e-05,
      "loss": 0.3421,
      "step": 4716
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.93810738953729e-05,
      "loss": 0.4397,
      "step": 4717
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.936866433786569e-05,
      "loss": 0.7736,
      "step": 4718
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.935625337647154e-05,
      "loss": 0.5699,
      "step": 4719
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.934384101209002e-05,
      "loss": 0.505,
      "step": 4720
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.93314272456208e-05,
      "loss": 0.6546,
      "step": 4721
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.931901207796369e-05,
      "loss": 0.7878,
      "step": 4722
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.930659551001853e-05,
      "loss": 0.1944,
      "step": 4723
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.929417754268532e-05,
      "loss": 0.4181,
      "step": 4724
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.928175817686415e-05,
      "loss": 0.7196,
      "step": 4725
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.92693374134552e-05,
      "loss": 0.4679,
      "step": 4726
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.925691525335875e-05,
      "loss": 0.4445,
      "step": 4727
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.92444916974752e-05,
      "loss": 0.5161,
      "step": 4728
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.923206674670502e-05,
      "loss": 0.6464,
      "step": 4729
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.921964040194881e-05,
      "loss": 0.6003,
      "step": 4730
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.920721266410729e-05,
      "loss": 0.7456,
      "step": 4731
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.91947835340812e-05,
      "loss": 0.6439,
      "step": 4732
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.918235301277148e-05,
      "loss": 0.3743,
      "step": 4733
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.916992110107909e-05,
      "loss": 0.8454,
      "step": 4734
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.915748779990514e-05,
      "loss": 0.5976,
      "step": 4735
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.914505311015083e-05,
      "loss": 0.5203,
      "step": 4736
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.913261703271743e-05,
      "loss": 0.6801,
      "step": 4737
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.912017956850636e-05,
      "loss": 0.4317,
      "step": 4738
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.910774071841912e-05,
      "loss": 1.1487,
      "step": 4739
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.909530048335729e-05,
      "loss": 0.871,
      "step": 4740
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.90828588642226e-05,
      "loss": 0.5027,
      "step": 4741
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.907041586191679e-05,
      "loss": 0.9874,
      "step": 4742
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.905797147734182e-05,
      "loss": 0.6407,
      "step": 4743
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.904552571139965e-05,
      "loss": 0.3832,
      "step": 4744
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.903307856499239e-05,
      "loss": 0.9625,
      "step": 4745
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.902063003902224e-05,
      "loss": 0.3795,
      "step": 4746
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.900818013439151e-05,
      "loss": 0.484,
      "step": 4747
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.899572885200258e-05,
      "loss": 0.6064,
      "step": 4748
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.898327619275795e-05,
      "loss": 0.6025,
      "step": 4749
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.897082215756025e-05,
      "loss": 1.4166,
      "step": 4750
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.895836674731212e-05,
      "loss": 0.9205,
      "step": 4751
    },
    {
      "epoch": 1.97,
      "learning_rate": 6.894590996291642e-05,
      "loss": 0.9212,
      "step": 4752
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.8933451805276e-05,
      "loss": 0.6396,
      "step": 4753
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.892099227529388e-05,
      "loss": 0.5732,
      "step": 4754
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.890853137387315e-05,
      "loss": 0.9694,
      "step": 4755
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.8896069101917e-05,
      "loss": 0.7978,
      "step": 4756
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.888360546032875e-05,
      "loss": 0.7227,
      "step": 4757
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.887114045001177e-05,
      "loss": 0.8244,
      "step": 4758
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.885867407186957e-05,
      "loss": 0.6342,
      "step": 4759
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.884620632680571e-05,
      "loss": 0.5176,
      "step": 4760
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.883373721572392e-05,
      "loss": 0.6098,
      "step": 4761
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.882126673952798e-05,
      "loss": 0.9349,
      "step": 4762
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.880879489912175e-05,
      "loss": 0.3614,
      "step": 4763
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.879632169540927e-05,
      "loss": 0.6129,
      "step": 4764
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.87838471292946e-05,
      "loss": 0.6864,
      "step": 4765
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.877137120168192e-05,
      "loss": 0.6591,
      "step": 4766
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.875889391347553e-05,
      "loss": 0.8729,
      "step": 4767
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.874641526557981e-05,
      "loss": 0.7488,
      "step": 4768
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.873393525889924e-05,
      "loss": 0.7285,
      "step": 4769
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.872145389433838e-05,
      "loss": 0.457,
      "step": 4770
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.870897117280195e-05,
      "loss": 0.5185,
      "step": 4771
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.86964870951947e-05,
      "loss": 0.8753,
      "step": 4772
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.868400166242151e-05,
      "loss": 0.5534,
      "step": 4773
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.867151487538735e-05,
      "loss": 0.609,
      "step": 4774
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.865902673499731e-05,
      "loss": 0.8693,
      "step": 4775
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.864653724215654e-05,
      "loss": 1.001,
      "step": 4776
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.863404639777032e-05,
      "loss": 0.6589,
      "step": 4777
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.862155420274402e-05,
      "loss": 0.6847,
      "step": 4778
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.860906065798306e-05,
      "loss": 0.4839,
      "step": 4779
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.859656576439309e-05,
      "loss": 0.5832,
      "step": 4780
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.85840695228797e-05,
      "loss": 0.6573,
      "step": 4781
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.857157193434864e-05,
      "loss": 0.4598,
      "step": 4782
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.855907299970583e-05,
      "loss": 0.553,
      "step": 4783
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.854657271985717e-05,
      "loss": 0.688,
      "step": 4784
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.853407109570871e-05,
      "loss": 0.9346,
      "step": 4785
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.852156812816661e-05,
      "loss": 0.6217,
      "step": 4786
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.850906381813713e-05,
      "loss": 0.6964,
      "step": 4787
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.849655816652659e-05,
      "loss": 0.82,
      "step": 4788
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.848405117424143e-05,
      "loss": 0.5997,
      "step": 4789
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.847154284218821e-05,
      "loss": 0.6896,
      "step": 4790
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.845903317127352e-05,
      "loss": 1.1809,
      "step": 4791
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.844652216240416e-05,
      "loss": 0.4683,
      "step": 4792
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.843400981648689e-05,
      "loss": 0.9061,
      "step": 4793
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.842149613442865e-05,
      "loss": 0.8442,
      "step": 4794
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.84089811171365e-05,
      "loss": 0.7228,
      "step": 4795
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.839646476551751e-05,
      "loss": 0.6781,
      "step": 4796
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.838394708047891e-05,
      "loss": 0.629,
      "step": 4797
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.837142806292804e-05,
      "loss": 0.6082,
      "step": 4798
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.835890771377227e-05,
      "loss": 0.7132,
      "step": 4799
    },
    {
      "epoch": 1.99,
      "learning_rate": 6.834638603391912e-05,
      "loss": 0.9135,
      "step": 4800
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.83338630242762e-05,
      "loss": 0.5752,
      "step": 4801
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.83213386857512e-05,
      "loss": 0.4809,
      "step": 4802
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.83088130192519e-05,
      "loss": 0.951,
      "step": 4803
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.82962860256862e-05,
      "loss": 0.4714,
      "step": 4804
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.82837577059621e-05,
      "loss": 1.0659,
      "step": 4805
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.827122806098767e-05,
      "loss": 0.5172,
      "step": 4806
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.82586970916711e-05,
      "loss": 0.9011,
      "step": 4807
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.824616479892063e-05,
      "loss": 0.5255,
      "step": 4808
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.823363118364468e-05,
      "loss": 0.7793,
      "step": 4809
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.822109624675168e-05,
      "loss": 0.5049,
      "step": 4810
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.82085599891502e-05,
      "loss": 0.925,
      "step": 4811
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.819602241174891e-05,
      "loss": 0.7384,
      "step": 4812
    },
    {
      "epoch": 2.0,
      "learning_rate": 6.818348351545655e-05,
      "loss": 0.5332,
      "step": 4813
    },
    {
      "epoch": 2.0,
      "eval_loss": 1.0745819807052612,
      "eval_runtime": 340.267,
      "eval_samples_per_second": 6.945,
      "eval_steps_per_second": 1.737,
      "step": 4813
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 12030,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 8.68908612144384e+16,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}