{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.9994086339444115,
  "global_step": 3804,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.7391304347826088e-07,
      "loss": 1.8444,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.4782608695652175e-07,
      "loss": 1.8409,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.217391304347826e-07,
      "loss": 1.1537,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.956521739130435e-07,
      "loss": 1.1071,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.695652173913044e-07,
      "loss": 1.08,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.0434782608695653e-06,
      "loss": 0.9888,
      "step": 6
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2173913043478262e-06,
      "loss": 1.0178,
      "step": 7
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.391304347826087e-06,
      "loss": 0.9369,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.565217391304348e-06,
      "loss": 0.9617,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.7391304347826088e-06,
      "loss": 0.9597,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9130434782608697e-06,
      "loss": 0.9355,
      "step": 11
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.0869565217391305e-06,
      "loss": 0.9507,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.2608695652173913e-06,
      "loss": 0.9251,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.4347826086956525e-06,
      "loss": 0.8563,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6086956521739132e-06,
      "loss": 0.8598,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.782608695652174e-06,
      "loss": 0.9135,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.956521739130435e-06,
      "loss": 0.9001,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.130434782608696e-06,
      "loss": 0.8269,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.3043478260869567e-06,
      "loss": 0.88,
      "step": 19
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.4782608695652175e-06,
      "loss": 0.863,
      "step": 20
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.6521739130434787e-06,
      "loss": 0.8473,
      "step": 21
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.8260869565217395e-06,
      "loss": 0.8219,
      "step": 22
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.8111,
      "step": 23
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.173913043478261e-06,
      "loss": 0.7826,
      "step": 24
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.347826086956522e-06,
      "loss": 0.8028,
      "step": 25
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.5217391304347826e-06,
      "loss": 0.8146,
      "step": 26
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.695652173913044e-06,
      "loss": 0.7722,
      "step": 27
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.869565217391305e-06,
      "loss": 0.7723,
      "step": 28
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.043478260869565e-06,
      "loss": 0.7455,
      "step": 29
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.2173913043478265e-06,
      "loss": 0.7708,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.391304347826088e-06,
      "loss": 0.7726,
      "step": 31
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.565217391304348e-06,
      "loss": 0.7273,
      "step": 32
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.739130434782609e-06,
      "loss": 0.7327,
      "step": 33
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.91304347826087e-06,
      "loss": 0.746,
      "step": 34
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.086956521739132e-06,
      "loss": 0.783,
      "step": 35
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.260869565217392e-06,
      "loss": 0.7776,
      "step": 36
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.434782608695652e-06,
      "loss": 0.7752,
      "step": 37
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.6086956521739135e-06,
      "loss": 0.758,
      "step": 38
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.782608695652174e-06,
      "loss": 0.7572,
      "step": 39
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.956521739130435e-06,
      "loss": 0.7572,
      "step": 40
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.130434782608696e-06,
      "loss": 0.7766,
      "step": 41
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.304347826086957e-06,
      "loss": 0.7276,
      "step": 42
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.478260869565218e-06,
      "loss": 0.7448,
      "step": 43
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.652173913043479e-06,
      "loss": 0.8111,
      "step": 44
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.82608695652174e-06,
      "loss": 0.6756,
      "step": 45
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.7062,
      "step": 46
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.173913043478263e-06,
      "loss": 0.7172,
      "step": 47
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.347826086956522e-06,
      "loss": 0.7153,
      "step": 48
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.521739130434783e-06,
      "loss": 0.719,
      "step": 49
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.695652173913044e-06,
      "loss": 0.6758,
      "step": 50
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.869565217391306e-06,
      "loss": 0.7054,
      "step": 51
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.043478260869565e-06,
      "loss": 0.7084,
      "step": 52
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.217391304347826e-06,
      "loss": 0.7191,
      "step": 53
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.391304347826087e-06,
      "loss": 0.705,
      "step": 54
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.565217391304349e-06,
      "loss": 0.7231,
      "step": 55
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.73913043478261e-06,
      "loss": 0.6881,
      "step": 56
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.913043478260871e-06,
      "loss": 0.7033,
      "step": 57
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.008695652173913e-05,
      "loss": 0.7025,
      "step": 58
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0260869565217393e-05,
      "loss": 0.6851,
      "step": 59
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0434782608695653e-05,
      "loss": 0.6887,
      "step": 60
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0608695652173914e-05,
      "loss": 0.6865,
      "step": 61
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0782608695652175e-05,
      "loss": 0.6935,
      "step": 62
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0956521739130435e-05,
      "loss": 0.7024,
      "step": 63
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1130434782608696e-05,
      "loss": 0.6764,
      "step": 64
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1304347826086957e-05,
      "loss": 0.663,
      "step": 65
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1478260869565218e-05,
      "loss": 0.6779,
      "step": 66
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1652173913043478e-05,
      "loss": 0.6871,
      "step": 67
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.182608695652174e-05,
      "loss": 0.6732,
      "step": 68
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.2e-05,
      "loss": 0.676,
      "step": 69
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2173913043478263e-05,
      "loss": 0.7021,
      "step": 70
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2347826086956523e-05,
      "loss": 0.669,
      "step": 71
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2521739130434784e-05,
      "loss": 0.6857,
      "step": 72
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2695652173913045e-05,
      "loss": 0.7095,
      "step": 73
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2869565217391305e-05,
      "loss": 0.7006,
      "step": 74
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3043478260869566e-05,
      "loss": 0.6747,
      "step": 75
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3217391304347827e-05,
      "loss": 0.6447,
      "step": 76
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3391304347826088e-05,
      "loss": 0.7011,
      "step": 77
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3565217391304348e-05,
      "loss": 0.6584,
      "step": 78
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.373913043478261e-05,
      "loss": 0.6836,
      "step": 79
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.391304347826087e-05,
      "loss": 0.6681,
      "step": 80
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.4086956521739133e-05,
      "loss": 0.6594,
      "step": 81
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.4260869565217392e-05,
      "loss": 0.6672,
      "step": 82
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4434782608695654e-05,
      "loss": 0.6742,
      "step": 83
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4608695652173915e-05,
      "loss": 0.6562,
      "step": 84
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4782608695652174e-05,
      "loss": 0.689,
      "step": 85
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4956521739130436e-05,
      "loss": 0.6697,
      "step": 86
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.5130434782608697e-05,
      "loss": 0.6659,
      "step": 87
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.5304347826086958e-05,
      "loss": 0.6764,
      "step": 88
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.5478260869565217e-05,
      "loss": 0.6731,
      "step": 89
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.565217391304348e-05,
      "loss": 0.6731,
      "step": 90
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.582608695652174e-05,
      "loss": 0.6513,
      "step": 91
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.6699,
      "step": 92
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6173913043478262e-05,
      "loss": 0.6444,
      "step": 93
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6347826086956525e-05,
      "loss": 0.6598,
      "step": 94
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6521739130434785e-05,
      "loss": 0.6674,
      "step": 95
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.6695652173913044e-05,
      "loss": 0.6761,
      "step": 96
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.6869565217391307e-05,
      "loss": 0.6556,
      "step": 97
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.7043478260869566e-05,
      "loss": 0.662,
      "step": 98
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.721739130434783e-05,
      "loss": 0.6641,
      "step": 99
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.739130434782609e-05,
      "loss": 0.6543,
      "step": 100
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.756521739130435e-05,
      "loss": 0.6503,
      "step": 101
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.773913043478261e-05,
      "loss": 0.6458,
      "step": 102
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.791304347826087e-05,
      "loss": 0.6893,
      "step": 103
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.808695652173913e-05,
      "loss": 0.6707,
      "step": 104
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.8260869565217393e-05,
      "loss": 0.6702,
      "step": 105
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.8434782608695653e-05,
      "loss": 0.65,
      "step": 106
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.8608695652173912e-05,
      "loss": 0.6735,
      "step": 107
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.8782608695652175e-05,
      "loss": 0.6161,
      "step": 108
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.8956521739130434e-05,
      "loss": 0.676,
      "step": 109
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9130434782608697e-05,
      "loss": 0.6576,
      "step": 110
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9304347826086957e-05,
      "loss": 0.6302,
      "step": 111
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.947826086956522e-05,
      "loss": 0.6782,
      "step": 112
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.965217391304348e-05,
      "loss": 0.6671,
      "step": 113
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9826086956521742e-05,
      "loss": 0.6862,
      "step": 114
    },
    {
      "epoch": 0.09,
      "learning_rate": 2e-05,
      "loss": 0.6872,
      "step": 115
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.99999963737945e-05,
      "loss": 0.6849,
      "step": 116
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9999985495180608e-05,
      "loss": 0.676,
      "step": 117
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9999967364166225e-05,
      "loss": 0.6412,
      "step": 118
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9999941980764498e-05,
      "loss": 0.638,
      "step": 119
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.999990934499384e-05,
      "loss": 0.6091,
      "step": 120
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999869456877913e-05,
      "loss": 0.6393,
      "step": 121
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999822316445652e-05,
      "loss": 0.6376,
      "step": 122
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.999976792373124e-05,
      "loss": 0.6821,
      "step": 123
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999706278774126e-05,
      "loss": 0.6396,
      "step": 124
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999637381619017e-05,
      "loss": 0.6371,
      "step": 125
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999561232315886e-05,
      "loss": 0.6095,
      "step": 126
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999477830919952e-05,
      "loss": 0.6624,
      "step": 127
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999387177491703e-05,
      "loss": 0.6563,
      "step": 128
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999289272096886e-05,
      "loss": 0.6237,
      "step": 129
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999184114806506e-05,
      "loss": 0.6709,
      "step": 130
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.999907170569683e-05,
      "loss": 0.6401,
      "step": 131
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9998952044849376e-05,
      "loss": 0.6329,
      "step": 132
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9998825132350928e-05,
      "loss": 0.6069,
      "step": 133
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999869096829353e-05,
      "loss": 0.6238,
      "step": 134
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999854955277449e-05,
      "loss": 0.6162,
      "step": 135
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9998400885896355e-05,
      "loss": 0.6236,
      "step": 136
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999824496776695e-05,
      "loss": 0.6335,
      "step": 137
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9998081798499358e-05,
      "loss": 0.6356,
      "step": 138
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999791137821191e-05,
      "loss": 0.615,
      "step": 139
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9997733707028203e-05,
      "loss": 0.637,
      "step": 140
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9997548785077093e-05,
      "loss": 0.6373,
      "step": 141
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9997356612492694e-05,
      "loss": 0.63,
      "step": 142
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9997157189414373e-05,
      "loss": 0.6144,
      "step": 143
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9996950515986767e-05,
      "loss": 0.6388,
      "step": 144
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9996736592359755e-05,
      "loss": 0.6536,
      "step": 145
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9996515418688493e-05,
      "loss": 0.6483,
      "step": 146
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9996286995133374e-05,
      "loss": 0.6691,
      "step": 147
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999605132186007e-05,
      "loss": 0.6349,
      "step": 148
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9995808399039496e-05,
      "loss": 0.6509,
      "step": 149
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999555822684783e-05,
      "loss": 0.6359,
      "step": 150
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9995300805466508e-05,
      "loss": 0.6237,
      "step": 151
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9995036135082225e-05,
      "loss": 0.6392,
      "step": 152
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9994764215886925e-05,
      "loss": 0.6281,
      "step": 153
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9994485048077823e-05,
      "loss": 0.5934,
      "step": 154
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9994198631857377e-05,
      "loss": 0.6216,
      "step": 155
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9993904967433307e-05,
      "loss": 0.6455,
      "step": 156
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999360405501859e-05,
      "loss": 0.6086,
      "step": 157
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9993295894831467e-05,
      "loss": 0.6123,
      "step": 158
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9992980487095426e-05,
      "loss": 0.59,
      "step": 159
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.999265783203921e-05,
      "loss": 0.6211,
      "step": 160
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.999232792989682e-05,
      "loss": 0.6124,
      "step": 161
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.999199078090752e-05,
      "loss": 0.6213,
      "step": 162
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9991646385315826e-05,
      "loss": 0.6007,
      "step": 163
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.99912947433715e-05,
      "loss": 0.6445,
      "step": 164
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9990935855329575e-05,
      "loss": 0.6368,
      "step": 165
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9990569721450324e-05,
      "loss": 0.6494,
      "step": 166
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.999019634199929e-05,
      "loss": 0.6313,
      "step": 167
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9989815717247256e-05,
      "loss": 0.6174,
      "step": 168
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.998942784747027e-05,
      "loss": 0.634,
      "step": 169
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9989032732949632e-05,
      "loss": 0.6045,
      "step": 170
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9988630373971896e-05,
      "loss": 0.619,
      "step": 171
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9988220770828867e-05,
      "loss": 0.6168,
      "step": 172
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9987803923817606e-05,
      "loss": 0.6201,
      "step": 173
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998737983324043e-05,
      "loss": 0.5922,
      "step": 174
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.99869484994049e-05,
      "loss": 0.6352,
      "step": 175
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998650992262385e-05,
      "loss": 0.6246,
      "step": 176
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998606410321534e-05,
      "loss": 0.6337,
      "step": 177
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9985611041502704e-05,
      "loss": 0.609,
      "step": 178
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998515073781452e-05,
      "loss": 0.6211,
      "step": 179
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9984683192484614e-05,
      "loss": 0.614,
      "step": 180
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9984208405852082e-05,
      "loss": 0.6105,
      "step": 181
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998372637826125e-05,
      "loss": 0.6217,
      "step": 182
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9983237110061696e-05,
      "loss": 0.622,
      "step": 183
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9982740601608275e-05,
      "loss": 0.5823,
      "step": 184
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9982236853261067e-05,
      "loss": 0.6055,
      "step": 185
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9981725865385404e-05,
      "loss": 0.6129,
      "step": 186
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9981207638351886e-05,
      "loss": 0.5901,
      "step": 187
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9980682172536346e-05,
      "loss": 0.6094,
      "step": 188
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.998014946831988e-05,
      "loss": 0.6236,
      "step": 189
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9979609526088822e-05,
      "loss": 0.5961,
      "step": 190
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9979062346234762e-05,
      "loss": 0.605,
      "step": 191
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9978507929154534e-05,
      "loss": 0.6218,
      "step": 192
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9977946275250227e-05,
      "loss": 0.6045,
      "step": 193
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9977377384929176e-05,
      "loss": 0.6127,
      "step": 194
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9976801258603964e-05,
      "loss": 0.5887,
      "step": 195
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.997621789669242e-05,
      "loss": 0.6042,
      "step": 196
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.997562729961762e-05,
      "loss": 0.63,
      "step": 197
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9975029467807895e-05,
      "loss": 0.6143,
      "step": 198
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.997442440169681e-05,
      "loss": 0.5989,
      "step": 199
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.997381210172319e-05,
      "loss": 0.6161,
      "step": 200
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9973192568331096e-05,
      "loss": 0.5979,
      "step": 201
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9972565801969842e-05,
      "loss": 0.6018,
      "step": 202
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9971931803093983e-05,
      "loss": 0.577,
      "step": 203
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.997129057216332e-05,
      "loss": 0.5763,
      "step": 204
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9970642109642906e-05,
      "loss": 0.6051,
      "step": 205
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9969986416003026e-05,
      "loss": 0.5868,
      "step": 206
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9969323491719216e-05,
      "loss": 0.5797,
      "step": 207
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9968653337272262e-05,
      "loss": 0.5807,
      "step": 208
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9967975953148182e-05,
      "loss": 0.6246,
      "step": 209
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9967291339838244e-05,
      "loss": 0.6059,
      "step": 210
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996659949783896e-05,
      "loss": 0.621,
      "step": 211
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996590042765208e-05,
      "loss": 0.6115,
      "step": 212
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9965194129784597e-05,
      "loss": 0.6164,
      "step": 213
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996448060474875e-05,
      "loss": 0.6111,
      "step": 214
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9963759853062017e-05,
      "loss": 0.6185,
      "step": 215
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9963031875247113e-05,
      "loss": 0.5983,
      "step": 216
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9962296671832e-05,
      "loss": 0.5947,
      "step": 217
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9961554243349882e-05,
      "loss": 0.5964,
      "step": 218
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996080459033919e-05,
      "loss": 0.5949,
      "step": 219
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996004771334361e-05,
      "loss": 0.616,
      "step": 220
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.995928361291206e-05,
      "loss": 0.5814,
      "step": 221
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.995851228959869e-05,
      "loss": 0.6062,
      "step": 222
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9957733743962902e-05,
      "loss": 0.6388,
      "step": 223
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.995694797656933e-05,
      "loss": 0.5948,
      "step": 224
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.995615498798784e-05,
      "loss": 0.5786,
      "step": 225
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9955354778793547e-05,
      "loss": 0.6096,
      "step": 226
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9954547349566783e-05,
      "loss": 0.5985,
      "step": 227
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9953732700893143e-05,
      "loss": 0.608,
      "step": 228
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9952910833363435e-05,
      "loss": 0.5849,
      "step": 229
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.995208174757371e-05,
      "loss": 0.6022,
      "step": 230
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9951245444125266e-05,
      "loss": 0.6105,
      "step": 231
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9950401923624615e-05,
      "loss": 0.59,
      "step": 232
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9949551186683513e-05,
      "loss": 0.6053,
      "step": 233
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.994869323391895e-05,
      "loss": 0.5863,
      "step": 234
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9947828065953155e-05,
      "loss": 0.6069,
      "step": 235
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.994695568341358e-05,
      "loss": 0.5908,
      "step": 236
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9946076086932905e-05,
      "loss": 0.5834,
      "step": 237
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.994518927714906e-05,
      "loss": 0.6016,
      "step": 238
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9944295254705187e-05,
      "loss": 0.5925,
      "step": 239
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9943394020249674e-05,
      "loss": 0.5722,
      "step": 240
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.994248557443613e-05,
      "loss": 0.6184,
      "step": 241
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9941569917923402e-05,
      "loss": 0.5669,
      "step": 242
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.994064705137556e-05,
      "loss": 0.5862,
      "step": 243
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.99397169754619e-05,
      "loss": 0.572,
      "step": 244
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9938779690856954e-05,
      "loss": 0.5732,
      "step": 245
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9937835198240482e-05,
      "loss": 0.5778,
      "step": 246
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9936883498297468e-05,
      "loss": 0.5789,
      "step": 247
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.993592459171812e-05,
      "loss": 0.5935,
      "step": 248
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9934958479197882e-05,
      "loss": 0.617,
      "step": 249
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9933985161437414e-05,
      "loss": 0.5768,
      "step": 250
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9933004639142606e-05,
      "loss": 0.5967,
      "step": 251
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9932016913024578e-05,
      "loss": 0.5866,
      "step": 252
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9931021983799664e-05,
      "loss": 0.548,
      "step": 253
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.993001985218943e-05,
      "loss": 0.6138,
      "step": 254
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9929010518920667e-05,
      "loss": 0.5599,
      "step": 255
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9927993984725377e-05,
      "loss": 0.5792,
      "step": 256
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9926970250340796e-05,
      "loss": 0.5804,
      "step": 257
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.992593931650938e-05,
      "loss": 0.5632,
      "step": 258
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.99249011839788e-05,
      "loss": 0.5616,
      "step": 259
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9923855853501963e-05,
      "loss": 0.5651,
      "step": 260
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9922803325836973e-05,
      "loss": 0.562,
      "step": 261
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.992174360174717e-05,
      "loss": 0.5936,
      "step": 262
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9920676682001115e-05,
      "loss": 0.5723,
      "step": 263
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9919602567372575e-05,
      "loss": 0.5731,
      "step": 264
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9918521258640545e-05,
      "loss": 0.6082,
      "step": 265
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.991743275658924e-05,
      "loss": 0.608,
      "step": 266
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9916337062008076e-05,
      "loss": 0.5457,
      "step": 267
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.99152341756917e-05,
      "loss": 0.5823,
      "step": 268
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9914124098439976e-05,
      "loss": 0.5853,
      "step": 269
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9913006831057967e-05,
      "loss": 0.5972,
      "step": 270
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9911882374355975e-05,
      "loss": 0.5607,
      "step": 271
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9910750729149487e-05,
      "loss": 0.6046,
      "step": 272
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9909611896259233e-05,
      "loss": 0.5769,
      "step": 273
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.990846587651113e-05,
      "loss": 0.5871,
      "step": 274
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9907312670736322e-05,
      "loss": 0.571,
      "step": 275
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9906152279771162e-05,
      "loss": 0.6121,
      "step": 276
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9904984704457214e-05,
      "loss": 0.576,
      "step": 277
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9903809945641252e-05,
      "loss": 0.5921,
      "step": 278
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.990262800417526e-05,
      "loss": 0.5852,
      "step": 279
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.990143888091643e-05,
      "loss": 0.5862,
      "step": 280
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9900242576727157e-05,
      "loss": 0.5885,
      "step": 281
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.989903909247506e-05,
      "loss": 0.5743,
      "step": 282
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9897828429032946e-05,
      "loss": 0.566,
      "step": 283
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9896610587278845e-05,
      "loss": 0.5738,
      "step": 284
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.989538556809598e-05,
      "loss": 0.5784,
      "step": 285
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9894153372372792e-05,
      "loss": 0.5865,
      "step": 286
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9892914001002913e-05,
      "loss": 0.5748,
      "step": 287
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.989166745488519e-05,
      "loss": 0.5552,
      "step": 288
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9890413734923672e-05,
      "loss": 0.5564,
      "step": 289
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9889152842027607e-05,
      "loss": 0.5899,
      "step": 290
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9887884777111437e-05,
      "loss": 0.5745,
      "step": 291
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9886609541094826e-05,
      "loss": 0.5621,
      "step": 292
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9885327134902622e-05,
      "loss": 0.5749,
      "step": 293
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9884037559464885e-05,
      "loss": 0.5846,
      "step": 294
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.988274081571686e-05,
      "loss": 0.5581,
      "step": 295
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9881436904599005e-05,
      "loss": 0.5988,
      "step": 296
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9880125827056967e-05,
      "loss": 0.5809,
      "step": 297
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.987880758404159e-05,
      "loss": 0.5596,
      "step": 298
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9877482176508925e-05,
      "loss": 0.5493,
      "step": 299
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9876149605420206e-05,
      "loss": 0.5978,
      "step": 300
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9874809871741877e-05,
      "loss": 0.5773,
      "step": 301
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9873462976445553e-05,
      "loss": 0.5931,
      "step": 302
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9872108920508076e-05,
      "loss": 0.5588,
      "step": 303
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.987074770491145e-05,
      "loss": 0.5692,
      "step": 304
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9869379330642888e-05,
      "loss": 0.5826,
      "step": 305
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.986800379869479e-05,
      "loss": 0.5661,
      "step": 306
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9866621110064753e-05,
      "loss": 0.6025,
      "step": 307
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9865231265755556e-05,
      "loss": 0.593,
      "step": 308
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.986383426677517e-05,
      "loss": 0.5639,
      "step": 309
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9862430114136758e-05,
      "loss": 0.606,
      "step": 310
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.986101880885867e-05,
      "loss": 0.5423,
      "step": 311
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.985960035196444e-05,
      "loss": 0.558,
      "step": 312
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9858174744482795e-05,
      "loss": 0.5878,
      "step": 313
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9856741987447643e-05,
      "loss": 0.5819,
      "step": 314
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9855302081898076e-05,
      "loss": 0.5659,
      "step": 315
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9853855028878372e-05,
      "loss": 0.5764,
      "step": 316
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9852400829438e-05,
      "loss": 0.5681,
      "step": 317
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9850939484631598e-05,
      "loss": 0.6137,
      "step": 318
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9849470995518993e-05,
      "loss": 0.5453,
      "step": 319
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.98479953631652e-05,
      "loss": 0.5571,
      "step": 320
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9846512588640403e-05,
      "loss": 0.578,
      "step": 321
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9845022673019973e-05,
      "loss": 0.5597,
      "step": 322
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.984352561738446e-05,
      "loss": 0.552,
      "step": 323
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9842021422819587e-05,
      "loss": 0.5845,
      "step": 324
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.984051009041626e-05,
      "loss": 0.5642,
      "step": 325
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9838991621270555e-05,
      "loss": 0.5711,
      "step": 326
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.983746601648373e-05,
      "loss": 0.5697,
      "step": 327
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9835933277162222e-05,
      "loss": 0.552,
      "step": 328
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.983439340441763e-05,
      "loss": 0.5807,
      "step": 329
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9832846399366738e-05,
      "loss": 0.5564,
      "step": 330
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9831292263131492e-05,
      "loss": 0.5578,
      "step": 331
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.982973099683902e-05,
      "loss": 0.5585,
      "step": 332
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9828162601621614e-05,
      "loss": 0.591,
      "step": 333
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.982658707861674e-05,
      "loss": 0.5559,
      "step": 334
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.982500442896703e-05,
      "loss": 0.572,
      "step": 335
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.982341465382029e-05,
      "loss": 0.5698,
      "step": 336
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9821817754329488e-05,
      "loss": 0.548,
      "step": 337
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.982021373165276e-05,
      "loss": 0.5918,
      "step": 338
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9818602586953414e-05,
      "loss": 0.5863,
      "step": 339
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9816984321399916e-05,
      "loss": 0.5461,
      "step": 340
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9815358936165894e-05,
      "loss": 0.5578,
      "step": 341
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9813726432430148e-05,
      "loss": 0.5615,
      "step": 342
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.981208681137664e-05,
      "loss": 0.539,
      "step": 343
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9810440074194484e-05,
      "loss": 0.5909,
      "step": 344
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.980878622207797e-05,
      "loss": 0.5557,
      "step": 345
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9807125256226532e-05,
      "loss": 0.5596,
      "step": 346
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.980545717784477e-05,
      "loss": 0.5373,
      "step": 347
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.980378198814245e-05,
      "loss": 0.5612,
      "step": 348
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9802099688334483e-05,
      "loss": 0.5497,
      "step": 349
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9800410279640943e-05,
      "loss": 0.5541,
      "step": 350
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9798713763287057e-05,
      "loss": 0.5686,
      "step": 351
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9797010140503214e-05,
      "loss": 0.5756,
      "step": 352
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9795299412524948e-05,
      "loss": 0.5744,
      "step": 353
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9793581580592948e-05,
      "loss": 0.5512,
      "step": 354
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9791856645953054e-05,
      "loss": 0.5653,
      "step": 355
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9790124609856266e-05,
      "loss": 0.5772,
      "step": 356
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9788385473558724e-05,
      "loss": 0.5707,
      "step": 357
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9786639238321723e-05,
      "loss": 0.544,
      "step": 358
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9784885905411705e-05,
      "loss": 0.5477,
      "step": 359
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9783125476100254e-05,
      "loss": 0.5595,
      "step": 360
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.978135795166411e-05,
      "loss": 0.6036,
      "step": 361
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9779583333385154e-05,
      "loss": 0.5987,
      "step": 362
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.977780162255041e-05,
      "loss": 0.5654,
      "step": 363
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9776012820452052e-05,
      "loss": 0.5918,
      "step": 364
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9774216928387387e-05,
      "loss": 0.579,
      "step": 365
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9772413947658875e-05,
      "loss": 0.5399,
      "step": 366
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9770603879574108e-05,
      "loss": 0.5569,
      "step": 367
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9768786725445824e-05,
      "loss": 0.5499,
      "step": 368
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9766962486591902e-05,
      "loss": 0.5604,
      "step": 369
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9765131164335345e-05,
      "loss": 0.5601,
      "step": 370
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9763292760004312e-05,
      "loss": 0.5401,
      "step": 371
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.976144727493208e-05,
      "loss": 0.5841,
      "step": 372
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9759594710457084e-05,
      "loss": 0.539,
      "step": 373
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.975773506792287e-05,
      "loss": 0.5609,
      "step": 374
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9755868348678125e-05,
      "loss": 0.5494,
      "step": 375
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.975399455407668e-05,
      "loss": 0.5771,
      "step": 376
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.975211368547748e-05,
      "loss": 0.5254,
      "step": 377
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.975022574424462e-05,
      "loss": 0.5656,
      "step": 378
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9748330731747295e-05,
      "loss": 0.5466,
      "step": 379
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.974642864935986e-05,
      "loss": 0.5886,
      "step": 380
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.974451949846177e-05,
      "loss": 0.5236,
      "step": 381
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9742603280437635e-05,
      "loss": 0.5347,
      "step": 382
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9740679996677165e-05,
      "loss": 0.5524,
      "step": 383
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.973874964857521e-05,
      "loss": 0.6005,
      "step": 384
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9736812237531735e-05,
      "loss": 0.5782,
      "step": 385
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.973486776495183e-05,
      "loss": 0.5458,
      "step": 386
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9732916232245703e-05,
      "loss": 0.5349,
      "step": 387
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.973095764082869e-05,
      "loss": 0.5555,
      "step": 388
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9728991992121244e-05,
      "loss": 0.5491,
      "step": 389
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.972701928754893e-05,
      "loss": 0.5688,
      "step": 390
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.972503952854244e-05,
      "loss": 0.5353,
      "step": 391
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9723052716537564e-05,
      "loss": 0.5512,
      "step": 392
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9721058852975234e-05,
      "loss": 0.5332,
      "step": 393
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9719057939301477e-05,
      "loss": 0.5399,
      "step": 394
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9717049976967437e-05,
      "loss": 0.557,
      "step": 395
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.971503496742937e-05,
      "loss": 0.5777,
      "step": 396
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9713012912148644e-05,
      "loss": 0.5329,
      "step": 397
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.971098381259174e-05,
      "loss": 0.5426,
      "step": 398
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9708947670230238e-05,
      "loss": 0.5721,
      "step": 399
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9706904486540834e-05,
      "loss": 0.5604,
      "step": 400
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9704854263005332e-05,
      "loss": 0.5284,
      "step": 401
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9702797001110642e-05,
      "loss": 0.5534,
      "step": 402
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9700732702348763e-05,
      "loss": 0.5685,
      "step": 403
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9698661368216818e-05,
      "loss": 0.5513,
      "step": 404
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9696583000217024e-05,
      "loss": 0.5221,
      "step": 405
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9694497599856696e-05,
      "loss": 0.5454,
      "step": 406
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.969240516864825e-05,
      "loss": 0.511,
      "step": 407
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.969030570810921e-05,
      "loss": 0.5551,
      "step": 408
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9688199219762183e-05,
      "loss": 0.558,
      "step": 409
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9686085705134888e-05,
      "loss": 0.567,
      "step": 410
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9683965165760127e-05,
      "loss": 0.5536,
      "step": 411
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9681837603175804e-05,
      "loss": 0.5489,
      "step": 412
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9679703018924917e-05,
      "loss": 0.5538,
      "step": 413
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9677561414555552e-05,
      "loss": 0.5862,
      "step": 414
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.967541279162089e-05,
      "loss": 0.5681,
      "step": 415
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.96732571516792e-05,
      "loss": 0.5529,
      "step": 416
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.967109449629384e-05,
      "loss": 0.534,
      "step": 417
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.966892482703326e-05,
      "loss": 0.5313,
      "step": 418
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9666748145470987e-05,
      "loss": 0.5617,
      "step": 419
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9664564453185645e-05,
      "loss": 0.5557,
      "step": 420
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9662373751760934e-05,
      "loss": 0.5527,
      "step": 421
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9660176042785646e-05,
      "loss": 0.5864,
      "step": 422
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9657971327853644e-05,
      "loss": 0.5659,
      "step": 423
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9655759608563882e-05,
      "loss": 0.5394,
      "step": 424
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9653540886520387e-05,
      "loss": 0.5351,
      "step": 425
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.965131516333227e-05,
      "loss": 0.5403,
      "step": 426
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9649082440613717e-05,
      "loss": 0.5508,
      "step": 427
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9646842719983983e-05,
      "loss": 0.5375,
      "step": 428
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9644596003067415e-05,
      "loss": 0.5725,
      "step": 429
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.964234229149342e-05,
      "loss": 0.5349,
      "step": 430
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9640081586896485e-05,
      "loss": 0.5632,
      "step": 431
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9637813890916158e-05,
      "loss": 0.5605,
      "step": 432
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9635539205197078e-05,
      "loss": 0.5292,
      "step": 433
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.963325753138893e-05,
      "loss": 0.5402,
      "step": 434
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9630968871146483e-05,
      "loss": 0.5751,
      "step": 435
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9628673226129564e-05,
      "loss": 0.5274,
      "step": 436
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.962637059800307e-05,
      "loss": 0.5666,
      "step": 437
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9624060988436968e-05,
      "loss": 0.5293,
      "step": 438
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9621744399106267e-05,
      "loss": 0.5267,
      "step": 439
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.961942083169107e-05,
      "loss": 0.5462,
      "step": 440
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9617090287876515e-05,
      "loss": 0.5631,
      "step": 441
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.961475276935281e-05,
      "loss": 0.553,
      "step": 442
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.961240827781521e-05,
      "loss": 0.5163,
      "step": 443
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9610056814964053e-05,
      "loss": 0.5141,
      "step": 444
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9607698382504703e-05,
      "loss": 0.5391,
      "step": 445
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.96053329821476e-05,
      "loss": 0.5818,
      "step": 446
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9602960615608227e-05,
      "loss": 0.573,
      "step": 447
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9600581284607124e-05,
      "loss": 0.5388,
      "step": 448
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9598194990869877e-05,
      "loss": 0.5267,
      "step": 449
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.959580173612712e-05,
      "loss": 0.5261,
      "step": 450
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.959340152211455e-05,
      "loss": 0.5296,
      "step": 451
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9590994350572896e-05,
      "loss": 0.534,
      "step": 452
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9588580223247937e-05,
      "loss": 0.5446,
      "step": 453
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9586159141890494e-05,
      "loss": 0.5438,
      "step": 454
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.958373110825644e-05,
      "loss": 0.5606,
      "step": 455
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9581296124106682e-05,
      "loss": 0.5398,
      "step": 456
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9578854191207172e-05,
      "loss": 0.5406,
      "step": 457
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.95764053113289e-05,
      "loss": 0.5437,
      "step": 458
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9573949486247893e-05,
      "loss": 0.5673,
      "step": 459
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9571486717745222e-05,
      "loss": 0.5639,
      "step": 460
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9569017007606978e-05,
      "loss": 0.554,
      "step": 461
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9566540357624303e-05,
      "loss": 0.5257,
      "step": 462
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9564056769593364e-05,
      "loss": 0.5405,
      "step": 463
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.956156624531536e-05,
      "loss": 0.5224,
      "step": 464
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9559068786596526e-05,
      "loss": 0.5657,
      "step": 465
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9556564395248114e-05,
      "loss": 0.554,
      "step": 466
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9554053073086415e-05,
      "loss": 0.554,
      "step": 467
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.955153482193275e-05,
      "loss": 0.5264,
      "step": 468
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9549009643613448e-05,
      "loss": 0.5364,
      "step": 469
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9546477539959875e-05,
      "loss": 0.5592,
      "step": 470
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9543938512808422e-05,
      "loss": 0.5362,
      "step": 471
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.954139256400049e-05,
      "loss": 0.537,
      "step": 472
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9538839695382506e-05,
      "loss": 0.5578,
      "step": 473
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.953627990880592e-05,
      "loss": 0.526,
      "step": 474
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.953371320612719e-05,
      "loss": 0.5433,
      "step": 475
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9531139589207794e-05,
      "loss": 0.5491,
      "step": 476
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9528559059914224e-05,
      "loss": 0.5243,
      "step": 477
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.952597162011799e-05,
      "loss": 0.4991,
      "step": 478
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.952337727169561e-05,
      "loss": 0.5305,
      "step": 479
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.952077601652861e-05,
      "loss": 0.5122,
      "step": 480
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9518167856503526e-05,
      "loss": 0.5506,
      "step": 481
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9515552793511902e-05,
      "loss": 0.5291,
      "step": 482
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.951293082945029e-05,
      "loss": 0.5283,
      "step": 483
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9510301966220252e-05,
      "loss": 0.5651,
      "step": 484
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9507666205728338e-05,
      "loss": 0.5453,
      "step": 485
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.950502354988612e-05,
      "loss": 0.5368,
      "step": 486
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9502374000610152e-05,
      "loss": 0.5388,
      "step": 487
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9499717559821997e-05,
      "loss": 0.5232,
      "step": 488
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9497054229448223e-05,
      "loss": 0.5482,
      "step": 489
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9494384011420376e-05,
      "loss": 0.5227,
      "step": 490
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9491706907675016e-05,
      "loss": 0.5194,
      "step": 491
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9489022920153682e-05,
      "loss": 0.5495,
      "step": 492
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.948633205080292e-05,
      "loss": 0.5206,
      "step": 493
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9483634301574254e-05,
      "loss": 0.5461,
      "step": 494
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9480929674424202e-05,
      "loss": 0.519,
      "step": 495
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9478218171314272e-05,
      "loss": 0.515,
      "step": 496
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.947549979421096e-05,
      "loss": 0.5413,
      "step": 497
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.947277454508574e-05,
      "loss": 0.5133,
      "step": 498
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9470042425915078e-05,
      "loss": 0.548,
      "step": 499
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9467303438680414e-05,
      "loss": 0.5167,
      "step": 500
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9464557585368182e-05,
      "loss": 0.5451,
      "step": 501
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9461804867969784e-05,
      "loss": 0.5146,
      "step": 502
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9459045288481605e-05,
      "loss": 0.5426,
      "step": 503
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9456278848905e-05,
      "loss": 0.5231,
      "step": 504
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.945350555124631e-05,
      "loss": 0.5562,
      "step": 505
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.945072539751685e-05,
      "loss": 0.5161,
      "step": 506
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.944793838973289e-05,
      "loss": 0.5372,
      "step": 507
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.944514452991569e-05,
      "loss": 0.5314,
      "step": 508
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9442343820091465e-05,
      "loss": 0.508,
      "step": 509
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.943953626229141e-05,
      "loss": 0.529,
      "step": 510
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9436721858551684e-05,
      "loss": 0.535,
      "step": 511
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9433900610913406e-05,
      "loss": 0.5259,
      "step": 512
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9431072521422655e-05,
      "loss": 0.5069,
      "step": 513
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9428237592130487e-05,
      "loss": 0.5167,
      "step": 514
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9425395825092902e-05,
      "loss": 0.5279,
      "step": 515
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9422547222370873e-05,
      "loss": 0.505,
      "step": 516
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9419691786030314e-05,
      "loss": 0.5194,
      "step": 517
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.941682951814212e-05,
      "loss": 0.5614,
      "step": 518
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9413960420782107e-05,
      "loss": 0.5544,
      "step": 519
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9411084496031073e-05,
      "loss": 0.542,
      "step": 520
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.940820174597476e-05,
      "loss": 0.5379,
      "step": 521
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9405312172703846e-05,
      "loss": 0.5283,
      "step": 522
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9402415778313978e-05,
      "loss": 0.5422,
      "step": 523
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9399512564905736e-05,
      "loss": 0.5523,
      "step": 524
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.939660253458465e-05,
      "loss": 0.5378,
      "step": 525
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9393685689461193e-05,
      "loss": 0.5545,
      "step": 526
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9390762031650782e-05,
      "loss": 0.4967,
      "step": 527
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9387831563273775e-05,
      "loss": 0.5147,
      "step": 528
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9384894286455465e-05,
      "loss": 0.519,
      "step": 529
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.938195020332609e-05,
      "loss": 0.5084,
      "step": 530
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9378999316020816e-05,
      "loss": 0.5408,
      "step": 531
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9376041626679746e-05,
      "loss": 0.5009,
      "step": 532
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9373077137447928e-05,
      "loss": 0.5259,
      "step": 533
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.937010585047532e-05,
      "loss": 0.5246,
      "step": 534
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9367127767916828e-05,
      "loss": 0.5627,
      "step": 535
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9364142891932278e-05,
      "loss": 0.4974,
      "step": 536
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9361151224686426e-05,
      "loss": 0.5183,
      "step": 537
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.935815276834895e-05,
      "loss": 0.5143,
      "step": 538
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9355147525094453e-05,
      "loss": 0.5347,
      "step": 539
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9352135497102464e-05,
      "loss": 0.4962,
      "step": 540
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9349116686557427e-05,
      "loss": 0.5527,
      "step": 541
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9346091095648712e-05,
      "loss": 0.531,
      "step": 542
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9343058726570594e-05,
      "loss": 0.5295,
      "step": 543
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9340019581522277e-05,
      "loss": 0.5215,
      "step": 544
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9336973662707876e-05,
      "loss": 0.5531,
      "step": 545
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.933392097233641e-05,
      "loss": 0.561,
      "step": 546
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.933086151262182e-05,
      "loss": 0.4954,
      "step": 547
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9327795285782954e-05,
      "loss": 0.5181,
      "step": 548
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.932472229404356e-05,
      "loss": 0.5172,
      "step": 549
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.93216425396323e-05,
      "loss": 0.5108,
      "step": 550
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9318556024782742e-05,
      "loss": 0.5372,
      "step": 551
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9315462751733346e-05,
      "loss": 0.5328,
      "step": 552
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9312362722727492e-05,
      "loss": 0.5204,
      "step": 553
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9309255940013436e-05,
      "loss": 0.5223,
      "step": 554
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.930614240584435e-05,
      "loss": 0.5287,
      "step": 555
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9303022122478303e-05,
      "loss": 0.536,
      "step": 556
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.929989509217824e-05,
      "loss": 0.5236,
      "step": 557
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9296761317212024e-05,
      "loss": 0.5133,
      "step": 558
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9293620799852388e-05,
      "loss": 0.5082,
      "step": 559
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9290473542376972e-05,
      "loss": 0.5035,
      "step": 560
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9287319547068294e-05,
      "loss": 0.5278,
      "step": 561
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9284158816213758e-05,
      "loss": 0.5555,
      "step": 562
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9280991352105656e-05,
      "loss": 0.4965,
      "step": 563
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.927781715704117e-05,
      "loss": 0.5548,
      "step": 564
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.927463623332235e-05,
      "loss": 0.5364,
      "step": 565
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.927144858325613e-05,
      "loss": 0.534,
      "step": 566
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9268254209154333e-05,
      "loss": 0.5205,
      "step": 567
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9265053113333646e-05,
      "loss": 0.4974,
      "step": 568
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9261845298115637e-05,
      "loss": 0.5335,
      "step": 569
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.925863076582674e-05,
      "loss": 0.5271,
      "step": 570
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9255409518798278e-05,
      "loss": 0.5079,
      "step": 571
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.925218155936642e-05,
      "loss": 0.524,
      "step": 572
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.924894688987222e-05,
      "loss": 0.5285,
      "step": 573
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9245705512661594e-05,
      "loss": 0.5226,
      "step": 574
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9242457430085316e-05,
      "loss": 0.5184,
      "step": 575
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9239202644499034e-05,
      "loss": 0.5413,
      "step": 576
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9235941158263253e-05,
      "loss": 0.4951,
      "step": 577
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9232672973743333e-05,
      "loss": 0.5271,
      "step": 578
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.92293980933095e-05,
      "loss": 0.5263,
      "step": 579
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.922611651933683e-05,
      "loss": 0.4925,
      "step": 580
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9222828254205255e-05,
      "loss": 0.5169,
      "step": 581
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.921953330029956e-05,
      "loss": 0.5079,
      "step": 582
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9216231660009383e-05,
      "loss": 0.5294,
      "step": 583
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9212923335729206e-05,
      "loss": 0.5058,
      "step": 584
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9209608329858365e-05,
      "loss": 0.4958,
      "step": 585
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9206286644801034e-05,
      "loss": 0.5366,
      "step": 586
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9202958282966237e-05,
      "loss": 0.5142,
      "step": 587
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9199623246767844e-05,
      "loss": 0.5221,
      "step": 588
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9196281538624557e-05,
      "loss": 0.539,
      "step": 589
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9192933160959917e-05,
      "loss": 0.5495,
      "step": 590
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.918957811620231e-05,
      "loss": 0.524,
      "step": 591
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9186216406784947e-05,
      "loss": 0.5012,
      "step": 592
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9182848035145884e-05,
      "loss": 0.5359,
      "step": 593
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9179473003727996e-05,
      "loss": 0.5291,
      "step": 594
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9176091314979e-05,
      "loss": 0.5303,
      "step": 595
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.917270297135143e-05,
      "loss": 0.5308,
      "step": 596
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9169307975302662e-05,
      "loss": 0.5122,
      "step": 597
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9165906329294875e-05,
      "loss": 0.4933,
      "step": 598
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.916249803579509e-05,
      "loss": 0.5237,
      "step": 599
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.915908309727514e-05,
      "loss": 0.5308,
      "step": 600
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.915566151621168e-05,
      "loss": 0.5204,
      "step": 601
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9152233295086177e-05,
      "loss": 0.5154,
      "step": 602
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9148798436384917e-05,
      "loss": 0.5297,
      "step": 603
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9145356942599007e-05,
      "loss": 0.5172,
      "step": 604
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9141908816224356e-05,
      "loss": 0.5201,
      "step": 605
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.913845405976169e-05,
      "loss": 0.5371,
      "step": 606
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9134992675716536e-05,
      "loss": 0.5057,
      "step": 607
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9131524666599233e-05,
      "loss": 0.5103,
      "step": 608
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9128050034924924e-05,
      "loss": 0.5214,
      "step": 609
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.912456878321356e-05,
      "loss": 0.5224,
      "step": 610
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.912108091398988e-05,
      "loss": 0.5129,
      "step": 611
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9117586429783433e-05,
      "loss": 0.5466,
      "step": 612
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9114085333128564e-05,
      "loss": 0.5322,
      "step": 613
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.911057762656441e-05,
      "loss": 0.52,
      "step": 614
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9107063312634908e-05,
      "loss": 0.497,
      "step": 615
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9103542393888778e-05,
      "loss": 0.4946,
      "step": 616
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9100014872879536e-05,
      "loss": 0.4949,
      "step": 617
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.909648075216549e-05,
      "loss": 0.5194,
      "step": 618
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.909294003430972e-05,
      "loss": 0.5039,
      "step": 619
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.908939272188011e-05,
      "loss": 0.5163,
      "step": 620
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9085838817449312e-05,
      "loss": 0.5047,
      "step": 621
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9082278323594763e-05,
      "loss": 0.5054,
      "step": 622
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9078711242898686e-05,
      "loss": 0.5354,
      "step": 623
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.907513757794806e-05,
      "loss": 0.5117,
      "step": 624
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.907155733133467e-05,
      "loss": 0.5578,
      "step": 625
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.906797050565505e-05,
      "loss": 0.5279,
      "step": 626
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9064377103510513e-05,
      "loss": 0.4939,
      "step": 627
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9060777127507145e-05,
      "loss": 0.4969,
      "step": 628
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9057170580255793e-05,
      "loss": 0.5087,
      "step": 629
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9053557464372073e-05,
      "loss": 0.5368,
      "step": 630
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9049937782476372e-05,
      "loss": 0.5201,
      "step": 631
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9046311537193826e-05,
      "loss": 0.5211,
      "step": 632
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9042678731154337e-05,
      "loss": 0.5256,
      "step": 633
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.903903936699257e-05,
      "loss": 0.5061,
      "step": 634
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9035393447347937e-05,
      "loss": 0.473,
      "step": 635
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9031740974864607e-05,
      "loss": 0.5209,
      "step": 636
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.902808195219151e-05,
      "loss": 0.5066,
      "step": 637
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9024416381982315e-05,
      "loss": 0.5173,
      "step": 638
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9020744266895446e-05,
      "loss": 0.498,
      "step": 639
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.901706560959407e-05,
      "loss": 0.5137,
      "step": 640
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.90133804127461e-05,
      "loss": 0.4868,
      "step": 641
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.900968867902419e-05,
      "loss": 0.475,
      "step": 642
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9005990411105747e-05,
      "loss": 0.5177,
      "step": 643
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9002285611672898e-05,
      "loss": 0.4984,
      "step": 644
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8998574283412518e-05,
      "loss": 0.5136,
      "step": 645
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8994856429016218e-05,
      "loss": 0.5308,
      "step": 646
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8991132051180332e-05,
      "loss": 0.5151,
      "step": 647
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8987401152605938e-05,
      "loss": 0.5011,
      "step": 648
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8983663735998832e-05,
      "loss": 0.5252,
      "step": 649
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8979919804069545e-05,
      "loss": 0.5028,
      "step": 650
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8976169359533334e-05,
      "loss": 0.5304,
      "step": 651
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8972412405110167e-05,
      "loss": 0.5188,
      "step": 652
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8968648943524753e-05,
      "loss": 0.5071,
      "step": 653
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8964878977506496e-05,
      "loss": 0.5194,
      "step": 654
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.896110250978954e-05,
      "loss": 0.5142,
      "step": 655
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.895731954311273e-05,
      "loss": 0.5368,
      "step": 656
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.895353008021963e-05,
      "loss": 0.5052,
      "step": 657
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8949734123858516e-05,
      "loss": 0.5204,
      "step": 658
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8945931676782373e-05,
      "loss": 0.5276,
      "step": 659
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8942122741748883e-05,
      "loss": 0.4896,
      "step": 660
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8938307321520453e-05,
      "loss": 0.5268,
      "step": 661
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8934485418864174e-05,
      "loss": 0.4934,
      "step": 662
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8930657036551854e-05,
      "loss": 0.4858,
      "step": 663
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8926822177359988e-05,
      "loss": 0.4894,
      "step": 664
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8922980844069775e-05,
      "loss": 0.4996,
      "step": 665
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.891913303946711e-05,
      "loss": 0.5082,
      "step": 666
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8915278766342578e-05,
      "loss": 0.5065,
      "step": 667
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8911418027491453e-05,
      "loss": 0.5052,
      "step": 668
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8907550825713706e-05,
      "loss": 0.5275,
      "step": 669
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8903677163813987e-05,
      "loss": 0.5041,
      "step": 670
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8899797044601638e-05,
      "loss": 0.5069,
      "step": 671
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8895910470890678e-05,
      "loss": 0.5041,
      "step": 672
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8892017445499812e-05,
      "loss": 0.499,
      "step": 673
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8888117971252426e-05,
      "loss": 0.5014,
      "step": 674
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8884212050976568e-05,
      "loss": 0.5509,
      "step": 675
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.888029968750498e-05,
      "loss": 0.5072,
      "step": 676
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.887638088367507e-05,
      "loss": 0.5027,
      "step": 677
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.887245564232891e-05,
      "loss": 0.5132,
      "step": 678
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.886852396631325e-05,
      "loss": 0.4916,
      "step": 679
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8864585858479504e-05,
      "loss": 0.5311,
      "step": 680
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8860641321683746e-05,
      "loss": 0.4876,
      "step": 681
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.885669035878672e-05,
      "loss": 0.5256,
      "step": 682
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.885273297265382e-05,
      "loss": 0.5276,
      "step": 683
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8848769166155115e-05,
      "loss": 0.484,
      "step": 684
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8844798942165316e-05,
      "loss": 0.5014,
      "step": 685
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8840822303563785e-05,
      "loss": 0.4985,
      "step": 686
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.883683925323456e-05,
      "loss": 0.5338,
      "step": 687
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8832849794066294e-05,
      "loss": 0.5089,
      "step": 688
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.882885392895232e-05,
      "loss": 0.4916,
      "step": 689
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.88248516607906e-05,
      "loss": 0.5174,
      "step": 690
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8820842992483747e-05,
      "loss": 0.5227,
      "step": 691
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8816827926939003e-05,
      "loss": 0.4949,
      "step": 692
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.881280646706827e-05,
      "loss": 0.5145,
      "step": 693
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8808778615788062e-05,
      "loss": 0.5085,
      "step": 694
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8804744376019552e-05,
      "loss": 0.4897,
      "step": 695
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8800703750688536e-05,
      "loss": 0.5661,
      "step": 696
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.879665674272544e-05,
      "loss": 0.5076,
      "step": 697
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.879260335506532e-05,
      "loss": 0.503,
      "step": 698
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.878854359064786e-05,
      "loss": 0.5257,
      "step": 699
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.878447745241737e-05,
      "loss": 0.5477,
      "step": 700
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.878040494332278e-05,
      "loss": 0.5104,
      "step": 701
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8776326066317634e-05,
      "loss": 0.4869,
      "step": 702
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.877224082436011e-05,
      "loss": 0.5018,
      "step": 703
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.876814922041299e-05,
      "loss": 0.4869,
      "step": 704
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8764051257443672e-05,
      "loss": 0.5267,
      "step": 705
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.875994693842417e-05,
      "loss": 0.4895,
      "step": 706
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8755836266331102e-05,
      "loss": 0.5072,
      "step": 707
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8751719244145698e-05,
      "loss": 0.5009,
      "step": 708
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.874759587485379e-05,
      "loss": 0.5418,
      "step": 709
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8743466161445823e-05,
      "loss": 0.4774,
      "step": 710
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8739330106916826e-05,
      "loss": 0.4782,
      "step": 711
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8735187714266432e-05,
      "loss": 0.5238,
      "step": 712
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.873103898649889e-05,
      "loss": 0.4947,
      "step": 713
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.872688392662301e-05,
      "loss": 0.5247,
      "step": 714
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8722722537652225e-05,
      "loss": 0.5073,
      "step": 715
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.871855482260454e-05,
      "loss": 0.4958,
      "step": 716
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8714380784502553e-05,
      "loss": 0.4673,
      "step": 717
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.871020042637345e-05,
      "loss": 0.4703,
      "step": 718
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8706013751249e-05,
      "loss": 0.5186,
      "step": 719
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8701820762165547e-05,
      "loss": 0.4887,
      "step": 720
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.869762146216402e-05,
      "loss": 0.5196,
      "step": 721
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.869341585428993e-05,
      "loss": 0.4905,
      "step": 722
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.868920394159335e-05,
      "loss": 0.4999,
      "step": 723
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8684985727128936e-05,
      "loss": 0.4999,
      "step": 724
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8680761213955904e-05,
      "loss": 0.5057,
      "step": 725
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8676530405138056e-05,
      "loss": 0.4612,
      "step": 726
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.867229330374374e-05,
      "loss": 0.4972,
      "step": 727
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8668049912845875e-05,
      "loss": 0.5159,
      "step": 728
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8663800235521945e-05,
      "loss": 0.4931,
      "step": 729
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8659544274853996e-05,
      "loss": 0.5126,
      "step": 730
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8655282033928618e-05,
      "loss": 0.5132,
      "step": 731
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8651013515836967e-05,
      "loss": 0.5196,
      "step": 732
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8646738723674744e-05,
      "loss": 0.4629,
      "step": 733
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.864245766054221e-05,
      "loss": 0.5075,
      "step": 734
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8638170329544164e-05,
      "loss": 0.5159,
      "step": 735
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8633876733789957e-05,
      "loss": 0.4987,
      "step": 736
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8629576876393477e-05,
      "loss": 0.4839,
      "step": 737
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8625270760473164e-05,
      "loss": 0.5186,
      "step": 738
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.862095838915198e-05,
      "loss": 0.5256,
      "step": 739
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.861663976555744e-05,
      "loss": 0.5016,
      "step": 740
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8612314892821593e-05,
      "loss": 0.5172,
      "step": 741
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8607983774081007e-05,
      "loss": 0.4951,
      "step": 742
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8603646412476787e-05,
      "loss": 0.4952,
      "step": 743
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8599302811154572e-05,
      "loss": 0.508,
      "step": 744
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8594952973264512e-05,
      "loss": 0.4862,
      "step": 745
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8590596901961296e-05,
      "loss": 0.5275,
      "step": 746
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8586234600404117e-05,
      "loss": 0.5087,
      "step": 747
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8581866071756707e-05,
      "loss": 0.5008,
      "step": 748
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8577491319187295e-05,
      "loss": 0.5007,
      "step": 749
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.857311034586863e-05,
      "loss": 0.5345,
      "step": 750
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.856872315497798e-05,
      "loss": 0.5347,
      "step": 751
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.856432974969711e-05,
      "loss": 0.5286,
      "step": 752
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8559930133212303e-05,
      "loss": 0.4872,
      "step": 753
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8555524308714335e-05,
      "loss": 0.4994,
      "step": 754
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8551112279398502e-05,
      "loss": 0.5257,
      "step": 755
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.854669404846458e-05,
      "loss": 0.4869,
      "step": 756
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.854226961911685e-05,
      "loss": 0.5137,
      "step": 757
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8537838994564102e-05,
      "loss": 0.4657,
      "step": 758
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8533402178019596e-05,
      "loss": 0.4682,
      "step": 759
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8528959172701096e-05,
      "loss": 0.5329,
      "step": 760
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8524509981830855e-05,
      "loss": 0.544,
      "step": 761
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8520054608635605e-05,
      "loss": 0.5125,
      "step": 762
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.851559305634657e-05,
      "loss": 0.4868,
      "step": 763
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8511125328199448e-05,
      "loss": 0.4941,
      "step": 764
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8506651427434417e-05,
      "loss": 0.5184,
      "step": 765
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8502171357296144e-05,
      "loss": 0.4824,
      "step": 766
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8497685121033747e-05,
      "loss": 0.5286,
      "step": 767
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.849319272190084e-05,
      "loss": 0.5081,
      "step": 768
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8488694163155485e-05,
      "loss": 0.4752,
      "step": 769
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.848418944806023e-05,
      "loss": 0.537,
      "step": 770
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8479678579882074e-05,
      "loss": 0.5071,
      "step": 771
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.847516156189249e-05,
      "loss": 0.4784,
      "step": 772
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8470638397367397e-05,
      "loss": 0.4985,
      "step": 773
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.846610908958719e-05,
      "loss": 0.4845,
      "step": 774
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8461573641836697e-05,
      "loss": 0.4908,
      "step": 775
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8457032057405222e-05,
      "loss": 0.4812,
      "step": 776
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8452484339586503e-05,
      "loss": 0.4899,
      "step": 777
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8447930491678735e-05,
      "loss": 0.4763,
      "step": 778
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8443370516984552e-05,
      "loss": 0.4857,
      "step": 779
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8438804418811038e-05,
      "loss": 0.4848,
      "step": 780
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8434232200469715e-05,
      "loss": 0.4946,
      "step": 781
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.842965386527654e-05,
      "loss": 0.5006,
      "step": 782
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8425069416551914e-05,
      "loss": 0.4956,
      "step": 783
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8420478857620667e-05,
      "loss": 0.4923,
      "step": 784
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.841588219181206e-05,
      "loss": 0.4963,
      "step": 785
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8411279422459783e-05,
      "loss": 0.469,
      "step": 786
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8406670552901958e-05,
      "loss": 0.4877,
      "step": 787
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.840205558648112e-05,
      "loss": 0.4698,
      "step": 788
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.839743452654424e-05,
      "loss": 0.4789,
      "step": 789
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.839280737644269e-05,
      "loss": 0.5049,
      "step": 790
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.838817413953228e-05,
      "loss": 0.5105,
      "step": 791
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8383534819173218e-05,
      "loss": 0.4945,
      "step": 792
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8378889418730132e-05,
      "loss": 0.5,
      "step": 793
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.837423794157206e-05,
      "loss": 0.4798,
      "step": 794
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8369580391072435e-05,
      "loss": 0.4832,
      "step": 795
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8364916770609113e-05,
      "loss": 0.5023,
      "step": 796
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8360247083564343e-05,
      "loss": 0.5037,
      "step": 797
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.835557133332477e-05,
      "loss": 0.5008,
      "step": 798
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8350889523281438e-05,
      "loss": 0.4964,
      "step": 799
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8346201656829795e-05,
      "loss": 0.4742,
      "step": 800
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.834150773736967e-05,
      "loss": 0.4739,
      "step": 801
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8336807768305286e-05,
      "loss": 0.4592,
      "step": 802
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.833210175304526e-05,
      "loss": 0.5025,
      "step": 803
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.832738969500258e-05,
      "loss": 0.4932,
      "step": 804
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8322671597594627e-05,
      "loss": 0.4932,
      "step": 805
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.831794746424316e-05,
      "loss": 0.4967,
      "step": 806
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8313217298374314e-05,
      "loss": 0.4882,
      "step": 807
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8308481103418597e-05,
      "loss": 0.539,
      "step": 808
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8303738882810897e-05,
      "loss": 0.5236,
      "step": 809
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8298990639990466e-05,
      "loss": 0.541,
      "step": 810
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8294236378400924e-05,
      "loss": 0.4753,
      "step": 811
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8289476101490257e-05,
      "loss": 0.4907,
      "step": 812
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8284709812710812e-05,
      "loss": 0.4637,
      "step": 813
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.82799375155193e-05,
      "loss": 0.5003,
      "step": 814
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8275159213376783e-05,
      "loss": 0.4891,
      "step": 815
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8270374909748686e-05,
      "loss": 0.5064,
      "step": 816
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8265584608104787e-05,
      "loss": 0.5097,
      "step": 817
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.82607883119192e-05,
      "loss": 0.4745,
      "step": 818
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.82559860246704e-05,
      "loss": 0.4628,
      "step": 819
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8251177749841205e-05,
      "loss": 0.5228,
      "step": 820
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8246363490918772e-05,
      "loss": 0.5281,
      "step": 821
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.82415432513946e-05,
      "loss": 0.4932,
      "step": 822
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8236717034764523e-05,
      "loss": 0.529,
      "step": 823
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8231884844528715e-05,
      "loss": 0.5039,
      "step": 824
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8227046684191672e-05,
      "loss": 0.5103,
      "step": 825
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8222202557262236e-05,
      "loss": 0.5121,
      "step": 826
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8217352467253563e-05,
      "loss": 0.4802,
      "step": 827
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8212496417683135e-05,
      "loss": 0.5016,
      "step": 828
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8207634412072765e-05,
      "loss": 0.5075,
      "step": 829
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8202766453948573e-05,
      "loss": 0.4873,
      "step": 830
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8197892546841003e-05,
      "loss": 0.4639,
      "step": 831
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.819301269428482e-05,
      "loss": 0.4996,
      "step": 832
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8188126899819082e-05,
      "loss": 0.4819,
      "step": 833
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8183235166987177e-05,
      "loss": 0.4975,
      "step": 834
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8178337499336788e-05,
      "loss": 0.4993,
      "step": 835
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.81734339004199e-05,
      "loss": 0.4903,
      "step": 836
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8168524373792814e-05,
      "loss": 0.4681,
      "step": 837
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8163608923016114e-05,
      "loss": 0.4853,
      "step": 838
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8158687551654687e-05,
      "loss": 0.5014,
      "step": 839
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8153760263277718e-05,
      "loss": 0.4967,
      "step": 840
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8148827061458676e-05,
      "loss": 0.4748,
      "step": 841
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8143887949775322e-05,
      "loss": 0.4947,
      "step": 842
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8138942931809702e-05,
      "loss": 0.464,
      "step": 843
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8133992011148147e-05,
      "loss": 0.4579,
      "step": 844
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8129035191381268e-05,
      "loss": 0.4921,
      "step": 845
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8124072476103957e-05,
      "loss": 0.497,
      "step": 846
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8119103868915375e-05,
      "loss": 0.4975,
      "step": 847
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.811412937341896e-05,
      "loss": 0.4995,
      "step": 848
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8109148993222425e-05,
      "loss": 0.5381,
      "step": 849
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8104162731937746e-05,
      "loss": 0.4743,
      "step": 850
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.809917059318116e-05,
      "loss": 0.4858,
      "step": 851
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8094172580573174e-05,
      "loss": 0.5014,
      "step": 852
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8089168697738554e-05,
      "loss": 0.5081,
      "step": 853
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8084158948306318e-05,
      "loss": 0.5277,
      "step": 854
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8079143335909742e-05,
      "loss": 0.4844,
      "step": 855
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8074121864186355e-05,
      "loss": 0.4853,
      "step": 856
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8069094536777938e-05,
      "loss": 0.4782,
      "step": 857
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8064061357330517e-05,
      "loss": 0.5019,
      "step": 858
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.805902232949435e-05,
      "loss": 0.4922,
      "step": 859
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8053977456923955e-05,
      "loss": 0.4872,
      "step": 860
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8048926743278083e-05,
      "loss": 0.4625,
      "step": 861
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8043870192219714e-05,
      "loss": 0.5049,
      "step": 862
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.803880780741607e-05,
      "loss": 0.4851,
      "step": 863
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8033739592538598e-05,
      "loss": 0.4808,
      "step": 864
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8028665551262974e-05,
      "loss": 0.4864,
      "step": 865
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8023585687269107e-05,
      "loss": 0.503,
      "step": 866
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8018500004241118e-05,
      "loss": 0.4995,
      "step": 867
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.801340850586736e-05,
      "loss": 0.4826,
      "step": 868
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.800831119584039e-05,
      "loss": 0.4736,
      "step": 869
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.8003208077856988e-05,
      "loss": 0.5026,
      "step": 870
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7998099155618147e-05,
      "loss": 0.4629,
      "step": 871
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7992984432829067e-05,
      "loss": 0.5012,
      "step": 872
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7987863913199153e-05,
      "loss": 0.5109,
      "step": 873
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.798273760044202e-05,
      "loss": 0.4836,
      "step": 874
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7977605498275475e-05,
      "loss": 0.4702,
      "step": 875
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7972467610421538e-05,
      "loss": 0.4973,
      "step": 876
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7967323940606408e-05,
      "loss": 0.4824,
      "step": 877
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7962174492560492e-05,
      "loss": 0.4697,
      "step": 878
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.795701927001838e-05,
      "loss": 0.5142,
      "step": 879
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7951858276718845e-05,
      "loss": 0.5064,
      "step": 880
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.794669151640486e-05,
      "loss": 0.467,
      "step": 881
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7941518992823567e-05,
      "loss": 0.4681,
      "step": 882
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7936340709726293e-05,
      "loss": 0.5232,
      "step": 883
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7931156670868544e-05,
      "loss": 0.4568,
      "step": 884
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7925966880009998e-05,
      "loss": 0.4834,
      "step": 885
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7920771340914502e-05,
      "loss": 0.4787,
      "step": 886
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7915570057350076e-05,
      "loss": 0.47,
      "step": 887
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7910363033088908e-05,
      "loss": 0.4892,
      "step": 888
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.790515027190734e-05,
      "loss": 0.4971,
      "step": 889
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.789993177758588e-05,
      "loss": 0.5058,
      "step": 890
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7894707553909202e-05,
      "loss": 0.5065,
      "step": 891
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7889477604666124e-05,
      "loss": 0.4643,
      "step": 892
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7884241933649614e-05,
      "loss": 0.4787,
      "step": 893
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7879000544656808e-05,
      "loss": 0.495,
      "step": 894
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7873753441488967e-05,
      "loss": 0.4832,
      "step": 895
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7868500627951514e-05,
      "loss": 0.5079,
      "step": 896
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7863242107853996e-05,
      "loss": 0.5006,
      "step": 897
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7857977885010108e-05,
      "loss": 0.4863,
      "step": 898
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.785270796323769e-05,
      "loss": 0.4724,
      "step": 899
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7847432346358703e-05,
      "loss": 0.4427,
      "step": 900
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7842151038199237e-05,
      "loss": 0.4982,
      "step": 901
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7836864042589518e-05,
      "loss": 0.4942,
      "step": 902
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7831571363363888e-05,
      "loss": 0.4814,
      "step": 903
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7826273004360816e-05,
      "loss": 0.5133,
      "step": 904
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7820968969422896e-05,
      "loss": 0.4737,
      "step": 905
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7815659262396825e-05,
      "loss": 0.5154,
      "step": 906
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7810343887133427e-05,
      "loss": 0.4707,
      "step": 907
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7805022847487624e-05,
      "loss": 0.4646,
      "step": 908
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7799696147318454e-05,
      "loss": 0.497,
      "step": 909
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7794363790489065e-05,
      "loss": 0.4865,
      "step": 910
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7789025780866694e-05,
      "loss": 0.4894,
      "step": 911
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7783682122322692e-05,
      "loss": 0.4932,
      "step": 912
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7778332818732492e-05,
      "loss": 0.5087,
      "step": 913
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7772977873975633e-05,
      "loss": 0.4698,
      "step": 914
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7767617291935742e-05,
      "loss": 0.5055,
      "step": 915
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7762251076500533e-05,
      "loss": 0.4891,
      "step": 916
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.77568792315618e-05,
      "loss": 0.4988,
      "step": 917
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7751501761015437e-05,
      "loss": 0.4616,
      "step": 918
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7746118668761398e-05,
      "loss": 0.4727,
      "step": 919
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7740729958703725e-05,
      "loss": 0.5219,
      "step": 920
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.773533563475053e-05,
      "loss": 0.478,
      "step": 921
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7729935700814007e-05,
      "loss": 0.4539,
      "step": 922
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.77245301608104e-05,
      "loss": 0.4783,
      "step": 923
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.771911901866003e-05,
      "loss": 0.5082,
      "step": 924
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.771370227828729e-05,
      "loss": 0.4752,
      "step": 925
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.770827994362061e-05,
      "loss": 0.4801,
      "step": 926
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7702852018592493e-05,
      "loss": 0.4637,
      "step": 927
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7697418507139497e-05,
      "loss": 0.4715,
      "step": 928
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7691979413202227e-05,
      "loss": 0.4835,
      "step": 929
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7686534740725336e-05,
      "loss": 0.503,
      "step": 930
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7681084493657526e-05,
      "loss": 0.4621,
      "step": 931
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7675628675951534e-05,
      "loss": 0.4595,
      "step": 932
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7670167291564152e-05,
      "loss": 0.4716,
      "step": 933
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7664700344456198e-05,
      "loss": 0.5086,
      "step": 934
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7659227838592524e-05,
      "loss": 0.5091,
      "step": 935
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.765374977794202e-05,
      "loss": 0.5072,
      "step": 936
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7648266166477596e-05,
      "loss": 0.5038,
      "step": 937
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7642777008176196e-05,
      "loss": 0.52,
      "step": 938
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7637282307018783e-05,
      "loss": 0.4791,
      "step": 939
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7631782066990335e-05,
      "loss": 0.4828,
      "step": 940
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.762627629207986e-05,
      "loss": 0.4928,
      "step": 941
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7620764986280367e-05,
      "loss": 0.5108,
      "step": 942
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.761524815358888e-05,
      "loss": 0.5001,
      "step": 943
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7609725798006443e-05,
      "loss": 0.4786,
      "step": 944
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7604197923538083e-05,
      "loss": 0.4003,
      "step": 945
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7598664534192847e-05,
      "loss": 0.4807,
      "step": 946
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.759312563398378e-05,
      "loss": 0.486,
      "step": 947
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.758758122692791e-05,
      "loss": 0.494,
      "step": 948
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.758203131704628e-05,
      "loss": 0.4637,
      "step": 949
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7576475908363904e-05,
      "loss": 0.4677,
      "step": 950
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7570915004909802e-05,
      "loss": 0.4647,
      "step": 951
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7565348610716963e-05,
      "loss": 0.4755,
      "step": 952
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7559776729822363e-05,
      "loss": 0.4688,
      "step": 953
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7554199366266963e-05,
      "loss": 0.4646,
      "step": 954
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7548616524095697e-05,
      "loss": 0.4838,
      "step": 955
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.754302820735747e-05,
      "loss": 0.4735,
      "step": 956
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.753743442010516e-05,
      "loss": 0.5099,
      "step": 957
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.753183516639561e-05,
      "loss": 0.4887,
      "step": 958
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7526230450289627e-05,
      "loss": 0.4805,
      "step": 959
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7520620275851988e-05,
      "loss": 0.4768,
      "step": 960
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7515004647151415e-05,
      "loss": 0.4606,
      "step": 961
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7509383568260597e-05,
      "loss": 0.4645,
      "step": 962
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7503757043256173e-05,
      "loss": 0.4819,
      "step": 963
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.749812507621872e-05,
      "loss": 0.4754,
      "step": 964
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7492487671232784e-05,
      "loss": 0.4836,
      "step": 965
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.748684483238684e-05,
      "loss": 0.4511,
      "step": 966
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7481196563773305e-05,
      "loss": 0.4954,
      "step": 967
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7475542869488533e-05,
      "loss": 0.4653,
      "step": 968
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7469883753632817e-05,
      "loss": 0.4461,
      "step": 969
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7464219220310383e-05,
      "loss": 0.5112,
      "step": 970
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7458549273629382e-05,
      "loss": 0.4905,
      "step": 971
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.745287391770189e-05,
      "loss": 0.4639,
      "step": 972
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7447193156643916e-05,
      "loss": 0.4954,
      "step": 973
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7441506994575372e-05,
      "loss": 0.4759,
      "step": 974
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.74358154356201e-05,
      "loss": 0.4572,
      "step": 975
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.743011848390585e-05,
      "loss": 0.4841,
      "step": 976
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7424416143564294e-05,
      "loss": 0.4552,
      "step": 977
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7418708418730993e-05,
      "loss": 0.4926,
      "step": 978
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7412995313545434e-05,
      "loss": 0.4719,
      "step": 979
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7407276832150984e-05,
      "loss": 0.4461,
      "step": 980
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7401552978694934e-05,
      "loss": 0.4788,
      "step": 981
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7395823757328445e-05,
      "loss": 0.4646,
      "step": 982
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7390089172206594e-05,
      "loss": 0.4313,
      "step": 983
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.738434922748833e-05,
      "loss": 0.4832,
      "step": 984
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.73786039273365e-05,
      "loss": 0.4736,
      "step": 985
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7372853275917837e-05,
      "loss": 0.5096,
      "step": 986
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7367097277402938e-05,
      "loss": 0.4771,
      "step": 987
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7361335935966303e-05,
      "loss": 0.4986,
      "step": 988
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7355569255786283e-05,
      "loss": 0.4739,
      "step": 989
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7349797241045115e-05,
      "loss": 0.5086,
      "step": 990
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7344019895928904e-05,
      "loss": 0.4755,
      "step": 991
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7338237224627616e-05,
      "loss": 0.4478,
      "step": 992
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7332449231335084e-05,
      "loss": 0.4432,
      "step": 993
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7326655920248993e-05,
      "loss": 0.4791,
      "step": 994
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7320857295570895e-05,
      "loss": 0.4685,
      "step": 995
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.731505336150619e-05,
      "loss": 0.4711,
      "step": 996
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.730924412226413e-05,
      "loss": 0.4779,
      "step": 997
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7303429582057813e-05,
      "loss": 0.4759,
      "step": 998
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7297609745104184e-05,
      "loss": 0.4463,
      "step": 999
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7291784615624023e-05,
      "loss": 0.4976,
      "step": 1000
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7285954197841964e-05,
      "loss": 0.479,
      "step": 1001
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7280118495986454e-05,
      "loss": 0.4678,
      "step": 1002
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.727427751428979e-05,
      "loss": 0.5062,
      "step": 1003
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.726843125698809e-05,
      "loss": 0.4862,
      "step": 1004
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7262579728321302e-05,
      "loss": 0.4762,
      "step": 1005
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7256722932533195e-05,
      "loss": 0.478,
      "step": 1006
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7250860873871357e-05,
      "loss": 0.5071,
      "step": 1007
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7244993556587192e-05,
      "loss": 0.495,
      "step": 1008
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.723912098493592e-05,
      "loss": 0.5375,
      "step": 1009
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7233243163176575e-05,
      "loss": 0.4567,
      "step": 1010
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7227360095571992e-05,
      "loss": 0.4792,
      "step": 1011
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7221471786388815e-05,
      "loss": 0.4765,
      "step": 1012
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.721557823989749e-05,
      "loss": 0.4678,
      "step": 1013
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.720967946037225e-05,
      "loss": 0.4629,
      "step": 1014
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.720377545209114e-05,
      "loss": 0.4557,
      "step": 1015
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.719786621933599e-05,
      "loss": 0.4363,
      "step": 1016
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7191951766392416e-05,
      "loss": 0.4581,
      "step": 1017
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7186032097549822e-05,
      "loss": 0.4798,
      "step": 1018
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.71801072171014e-05,
      "loss": 0.4619,
      "step": 1019
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7174177129344106e-05,
      "loss": 0.4785,
      "step": 1020
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7168241838578693e-05,
      "loss": 0.4911,
      "step": 1021
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7162301349109677e-05,
      "loss": 0.4532,
      "step": 1022
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7156355665245342e-05,
      "loss": 0.4859,
      "step": 1023
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7150404791297744e-05,
      "loss": 0.4733,
      "step": 1024
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7144448731582698e-05,
      "loss": 0.4713,
      "step": 1025
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7138487490419787e-05,
      "loss": 0.458,
      "step": 1026
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.713252107213235e-05,
      "loss": 0.4941,
      "step": 1027
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7126549481047473e-05,
      "loss": 0.486,
      "step": 1028
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7120572721496004e-05,
      "loss": 0.4668,
      "step": 1029
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7114590797812533e-05,
      "loss": 0.4716,
      "step": 1030
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7108603714335396e-05,
      "loss": 0.4762,
      "step": 1031
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7102611475406676e-05,
      "loss": 0.4606,
      "step": 1032
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7096614085372186e-05,
      "loss": 0.4801,
      "step": 1033
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.709061154858148e-05,
      "loss": 0.4678,
      "step": 1034
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7084603869387845e-05,
      "loss": 0.4836,
      "step": 1035
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.70785910521483e-05,
      "loss": 0.4634,
      "step": 1036
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7072573101223584e-05,
      "loss": 0.4884,
      "step": 1037
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7066550020978164e-05,
      "loss": 0.4654,
      "step": 1038
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7060521815780225e-05,
      "loss": 0.4671,
      "step": 1039
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.705448849000167e-05,
      "loss": 0.5056,
      "step": 1040
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7048450048018113e-05,
      "loss": 0.4801,
      "step": 1041
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.704240649420888e-05,
      "loss": 0.4916,
      "step": 1042
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7036357832957005e-05,
      "loss": 0.4723,
      "step": 1043
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7030304068649228e-05,
      "loss": 0.4862,
      "step": 1044
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7024245205675986e-05,
      "loss": 0.4551,
      "step": 1045
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7018181248431416e-05,
      "loss": 0.4672,
      "step": 1046
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.701211220131335e-05,
      "loss": 0.466,
      "step": 1047
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.7006038068723307e-05,
      "loss": 0.4501,
      "step": 1048
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6999958855066497e-05,
      "loss": 0.4667,
      "step": 1049
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6993874564751824e-05,
      "loss": 0.4962,
      "step": 1050
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6987785202191856e-05,
      "loss": 0.4711,
      "step": 1051
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6981690771802852e-05,
      "loss": 0.4486,
      "step": 1052
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6975591278004747e-05,
      "loss": 0.4897,
      "step": 1053
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.696948672522114e-05,
      "loss": 0.4665,
      "step": 1054
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6963377117879307e-05,
      "loss": 0.4967,
      "step": 1055
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6957262460410183e-05,
      "loss": 0.4407,
      "step": 1056
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.695114275724837e-05,
      "loss": 0.4844,
      "step": 1057
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.694501801283213e-05,
      "loss": 0.4573,
      "step": 1058
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6938888231603377e-05,
      "loss": 0.4795,
      "step": 1059
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6932753418007683e-05,
      "loss": 0.4726,
      "step": 1060
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6926613576494263e-05,
      "loss": 0.4632,
      "step": 1061
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.692046871151599e-05,
      "loss": 0.4764,
      "step": 1062
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.691431882752936e-05,
      "loss": 0.4571,
      "step": 1063
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6908163928994536e-05,
      "loss": 0.4661,
      "step": 1064
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6902004020375292e-05,
      "loss": 0.4659,
      "step": 1065
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6895839106139054e-05,
      "loss": 0.4785,
      "step": 1066
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.688966919075687e-05,
      "loss": 0.497,
      "step": 1067
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.688349427870341e-05,
      "loss": 0.4816,
      "step": 1068
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6877314374456983e-05,
      "loss": 0.4643,
      "step": 1069
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6871129482499502e-05,
      "loss": 0.4813,
      "step": 1070
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6864939607316513e-05,
      "loss": 0.503,
      "step": 1071
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6858744753397157e-05,
      "loss": 0.4954,
      "step": 1072
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.685254492523421e-05,
      "loss": 0.4814,
      "step": 1073
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.684634012732403e-05,
      "loss": 0.4767,
      "step": 1074
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6840130364166597e-05,
      "loss": 0.4819,
      "step": 1075
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6833915640265485e-05,
      "loss": 0.4344,
      "step": 1076
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6827695960127873e-05,
      "loss": 0.4502,
      "step": 1077
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6821471328264524e-05,
      "loss": 0.4665,
      "step": 1078
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.681524174918979e-05,
      "loss": 0.4843,
      "step": 1079
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.680900722742163e-05,
      "loss": 0.4662,
      "step": 1080
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.680276776748157e-05,
      "loss": 0.4466,
      "step": 1081
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.679652337389472e-05,
      "loss": 0.4541,
      "step": 1082
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6790274051189775e-05,
      "loss": 0.4813,
      "step": 1083
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6784019803899e-05,
      "loss": 0.4634,
      "step": 1084
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.677776063655823e-05,
      "loss": 0.4812,
      "step": 1085
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6771496553706874e-05,
      "loss": 0.4929,
      "step": 1086
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6765227559887898e-05,
      "loss": 0.4695,
      "step": 1087
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6758953659647838e-05,
      "loss": 0.4558,
      "step": 1088
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.675267485753678e-05,
      "loss": 0.4919,
      "step": 1089
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6746391158108374e-05,
      "loss": 0.4774,
      "step": 1090
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6740102565919817e-05,
      "loss": 0.4772,
      "step": 1091
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6733809085531852e-05,
      "loss": 0.4784,
      "step": 1092
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6727510721508766e-05,
      "loss": 0.4864,
      "step": 1093
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.67212074784184e-05,
      "loss": 0.4556,
      "step": 1094
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6714899360832118e-05,
      "loss": 0.4652,
      "step": 1095
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6708586373324828e-05,
      "loss": 0.4563,
      "step": 1096
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.670226852047497e-05,
      "loss": 0.4578,
      "step": 1097
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.669594580686451e-05,
      "loss": 0.4621,
      "step": 1098
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.668961823707894e-05,
      "loss": 0.4649,
      "step": 1099
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6683285815707276e-05,
      "loss": 0.4625,
      "step": 1100
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.667694854734204e-05,
      "loss": 0.4559,
      "step": 1101
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.667060643657929e-05,
      "loss": 0.4728,
      "step": 1102
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6664259488018585e-05,
      "loss": 0.463,
      "step": 1103
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6657907706262984e-05,
      "loss": 0.462,
      "step": 1104
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.665155109591907e-05,
      "loss": 0.4869,
      "step": 1105
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.664518966159691e-05,
      "loss": 0.4764,
      "step": 1106
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6638823407910085e-05,
      "loss": 0.452,
      "step": 1107
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6632452339475663e-05,
      "loss": 0.4744,
      "step": 1108
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.66260764609142e-05,
      "loss": 0.5035,
      "step": 1109
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6619695776849748e-05,
      "loss": 0.4832,
      "step": 1110
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6613310291909837e-05,
      "loss": 0.4437,
      "step": 1111
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.660692001072549e-05,
      "loss": 0.4874,
      "step": 1112
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6600524937931197e-05,
      "loss": 0.4578,
      "step": 1113
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6594125078164935e-05,
      "loss": 0.4644,
      "step": 1114
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.658772043606813e-05,
      "loss": 0.4426,
      "step": 1115
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.658131101628571e-05,
      "loss": 0.4906,
      "step": 1116
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.657489682346604e-05,
      "loss": 0.5024,
      "step": 1117
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6568477862260952e-05,
      "loss": 0.4716,
      "step": 1118
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6562054137325748e-05,
      "loss": 0.4511,
      "step": 1119
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6555625653319175e-05,
      "loss": 0.4705,
      "step": 1120
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6549192414903435e-05,
      "loss": 0.4808,
      "step": 1121
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6542754426744175e-05,
      "loss": 0.4688,
      "step": 1122
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.653631169351049e-05,
      "loss": 0.4371,
      "step": 1123
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.652986421987491e-05,
      "loss": 0.4679,
      "step": 1124
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.652341201051342e-05,
      "loss": 0.4538,
      "step": 1125
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6516955070105415e-05,
      "loss": 0.4709,
      "step": 1126
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.651049340333374e-05,
      "loss": 0.4797,
      "step": 1127
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6504027014884656e-05,
      "loss": 0.4334,
      "step": 1128
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.649755590944786e-05,
      "loss": 0.4661,
      "step": 1129
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6491080091716457e-05,
      "loss": 0.4578,
      "step": 1130
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6484599566386986e-05,
      "loss": 0.452,
      "step": 1131
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.647811433815938e-05,
      "loss": 0.4298,
      "step": 1132
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6471624411737002e-05,
      "loss": 0.4239,
      "step": 1133
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6465129791826605e-05,
      "loss": 0.4587,
      "step": 1134
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6458630483138354e-05,
      "loss": 0.4638,
      "step": 1135
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6452126490385827e-05,
      "loss": 0.4764,
      "step": 1136
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6445617818285974e-05,
      "loss": 0.4799,
      "step": 1137
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6439104471559157e-05,
      "loss": 0.4545,
      "step": 1138
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6432586454929125e-05,
      "loss": 0.4712,
      "step": 1139
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6426063773123004e-05,
      "loss": 0.4741,
      "step": 1140
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6419536430871318e-05,
      "loss": 0.4398,
      "step": 1141
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.641300443290796e-05,
      "loss": 0.4842,
      "step": 1142
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6406467783970205e-05,
      "loss": 0.4699,
      "step": 1143
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6399926488798702e-05,
      "loss": 0.4614,
      "step": 1144
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.639338055213746e-05,
      "loss": 0.4491,
      "step": 1145
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6386829978733868e-05,
      "loss": 0.4677,
      "step": 1146
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.638027477333867e-05,
      "loss": 0.494,
      "step": 1147
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.637371494070597e-05,
      "loss": 0.45,
      "step": 1148
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6367150485593223e-05,
      "loss": 0.4536,
      "step": 1149
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.636058141276125e-05,
      "loss": 0.4774,
      "step": 1150
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6354007726974205e-05,
      "loss": 0.4912,
      "step": 1151
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6347429432999602e-05,
      "loss": 0.4298,
      "step": 1152
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6340846535608284e-05,
      "loss": 0.4658,
      "step": 1153
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6334259039574445e-05,
      "loss": 0.4565,
      "step": 1154
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6327666949675603e-05,
      "loss": 0.4387,
      "step": 1155
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6321070270692614e-05,
      "loss": 0.4463,
      "step": 1156
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6314469007409657e-05,
      "loss": 0.4491,
      "step": 1157
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.630786316461425e-05,
      "loss": 0.5062,
      "step": 1158
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6301252747097213e-05,
      "loss": 0.4831,
      "step": 1159
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6294637759652692e-05,
      "loss": 0.4528,
      "step": 1160
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6288018207078153e-05,
      "loss": 0.4379,
      "step": 1161
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6281394094174363e-05,
      "loss": 0.4476,
      "step": 1162
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6274765425745403e-05,
      "loss": 0.4571,
      "step": 1163
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6268132206598656e-05,
      "loss": 0.4928,
      "step": 1164
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6261494441544805e-05,
      "loss": 0.415,
      "step": 1165
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6254852135397828e-05,
      "loss": 0.4936,
      "step": 1166
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6248205292975003e-05,
      "loss": 0.4656,
      "step": 1167
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.624155391909689e-05,
      "loss": 0.4642,
      "step": 1168
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6234898018587336e-05,
      "loss": 0.4631,
      "step": 1169
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.622823759627348e-05,
      "loss": 0.4743,
      "step": 1170
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6221572656985728e-05,
      "loss": 0.4496,
      "step": 1171
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6214903205557774e-05,
      "loss": 0.4594,
      "step": 1172
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6208229246826574e-05,
      "loss": 0.4507,
      "step": 1173
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.620155078563236e-05,
      "loss": 0.4728,
      "step": 1174
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.619486782681862e-05,
      "loss": 0.4452,
      "step": 1175
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.618818037523212e-05,
      "loss": 0.462,
      "step": 1176
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6181488435722867e-05,
      "loss": 0.4556,
      "step": 1177
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6174792013144133e-05,
      "loss": 0.4423,
      "step": 1178
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6168091112352443e-05,
      "loss": 0.4592,
      "step": 1179
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.616138573820756e-05,
      "loss": 0.469,
      "step": 1180
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.61546758955725e-05,
      "loss": 0.499,
      "step": 1181
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6147961589313515e-05,
      "loss": 0.4711,
      "step": 1182
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6141242824300096e-05,
      "loss": 0.4576,
      "step": 1183
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6134519605404966e-05,
      "loss": 0.4796,
      "step": 1184
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6127791937504086e-05,
      "loss": 0.4219,
      "step": 1185
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.612105982547663e-05,
      "loss": 0.4453,
      "step": 1186
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6114323274205008e-05,
      "loss": 0.4837,
      "step": 1187
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.610758228857484e-05,
      "loss": 0.4416,
      "step": 1188
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6100836873474966e-05,
      "loss": 0.4523,
      "step": 1189
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6094087033797438e-05,
      "loss": 0.4549,
      "step": 1190
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.608733277443752e-05,
      "loss": 0.4621,
      "step": 1191
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.608057410029368e-05,
      "loss": 0.4917,
      "step": 1192
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.607381101626758e-05,
      "loss": 0.4702,
      "step": 1193
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6067043527264093e-05,
      "loss": 0.4845,
      "step": 1194
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6060271638191275e-05,
      "loss": 0.4553,
      "step": 1195
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.605349535396038e-05,
      "loss": 0.4197,
      "step": 1196
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6046714679485848e-05,
      "loss": 0.4655,
      "step": 1197
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.60399296196853e-05,
      "loss": 0.4852,
      "step": 1198
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.6033140179479546e-05,
      "loss": 0.4385,
      "step": 1199
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.6026346363792565e-05,
      "loss": 0.4296,
      "step": 1200
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.601954817755151e-05,
      "loss": 0.4709,
      "step": 1201
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.6012745625686704e-05,
      "loss": 0.4638,
      "step": 1202
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.6005938713131644e-05,
      "loss": 0.4255,
      "step": 1203
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5999127444822972e-05,
      "loss": 0.458,
      "step": 1204
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.599231182570051e-05,
      "loss": 0.4903,
      "step": 1205
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5985491860707217e-05,
      "loss": 0.471,
      "step": 1206
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5978667554789216e-05,
      "loss": 0.4641,
      "step": 1207
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5971838912895775e-05,
      "loss": 0.4313,
      "step": 1208
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5965005939979303e-05,
      "loss": 0.4592,
      "step": 1209
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5958168640995356e-05,
      "loss": 0.4609,
      "step": 1210
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.595132702090262e-05,
      "loss": 0.4816,
      "step": 1211
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5944481084662923e-05,
      "loss": 0.4388,
      "step": 1212
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.593763083724122e-05,
      "loss": 0.4343,
      "step": 1213
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5930776283605585e-05,
      "loss": 0.4719,
      "step": 1214
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5923917428727225e-05,
      "loss": 0.4195,
      "step": 1215
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5917054277580474e-05,
      "loss": 0.4515,
      "step": 1216
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5910186835142757e-05,
      "loss": 0.4277,
      "step": 1217
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.590331510639463e-05,
      "loss": 0.4166,
      "step": 1218
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5896439096319758e-05,
      "loss": 0.4281,
      "step": 1219
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.58895588099049e-05,
      "loss": 0.4706,
      "step": 1220
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5882674252139928e-05,
      "loss": 0.4274,
      "step": 1221
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5875785428017797e-05,
      "loss": 0.4424,
      "step": 1222
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5868892342534575e-05,
      "loss": 0.4702,
      "step": 1223
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.586199500068941e-05,
      "loss": 0.4844,
      "step": 1224
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5855093407484533e-05,
      "loss": 0.463,
      "step": 1225
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5848187567925264e-05,
      "loss": 0.4322,
      "step": 1226
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5841277487020005e-05,
      "loss": 0.4373,
      "step": 1227
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5834363169780227e-05,
      "loss": 0.4429,
      "step": 1228
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.582744462122048e-05,
      "loss": 0.4798,
      "step": 1229
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5820521846358376e-05,
      "loss": 0.4581,
      "step": 1230
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.58135948502146e-05,
      "loss": 0.4436,
      "step": 1231
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.580666363781289e-05,
      "loss": 0.448,
      "step": 1232
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5799728214180054e-05,
      "loss": 0.4469,
      "step": 1233
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5792788584345936e-05,
      "loss": 0.4468,
      "step": 1234
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.578584475334345e-05,
      "loss": 0.4726,
      "step": 1235
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.577889672620854e-05,
      "loss": 0.4473,
      "step": 1236
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5771944507980208e-05,
      "loss": 0.4615,
      "step": 1237
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.576498810370048e-05,
      "loss": 0.4262,
      "step": 1238
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5758027518414434e-05,
      "loss": 0.4441,
      "step": 1239
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.575106275717017e-05,
      "loss": 0.4633,
      "step": 1240
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5744093825018817e-05,
      "loss": 0.4442,
      "step": 1241
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5737120727014535e-05,
      "loss": 0.4602,
      "step": 1242
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.57301434682145e-05,
      "loss": 0.4553,
      "step": 1243
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5723162053678902e-05,
      "loss": 0.4461,
      "step": 1244
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5716176488470953e-05,
      "loss": 0.462,
      "step": 1245
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5709186777656877e-05,
      "loss": 0.4273,
      "step": 1246
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5702192926305892e-05,
      "loss": 0.4578,
      "step": 1247
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5695194939490228e-05,
      "loss": 0.425,
      "step": 1248
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5688192822285116e-05,
      "loss": 0.45,
      "step": 1249
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.568118657976878e-05,
      "loss": 0.4398,
      "step": 1250
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.567417621702243e-05,
      "loss": 0.4391,
      "step": 1251
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.566716173913027e-05,
      "loss": 0.4874,
      "step": 1252
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.566014315117949e-05,
      "loss": 0.4348,
      "step": 1253
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5653120458260263e-05,
      "loss": 0.4141,
      "step": 1254
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5646093665465723e-05,
      "loss": 0.4746,
      "step": 1255
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5639062777892e-05,
      "loss": 0.4414,
      "step": 1256
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5632027800638177e-05,
      "loss": 0.446,
      "step": 1257
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5624988738806305e-05,
      "loss": 0.4631,
      "step": 1258
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.561794559750141e-05,
      "loss": 0.436,
      "step": 1259
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.561089838183146e-05,
      "loss": 0.4351,
      "step": 1260
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5603847096907393e-05,
      "loss": 0.4965,
      "step": 1261
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5596791747843083e-05,
      "loss": 0.4827,
      "step": 1262
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5589732339755362e-05,
      "loss": 0.4434,
      "step": 1263
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5582668877764005e-05,
      "loss": 0.4732,
      "step": 1264
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5575601366991725e-05,
      "loss": 0.4489,
      "step": 1265
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5568529812564166e-05,
      "loss": 0.4231,
      "step": 1266
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5561454219609915e-05,
      "loss": 0.4345,
      "step": 1267
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.555437459326048e-05,
      "loss": 0.4445,
      "step": 1268
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.55472909386503e-05,
      "loss": 0.3879,
      "step": 1269
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5540203260916728e-05,
      "loss": 0.3562,
      "step": 1270
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5533111565200046e-05,
      "loss": 0.33,
      "step": 1271
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.552601585664344e-05,
      "loss": 0.3358,
      "step": 1272
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5518916140393002e-05,
      "loss": 0.3425,
      "step": 1273
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5511812421597746e-05,
      "loss": 0.344,
      "step": 1274
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5504704705409582e-05,
      "loss": 0.3783,
      "step": 1275
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5497592996983316e-05,
      "loss": 0.3336,
      "step": 1276
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5490477301476648e-05,
      "loss": 0.3517,
      "step": 1277
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5483357624050176e-05,
      "loss": 0.3517,
      "step": 1278
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5476233969867388e-05,
      "loss": 0.3419,
      "step": 1279
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5469106344094636e-05,
      "loss": 0.3804,
      "step": 1280
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5461974751901184e-05,
      "loss": 0.3529,
      "step": 1281
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5454839198459145e-05,
      "loss": 0.3584,
      "step": 1282
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5447699688943516e-05,
      "loss": 0.352,
      "step": 1283
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5440556228532168e-05,
      "loss": 0.3129,
      "step": 1284
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.543340882240583e-05,
      "loss": 0.3397,
      "step": 1285
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5426257475748094e-05,
      "loss": 0.3459,
      "step": 1286
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.541910219374541e-05,
      "loss": 0.3592,
      "step": 1287
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.541194298158708e-05,
      "loss": 0.3532,
      "step": 1288
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5404779844465266e-05,
      "loss": 0.3548,
      "step": 1289
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5397612787574967e-05,
      "loss": 0.3665,
      "step": 1290
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5390441816114022e-05,
      "loss": 0.3404,
      "step": 1291
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5383266935283117e-05,
      "loss": 0.3497,
      "step": 1292
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5376088150285777e-05,
      "loss": 0.3561,
      "step": 1293
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5368905466328343e-05,
      "loss": 0.3591,
      "step": 1294
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5361718888619996e-05,
      "loss": 0.3324,
      "step": 1295
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5354528422372735e-05,
      "loss": 0.4063,
      "step": 1296
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5347334072801392e-05,
      "loss": 0.3426,
      "step": 1297
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.534013584512359e-05,
      "loss": 0.3392,
      "step": 1298
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.533293374455979e-05,
      "loss": 0.3443,
      "step": 1299
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5325727776333247e-05,
      "loss": 0.3637,
      "step": 1300
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.531851794567003e-05,
      "loss": 0.3554,
      "step": 1301
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5311304257799e-05,
      "loss": 0.3544,
      "step": 1302
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5304086717951817e-05,
      "loss": 0.344,
      "step": 1303
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5296865331362945e-05,
      "loss": 0.3548,
      "step": 1304
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5289640103269626e-05,
      "loss": 0.33,
      "step": 1305
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5282411038911892e-05,
      "loss": 0.3326,
      "step": 1306
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.527517814353256e-05,
      "loss": 0.364,
      "step": 1307
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5267941422377223e-05,
      "loss": 0.3666,
      "step": 1308
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5260700880694248e-05,
      "loss": 0.3546,
      "step": 1309
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5253456523734771e-05,
      "loss": 0.3666,
      "step": 1310
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5246208356752701e-05,
      "loss": 0.3621,
      "step": 1311
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5238956385004703e-05,
      "loss": 0.3419,
      "step": 1312
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.523170061375021e-05,
      "loss": 0.3708,
      "step": 1313
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.52244410482514e-05,
      "loss": 0.4015,
      "step": 1314
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5217177693773212e-05,
      "loss": 0.351,
      "step": 1315
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5209910555583327e-05,
      "loss": 0.3306,
      "step": 1316
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5202639638952173e-05,
      "loss": 0.3382,
      "step": 1317
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5195364949152918e-05,
      "loss": 0.3475,
      "step": 1318
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5188086491461467e-05,
      "loss": 0.3577,
      "step": 1319
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5180804271156452e-05,
      "loss": 0.36,
      "step": 1320
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5173518293519246e-05,
      "loss": 0.3201,
      "step": 1321
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5166228563833933e-05,
      "loss": 0.3681,
      "step": 1322
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5158935087387333e-05,
      "loss": 0.3622,
      "step": 1323
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.515163786946896e-05,
      "loss": 0.3611,
      "step": 1324
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5144336915371071e-05,
      "loss": 0.3568,
      "step": 1325
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5137032230388613e-05,
      "loss": 0.3621,
      "step": 1326
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.512972381981924e-05,
      "loss": 0.3686,
      "step": 1327
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5122411688963317e-05,
      "loss": 0.3394,
      "step": 1328
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5115095843123897e-05,
      "loss": 0.3462,
      "step": 1329
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.510777628760674e-05,
      "loss": 0.3423,
      "step": 1330
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5100453027720278e-05,
      "loss": 0.356,
      "step": 1331
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5093126068775646e-05,
      "loss": 0.3351,
      "step": 1332
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5085795416086655e-05,
      "loss": 0.3482,
      "step": 1333
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5078461074969797e-05,
      "loss": 0.3548,
      "step": 1334
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5071123050744234e-05,
      "loss": 0.3577,
      "step": 1335
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.506378134873181e-05,
      "loss": 0.3373,
      "step": 1336
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.505643597425702e-05,
      "loss": 0.3488,
      "step": 1337
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.504908693264704e-05,
      "loss": 0.3734,
      "step": 1338
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5041734229231688e-05,
      "loss": 0.3523,
      "step": 1339
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5034377869343453e-05,
      "loss": 0.3463,
      "step": 1340
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.502701785831747e-05,
      "loss": 0.3706,
      "step": 1341
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5019654201491515e-05,
      "loss": 0.3506,
      "step": 1342
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5012286904206021e-05,
      "loss": 0.3635,
      "step": 1343
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5004915971804055e-05,
      "loss": 0.3539,
      "step": 1344
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4997541409631317e-05,
      "loss": 0.3314,
      "step": 1345
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4990163223036143e-05,
      "loss": 0.3418,
      "step": 1346
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4982781417369496e-05,
      "loss": 0.3583,
      "step": 1347
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4975395997984968e-05,
      "loss": 0.3581,
      "step": 1348
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4968006970238766e-05,
      "loss": 0.3349,
      "step": 1349
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4960614339489721e-05,
      "loss": 0.3415,
      "step": 1350
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4953218111099266e-05,
      "loss": 0.3467,
      "step": 1351
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4945818290431458e-05,
      "loss": 0.3389,
      "step": 1352
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4938414882852943e-05,
      "loss": 0.3824,
      "step": 1353
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4931007893732981e-05,
      "loss": 0.3521,
      "step": 1354
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4923597328443423e-05,
      "loss": 0.3555,
      "step": 1355
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4916183192358717e-05,
      "loss": 0.3252,
      "step": 1356
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4908765490855899e-05,
      "loss": 0.3679,
      "step": 1357
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4901344229314592e-05,
      "loss": 0.3377,
      "step": 1358
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4893919413116998e-05,
      "loss": 0.3764,
      "step": 1359
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4886491047647895e-05,
      "loss": 0.34,
      "step": 1360
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4879059138294647e-05,
      "loss": 0.383,
      "step": 1361
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4871623690447177e-05,
      "loss": 0.3797,
      "step": 1362
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4864184709497975e-05,
      "loss": 0.3558,
      "step": 1363
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4856742200842097e-05,
      "loss": 0.3564,
      "step": 1364
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4849296169877158e-05,
      "loss": 0.3647,
      "step": 1365
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4841846622003327e-05,
      "loss": 0.3674,
      "step": 1366
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.483439356262332e-05,
      "loss": 0.3324,
      "step": 1367
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4826936997142399e-05,
      "loss": 0.3441,
      "step": 1368
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4819476930968373e-05,
      "loss": 0.3255,
      "step": 1369
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4812013369511595e-05,
      "loss": 0.3292,
      "step": 1370
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4804546318184938e-05,
      "loss": 0.3608,
      "step": 1371
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4797075782403818e-05,
      "loss": 0.3708,
      "step": 1372
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4789601767586174e-05,
      "loss": 0.3551,
      "step": 1373
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4782124279152472e-05,
      "loss": 0.3481,
      "step": 1374
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4774643322525691e-05,
      "loss": 0.3543,
      "step": 1375
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.476715890313133e-05,
      "loss": 0.3347,
      "step": 1376
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4759671026397393e-05,
      "loss": 0.3568,
      "step": 1377
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4752179697754403e-05,
      "loss": 0.3369,
      "step": 1378
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4744684922635374e-05,
      "loss": 0.3737,
      "step": 1379
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4737186706475826e-05,
      "loss": 0.3285,
      "step": 1380
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4729685054713776e-05,
      "loss": 0.3837,
      "step": 1381
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4722179972789725e-05,
      "loss": 0.3628,
      "step": 1382
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4714671466146673e-05,
      "loss": 0.3524,
      "step": 1383
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4707159540230094e-05,
      "loss": 0.3532,
      "step": 1384
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4699644200487948e-05,
      "loss": 0.3373,
      "step": 1385
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4692125452370664e-05,
      "loss": 0.31,
      "step": 1386
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4684603301331151e-05,
      "loss": 0.3586,
      "step": 1387
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.467707775282478e-05,
      "loss": 0.3666,
      "step": 1388
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.466954881230939e-05,
      "loss": 0.3713,
      "step": 1389
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4662016485245274e-05,
      "loss": 0.3769,
      "step": 1390
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4654480777095192e-05,
      "loss": 0.361,
      "step": 1391
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4646941693324346e-05,
      "loss": 0.344,
      "step": 1392
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4639399239400388e-05,
      "loss": 0.3667,
      "step": 1393
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4631853420793415e-05,
      "loss": 0.3811,
      "step": 1394
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4624304242975964e-05,
      "loss": 0.3641,
      "step": 1395
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4616751711423016e-05,
      "loss": 0.3806,
      "step": 1396
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.460919583161197e-05,
      "loss": 0.362,
      "step": 1397
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4601636609022665e-05,
      "loss": 0.3315,
      "step": 1398
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4594074049137357e-05,
      "loss": 0.3783,
      "step": 1399
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4586508157440728e-05,
      "loss": 0.3447,
      "step": 1400
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.457893893941987e-05,
      "loss": 0.3938,
      "step": 1401
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4571366400564293e-05,
      "loss": 0.3729,
      "step": 1402
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4563790546365914e-05,
      "loss": 0.3358,
      "step": 1403
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4556211382319054e-05,
      "loss": 0.3724,
      "step": 1404
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4548628913920434e-05,
      "loss": 0.3245,
      "step": 1405
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4541043146669167e-05,
      "loss": 0.3545,
      "step": 1406
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4533454086066772e-05,
      "loss": 0.3433,
      "step": 1407
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4525861737617142e-05,
      "loss": 0.3346,
      "step": 1408
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4518266106826562e-05,
      "loss": 0.3429,
      "step": 1409
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4510667199203697e-05,
      "loss": 0.3582,
      "step": 1410
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.450306502025958e-05,
      "loss": 0.3663,
      "step": 1411
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4495459575507632e-05,
      "loss": 0.3604,
      "step": 1412
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4487850870463629e-05,
      "loss": 0.3389,
      "step": 1413
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4480238910645721e-05,
      "loss": 0.3534,
      "step": 1414
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4472623701574409e-05,
      "loss": 0.3673,
      "step": 1415
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4465005248772558e-05,
      "loss": 0.3668,
      "step": 1416
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4457383557765385e-05,
      "loss": 0.3324,
      "step": 1417
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.444975863408045e-05,
      "loss": 0.3605,
      "step": 1418
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4442130483247666e-05,
      "loss": 0.3536,
      "step": 1419
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4434499110799273e-05,
      "loss": 0.3664,
      "step": 1420
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4426864522269867e-05,
      "loss": 0.3417,
      "step": 1421
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.441922672319636e-05,
      "loss": 0.3645,
      "step": 1422
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4411585719117996e-05,
      "loss": 0.3611,
      "step": 1423
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4403941515576344e-05,
      "loss": 0.3608,
      "step": 1424
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4396294118115302e-05,
      "loss": 0.3379,
      "step": 1425
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.438864353228107e-05,
      "loss": 0.3587,
      "step": 1426
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4380989763622169e-05,
      "loss": 0.3628,
      "step": 1427
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4373332817689426e-05,
      "loss": 0.3495,
      "step": 1428
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.436567270003598e-05,
      "loss": 0.3309,
      "step": 1429
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4358009416217252e-05,
      "loss": 0.3683,
      "step": 1430
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4350342971790979e-05,
      "loss": 0.313,
      "step": 1431
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4342673372317174e-05,
      "loss": 0.3739,
      "step": 1432
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4335000623358154e-05,
      "loss": 0.3623,
      "step": 1433
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4327324730478505e-05,
      "loss": 0.354,
      "step": 1434
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4319645699245106e-05,
      "loss": 0.3272,
      "step": 1435
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4311963535227102e-05,
      "loss": 0.3225,
      "step": 1436
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4304278243995915e-05,
      "loss": 0.3563,
      "step": 1437
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4296589831125234e-05,
      "loss": 0.3448,
      "step": 1438
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4288898302191012e-05,
      "loss": 0.3686,
      "step": 1439
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4281203662771466e-05,
      "loss": 0.3544,
      "step": 1440
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4273505918447054e-05,
      "loss": 0.3576,
      "step": 1441
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4265805074800509e-05,
      "loss": 0.3246,
      "step": 1442
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.425810113741679e-05,
      "loss": 0.3567,
      "step": 1443
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4250394111883117e-05,
      "loss": 0.357,
      "step": 1444
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4242684003788934e-05,
      "loss": 0.3527,
      "step": 1445
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4234970818725933e-05,
      "loss": 0.331,
      "step": 1446
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4227254562288036e-05,
      "loss": 0.3503,
      "step": 1447
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4219535240071378e-05,
      "loss": 0.359,
      "step": 1448
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4211812857674339e-05,
      "loss": 0.3178,
      "step": 1449
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4204087420697503e-05,
      "loss": 0.3454,
      "step": 1450
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4196358934743682e-05,
      "loss": 0.3665,
      "step": 1451
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.418862740541788e-05,
      "loss": 0.3752,
      "step": 1452
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4180892838327327e-05,
      "loss": 0.3366,
      "step": 1453
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4173155239081452e-05,
      "loss": 0.3423,
      "step": 1454
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4165414613291871e-05,
      "loss": 0.3455,
      "step": 1455
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4157670966572408e-05,
      "loss": 0.3376,
      "step": 1456
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4149924304539077e-05,
      "loss": 0.3453,
      "step": 1457
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4142174632810074e-05,
      "loss": 0.371,
      "step": 1458
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4134421957005775e-05,
      "loss": 0.3663,
      "step": 1459
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4126666282748746e-05,
      "loss": 0.3524,
      "step": 1460
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4118907615663715e-05,
      "loss": 0.3282,
      "step": 1461
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4111145961377592e-05,
      "loss": 0.3466,
      "step": 1462
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4103381325519445e-05,
      "loss": 0.3517,
      "step": 1463
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4095613713720503e-05,
      "loss": 0.3319,
      "step": 1464
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4087843131614165e-05,
      "loss": 0.3562,
      "step": 1465
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4080069584835971e-05,
      "loss": 0.327,
      "step": 1466
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4072293079023619e-05,
      "loss": 0.3417,
      "step": 1467
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4064513619816946e-05,
      "loss": 0.3271,
      "step": 1468
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4056731212857943e-05,
      "loss": 0.3705,
      "step": 1469
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4048945863790727e-05,
      "loss": 0.3601,
      "step": 1470
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4041157578261552e-05,
      "loss": 0.3574,
      "step": 1471
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4033366361918805e-05,
      "loss": 0.3727,
      "step": 1472
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4025572220412998e-05,
      "loss": 0.3543,
      "step": 1473
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4017775159396759e-05,
      "loss": 0.3594,
      "step": 1474
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.400997518452484e-05,
      "loss": 0.3401,
      "step": 1475
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4002172301454103e-05,
      "loss": 0.3035,
      "step": 1476
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.3994366515843515e-05,
      "loss": 0.324,
      "step": 1477
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.398655783335416e-05,
      "loss": 0.358,
      "step": 1478
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.397874625964921e-05,
      "loss": 0.3419,
      "step": 1479
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3970931800393943e-05,
      "loss": 0.3759,
      "step": 1480
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3963114461255722e-05,
      "loss": 0.3687,
      "step": 1481
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3955294247904006e-05,
      "loss": 0.3572,
      "step": 1482
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3947471166010332e-05,
      "loss": 0.3297,
      "step": 1483
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3939645221248325e-05,
      "loss": 0.3391,
      "step": 1484
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3931816419293677e-05,
      "loss": 0.3527,
      "step": 1485
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3923984765824159e-05,
      "loss": 0.3496,
      "step": 1486
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.391615026651961e-05,
      "loss": 0.3627,
      "step": 1487
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3908312927061926e-05,
      "loss": 0.364,
      "step": 1488
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3900472753135075e-05,
      "loss": 0.3739,
      "step": 1489
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3892629750425066e-05,
      "loss": 0.3456,
      "step": 1490
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3884783924619974e-05,
      "loss": 0.3302,
      "step": 1491
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3876935281409907e-05,
      "loss": 0.3418,
      "step": 1492
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3869083826487026e-05,
      "loss": 0.3316,
      "step": 1493
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3861229565545532e-05,
      "loss": 0.3554,
      "step": 1494
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3853372504281657e-05,
      "loss": 0.3555,
      "step": 1495
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3845512648393663e-05,
      "loss": 0.3443,
      "step": 1496
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3837650003581838e-05,
      "loss": 0.3222,
      "step": 1497
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3829784575548502e-05,
      "loss": 0.3345,
      "step": 1498
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3821916369997979e-05,
      "loss": 0.354,
      "step": 1499
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3814045392636622e-05,
      "loss": 0.3477,
      "step": 1500
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3806171649172782e-05,
      "loss": 0.3584,
      "step": 1501
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3798295145316823e-05,
      "loss": 0.3747,
      "step": 1502
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3790415886781112e-05,
      "loss": 0.3484,
      "step": 1503
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3782533879280009e-05,
      "loss": 0.3506,
      "step": 1504
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.377464912852987e-05,
      "loss": 0.3459,
      "step": 1505
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.376676164024904e-05,
      "loss": 0.3276,
      "step": 1506
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.375887142015785e-05,
      "loss": 0.3259,
      "step": 1507
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3750978473978611e-05,
      "loss": 0.34,
      "step": 1508
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3743082807435615e-05,
      "loss": 0.3321,
      "step": 1509
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3735184426255117e-05,
      "loss": 0.3384,
      "step": 1510
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3727283336165356e-05,
      "loss": 0.3441,
      "step": 1511
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3719379542896523e-05,
      "loss": 0.3355,
      "step": 1512
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3711473052180772e-05,
      "loss": 0.3704,
      "step": 1513
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3703563869752222e-05,
      "loss": 0.3416,
      "step": 1514
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3695652001346928e-05,
      "loss": 0.3373,
      "step": 1515
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3687737452702909e-05,
      "loss": 0.3409,
      "step": 1516
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3679820229560119e-05,
      "loss": 0.363,
      "step": 1517
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3671900337660454e-05,
      "loss": 0.3464,
      "step": 1518
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3663977782747741e-05,
      "loss": 0.3572,
      "step": 1519
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3656052570567751e-05,
      "loss": 0.3384,
      "step": 1520
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3648124706868166e-05,
      "loss": 0.345,
      "step": 1521
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.36401941973986e-05,
      "loss": 0.3319,
      "step": 1522
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3632261047910587e-05,
      "loss": 0.3506,
      "step": 1523
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3624325264157571e-05,
      "loss": 0.3094,
      "step": 1524
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3616386851894909e-05,
      "loss": 0.3563,
      "step": 1525
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3608445816879865e-05,
      "loss": 0.3431,
      "step": 1526
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.36005021648716e-05,
      "loss": 0.3458,
      "step": 1527
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3592555901631181e-05,
      "loss": 0.3493,
      "step": 1528
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3584607032921566e-05,
      "loss": 0.3618,
      "step": 1529
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3576655564507596e-05,
      "loss": 0.3598,
      "step": 1530
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3568701502156009e-05,
      "loss": 0.323,
      "step": 1531
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3560744851635415e-05,
      "loss": 0.3503,
      "step": 1532
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.35527856187163e-05,
      "loss": 0.3426,
      "step": 1533
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3544823809171031e-05,
      "loss": 0.3476,
      "step": 1534
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3536859428773841e-05,
      "loss": 0.3631,
      "step": 1535
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3528892483300821e-05,
      "loss": 0.3425,
      "step": 1536
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3520922978529931e-05,
      "loss": 0.3371,
      "step": 1537
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3512950920240984e-05,
      "loss": 0.3634,
      "step": 1538
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3504976314215643e-05,
      "loss": 0.3853,
      "step": 1539
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.349699916623742e-05,
      "loss": 0.3445,
      "step": 1540
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.348901948209167e-05,
      "loss": 0.3388,
      "step": 1541
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3481037267565587e-05,
      "loss": 0.3344,
      "step": 1542
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3473052528448203e-05,
      "loss": 0.3349,
      "step": 1543
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.346506527053038e-05,
      "loss": 0.3403,
      "step": 1544
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3457075499604803e-05,
      "loss": 0.355,
      "step": 1545
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3449083221465986e-05,
      "loss": 0.3409,
      "step": 1546
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3441088441910254e-05,
      "loss": 0.352,
      "step": 1547
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.343309116673575e-05,
      "loss": 0.3196,
      "step": 1548
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3425091401742428e-05,
      "loss": 0.325,
      "step": 1549
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3417089152732049e-05,
      "loss": 0.3352,
      "step": 1550
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3409084425508165e-05,
      "loss": 0.3125,
      "step": 1551
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.340107722587614e-05,
      "loss": 0.3396,
      "step": 1552
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3393067559643125e-05,
      "loss": 0.3401,
      "step": 1553
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3385055432618053e-05,
      "loss": 0.3416,
      "step": 1554
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3377040850611652e-05,
      "loss": 0.336,
      "step": 1555
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3369023819436424e-05,
      "loss": 0.3221,
      "step": 1556
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3361004344906652e-05,
      "loss": 0.3792,
      "step": 1557
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3352982432838388e-05,
      "loss": 0.3552,
      "step": 1558
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3344958089049452e-05,
      "loss": 0.3517,
      "step": 1559
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3336931319359428e-05,
      "loss": 0.3642,
      "step": 1560
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.332890212958966e-05,
      "loss": 0.3265,
      "step": 1561
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3320870525563243e-05,
      "loss": 0.3316,
      "step": 1562
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.331283651310503e-05,
      "loss": 0.347,
      "step": 1563
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.330480009804162e-05,
      "loss": 0.3291,
      "step": 1564
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3296761286201343e-05,
      "loss": 0.3475,
      "step": 1565
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3288720083414282e-05,
      "loss": 0.328,
      "step": 1566
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3280676495512244e-05,
      "loss": 0.3663,
      "step": 1567
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3272630528328774e-05,
      "loss": 0.3343,
      "step": 1568
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3264582187699136e-05,
      "loss": 0.3215,
      "step": 1569
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3256531479460315e-05,
      "loss": 0.3666,
      "step": 1570
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3248478409451017e-05,
      "loss": 0.3443,
      "step": 1571
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.324042298351166e-05,
      "loss": 0.3406,
      "step": 1572
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3232365207484373e-05,
      "loss": 0.3763,
      "step": 1573
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3224305087212979e-05,
      "loss": 0.3029,
      "step": 1574
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3216242628543017e-05,
      "loss": 0.3273,
      "step": 1575
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3208177837321706e-05,
      "loss": 0.3462,
      "step": 1576
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3200110719397969e-05,
      "loss": 0.3026,
      "step": 1577
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3192041280622409e-05,
      "loss": 0.365,
      "step": 1578
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3183969526847318e-05,
      "loss": 0.334,
      "step": 1579
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3175895463926657e-05,
      "loss": 0.3477,
      "step": 1580
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3167819097716077e-05,
      "loss": 0.3407,
      "step": 1581
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.315974043407288e-05,
      "loss": 0.3363,
      "step": 1582
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3151659478856056e-05,
      "loss": 0.328,
      "step": 1583
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3143576237926237e-05,
      "loss": 0.351,
      "step": 1584
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3135490717145726e-05,
      "loss": 0.38,
      "step": 1585
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3127402922378479e-05,
      "loss": 0.3377,
      "step": 1586
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3119312859490089e-05,
      "loss": 0.3371,
      "step": 1587
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3111220534347805e-05,
      "loss": 0.3333,
      "step": 1588
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3103125952820515e-05,
      "loss": 0.3256,
      "step": 1589
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3095029120778743e-05,
      "loss": 0.3445,
      "step": 1590
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3086930044094642e-05,
      "loss": 0.3453,
      "step": 1591
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3078828728641994e-05,
      "loss": 0.35,
      "step": 1592
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3070725180296213e-05,
      "loss": 0.3555,
      "step": 1593
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3062619404934319e-05,
      "loss": 0.3454,
      "step": 1594
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3054511408434954e-05,
      "loss": 0.3449,
      "step": 1595
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.304640119667837e-05,
      "loss": 0.3292,
      "step": 1596
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.303828877554643e-05,
      "loss": 0.3324,
      "step": 1597
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.303017415092259e-05,
      "loss": 0.3156,
      "step": 1598
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3022057328691915e-05,
      "loss": 0.31,
      "step": 1599
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3013938314741053e-05,
      "loss": 0.372,
      "step": 1600
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3005817114958248e-05,
      "loss": 0.331,
      "step": 1601
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.299769373523333e-05,
      "loss": 0.3392,
      "step": 1602
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.2989568181457704e-05,
      "loss": 0.3502,
      "step": 1603
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.298144045952436e-05,
      "loss": 0.3325,
      "step": 1604
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2973310575327855e-05,
      "loss": 0.3368,
      "step": 1605
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2965178534764311e-05,
      "loss": 0.3487,
      "step": 1606
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2957044343731422e-05,
      "loss": 0.3271,
      "step": 1607
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2948908008128437e-05,
      "loss": 0.3398,
      "step": 1608
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.294076953385616e-05,
      "loss": 0.3558,
      "step": 1609
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2932628926816948e-05,
      "loss": 0.3364,
      "step": 1610
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2924486192914705e-05,
      "loss": 0.3365,
      "step": 1611
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2916341338054872e-05,
      "loss": 0.3359,
      "step": 1612
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2908194368144437e-05,
      "loss": 0.3229,
      "step": 1613
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2900045289091916e-05,
      "loss": 0.3511,
      "step": 1614
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2891894106807354e-05,
      "loss": 0.3374,
      "step": 1615
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2883740827202332e-05,
      "loss": 0.3574,
      "step": 1616
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2875585456189929e-05,
      "loss": 0.33,
      "step": 1617
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2867427999684765e-05,
      "loss": 0.3305,
      "step": 1618
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2859268463602965e-05,
      "loss": 0.3547,
      "step": 1619
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.285110685386215e-05,
      "loss": 0.3375,
      "step": 1620
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2842943176381463e-05,
      "loss": 0.3139,
      "step": 1621
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2834777437081535e-05,
      "loss": 0.3205,
      "step": 1622
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.28266096418845e-05,
      "loss": 0.338,
      "step": 1623
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2818439796713973e-05,
      "loss": 0.3469,
      "step": 1624
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2810267907495063e-05,
      "loss": 0.3354,
      "step": 1625
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2802093980154363e-05,
      "loss": 0.3417,
      "step": 1626
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2793918020619937e-05,
      "loss": 0.3288,
      "step": 1627
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.278574003482133e-05,
      "loss": 0.3398,
      "step": 1628
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2777560028689551e-05,
      "loss": 0.3379,
      "step": 1629
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2769378008157083e-05,
      "loss": 0.3383,
      "step": 1630
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2761193979157852e-05,
      "loss": 0.3467,
      "step": 1631
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2753007947627263e-05,
      "loss": 0.2984,
      "step": 1632
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2744819919502156e-05,
      "loss": 0.3581,
      "step": 1633
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2736629900720832e-05,
      "loss": 0.353,
      "step": 1634
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.272843789722302e-05,
      "loss": 0.3359,
      "step": 1635
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2720243914949906e-05,
      "loss": 0.3543,
      "step": 1636
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2712047959844097e-05,
      "loss": 0.3167,
      "step": 1637
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.270385003784964e-05,
      "loss": 0.3377,
      "step": 1638
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2695650154912005e-05,
      "loss": 0.3408,
      "step": 1639
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2687448316978084e-05,
      "loss": 0.3356,
      "step": 1640
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2679244529996182e-05,
      "loss": 0.2986,
      "step": 1641
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2671038799916028e-05,
      "loss": 0.3357,
      "step": 1642
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2662831132688755e-05,
      "loss": 0.3324,
      "step": 1643
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2654621534266897e-05,
      "loss": 0.3532,
      "step": 1644
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2646410010604397e-05,
      "loss": 0.3554,
      "step": 1645
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2638196567656585e-05,
      "loss": 0.3314,
      "step": 1646
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2629981211380188e-05,
      "loss": 0.3712,
      "step": 1647
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.262176394773332e-05,
      "loss": 0.337,
      "step": 1648
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2613544782675485e-05,
      "loss": 0.3317,
      "step": 1649
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2605323722167549e-05,
      "loss": 0.3233,
      "step": 1650
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2597100772171767e-05,
      "loss": 0.3503,
      "step": 1651
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2588875938651762e-05,
      "loss": 0.32,
      "step": 1652
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2580649227572522e-05,
      "loss": 0.355,
      "step": 1653
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.257242064490039e-05,
      "loss": 0.3359,
      "step": 1654
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.256419019660308e-05,
      "loss": 0.3452,
      "step": 1655
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2555957888649648e-05,
      "loss": 0.3409,
      "step": 1656
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2547723727010501e-05,
      "loss": 0.3377,
      "step": 1657
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2539487717657393e-05,
      "loss": 0.3466,
      "step": 1658
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2531249866563415e-05,
      "loss": 0.3534,
      "step": 1659
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2523010179703e-05,
      "loss": 0.3338,
      "step": 1660
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2514768663051902e-05,
      "loss": 0.3119,
      "step": 1661
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2506525322587207e-05,
      "loss": 0.3119,
      "step": 1662
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2498280164287331e-05,
      "loss": 0.3687,
      "step": 1663
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2490033194131992e-05,
      "loss": 0.3664,
      "step": 1664
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.248178441810224e-05,
      "loss": 0.319,
      "step": 1665
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2473533842180423e-05,
      "loss": 0.3413,
      "step": 1666
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2465281472350199e-05,
      "loss": 0.35,
      "step": 1667
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2457027314596524e-05,
      "loss": 0.3289,
      "step": 1668
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2448771374905655e-05,
      "loss": 0.353,
      "step": 1669
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2440513659265136e-05,
      "loss": 0.3706,
      "step": 1670
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2432254173663805e-05,
      "loss": 0.3223,
      "step": 1671
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2423992924091776e-05,
      "loss": 0.3586,
      "step": 1672
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.241572991654045e-05,
      "loss": 0.3501,
      "step": 1673
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2407465157002502e-05,
      "loss": 0.3859,
      "step": 1674
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.239919865147187e-05,
      "loss": 0.3288,
      "step": 1675
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2390930405943766e-05,
      "loss": 0.3344,
      "step": 1676
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2382660426414661e-05,
      "loss": 0.3366,
      "step": 1677
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2374388718882289e-05,
      "loss": 0.3391,
      "step": 1678
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.236611528934562e-05,
      "loss": 0.3247,
      "step": 1679
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2357840143804898e-05,
      "loss": 0.3307,
      "step": 1680
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2349563288261592e-05,
      "loss": 0.3386,
      "step": 1681
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2341284728718419e-05,
      "loss": 0.3689,
      "step": 1682
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.233300447117933e-05,
      "loss": 0.3301,
      "step": 1683
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.232472252164951e-05,
      "loss": 0.3259,
      "step": 1684
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2316438886135371e-05,
      "loss": 0.3396,
      "step": 1685
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2308153570644541e-05,
      "loss": 0.359,
      "step": 1686
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2299866581185872e-05,
      "loss": 0.3671,
      "step": 1687
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2291577923769433e-05,
      "loss": 0.3527,
      "step": 1688
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.22832876044065e-05,
      "loss": 0.3557,
      "step": 1689
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2274995629109545e-05,
      "loss": 0.3504,
      "step": 1690
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2266702003892256e-05,
      "loss": 0.3575,
      "step": 1691
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.225840673476951e-05,
      "loss": 0.3169,
      "step": 1692
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2250109827757379e-05,
      "loss": 0.3384,
      "step": 1693
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2241811288873116e-05,
      "loss": 0.3242,
      "step": 1694
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2233511124135164e-05,
      "loss": 0.3426,
      "step": 1695
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2225209339563144e-05,
      "loss": 0.3201,
      "step": 1696
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2216905941177854e-05,
      "loss": 0.3039,
      "step": 1697
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2208600935001258e-05,
      "loss": 0.3553,
      "step": 1698
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2200294327056488e-05,
      "loss": 0.3462,
      "step": 1699
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2191986123367838e-05,
      "loss": 0.3215,
      "step": 1700
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2183676329960756e-05,
      "loss": 0.3748,
      "step": 1701
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2175364952861849e-05,
      "loss": 0.3448,
      "step": 1702
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2167051998098866e-05,
      "loss": 0.3369,
      "step": 1703
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.215873747170071e-05,
      "loss": 0.3418,
      "step": 1704
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2150421379697408e-05,
      "loss": 0.3513,
      "step": 1705
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2142103728120138e-05,
      "loss": 0.3385,
      "step": 1706
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2133784523001202e-05,
      "loss": 0.3217,
      "step": 1707
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.212546377037403e-05,
      "loss": 0.3506,
      "step": 1708
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2117141476273169e-05,
      "loss": 0.3367,
      "step": 1709
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2108817646734292e-05,
      "loss": 0.3377,
      "step": 1710
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2100492287794186e-05,
      "loss": 0.318,
      "step": 1711
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2092165405490737e-05,
      "loss": 0.3475,
      "step": 1712
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2083837005862946e-05,
      "loss": 0.3381,
      "step": 1713
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2075507094950912e-05,
      "loss": 0.3733,
      "step": 1714
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2067175678795826e-05,
      "loss": 0.3364,
      "step": 1715
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2058842763439972e-05,
      "loss": 0.3039,
      "step": 1716
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2050508354926727e-05,
      "loss": 0.3661,
      "step": 1717
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2042172459300546e-05,
      "loss": 0.3165,
      "step": 1718
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.203383508260696e-05,
      "loss": 0.3502,
      "step": 1719
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.2025496230892577e-05,
      "loss": 0.3415,
      "step": 1720
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.2017155910205079e-05,
      "loss": 0.3275,
      "step": 1721
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.2008814126593207e-05,
      "loss": 0.3524,
      "step": 1722
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.2000470886106766e-05,
      "loss": 0.3396,
      "step": 1723
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1992126194796617e-05,
      "loss": 0.3185,
      "step": 1724
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.198378005871467e-05,
      "loss": 0.344,
      "step": 1725
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1975432483913892e-05,
      "loss": 0.3747,
      "step": 1726
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1967083476448282e-05,
      "loss": 0.3256,
      "step": 1727
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1958733042372886e-05,
      "loss": 0.3773,
      "step": 1728
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1950381187743782e-05,
      "loss": 0.3127,
      "step": 1729
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1942027918618076e-05,
      "loss": 0.335,
      "step": 1730
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1933673241053905e-05,
      "loss": 0.3301,
      "step": 1731
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.192531716111042e-05,
      "loss": 0.3481,
      "step": 1732
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1916959684847802e-05,
      "loss": 0.3356,
      "step": 1733
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1908600818327227e-05,
      "loss": 0.3344,
      "step": 1734
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1900240567610894e-05,
      "loss": 0.3145,
      "step": 1735
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1891878938762001e-05,
      "loss": 0.3354,
      "step": 1736
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.188351593784474e-05,
      "loss": 0.3236,
      "step": 1737
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1875151570924307e-05,
      "loss": 0.3305,
      "step": 1738
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1866785844066884e-05,
      "loss": 0.3139,
      "step": 1739
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.185841876333964e-05,
      "loss": 0.3375,
      "step": 1740
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1850050334810725e-05,
      "loss": 0.3511,
      "step": 1741
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1841680564549268e-05,
      "loss": 0.3452,
      "step": 1742
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1833309458625368e-05,
      "loss": 0.3225,
      "step": 1743
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1824937023110102e-05,
      "loss": 0.3221,
      "step": 1744
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1816563264075498e-05,
      "loss": 0.3524,
      "step": 1745
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1808188187594549e-05,
      "loss": 0.3098,
      "step": 1746
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1799811799741209e-05,
      "loss": 0.3394,
      "step": 1747
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.179143410659038e-05,
      "loss": 0.3246,
      "step": 1748
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1783055114217905e-05,
      "loss": 0.3618,
      "step": 1749
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1774674828700574e-05,
      "loss": 0.3709,
      "step": 1750
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1766293256116118e-05,
      "loss": 0.3324,
      "step": 1751
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1757910402543193e-05,
      "loss": 0.335,
      "step": 1752
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1749526274061394e-05,
      "loss": 0.3448,
      "step": 1753
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1741140876751232e-05,
      "loss": 0.3345,
      "step": 1754
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1732754216694145e-05,
      "loss": 0.3593,
      "step": 1755
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.172436629997248e-05,
      "loss": 0.3362,
      "step": 1756
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.17159771326695e-05,
      "loss": 0.328,
      "step": 1757
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1707586720869375e-05,
      "loss": 0.3432,
      "step": 1758
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1699195070657179e-05,
      "loss": 0.3424,
      "step": 1759
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1690802188118878e-05,
      "loss": 0.3119,
      "step": 1760
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1682408079341333e-05,
      "loss": 0.3656,
      "step": 1761
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1674012750412302e-05,
      "loss": 0.3431,
      "step": 1762
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1665616207420421e-05,
      "loss": 0.3458,
      "step": 1763
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1657218456455207e-05,
      "loss": 0.3393,
      "step": 1764
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1648819503607052e-05,
      "loss": 0.3324,
      "step": 1765
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1640419354967225e-05,
      "loss": 0.3219,
      "step": 1766
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1632018016627859e-05,
      "loss": 0.3579,
      "step": 1767
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1623615494681952e-05,
      "loss": 0.3222,
      "step": 1768
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1615211795223354e-05,
      "loss": 0.3384,
      "step": 1769
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1606806924346776e-05,
      "loss": 0.3429,
      "step": 1770
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1598400888147771e-05,
      "loss": 0.3469,
      "step": 1771
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.158999369272275e-05,
      "loss": 0.3306,
      "step": 1772
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1581585344168951e-05,
      "loss": 0.3323,
      "step": 1773
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1573175848584455e-05,
      "loss": 0.3276,
      "step": 1774
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1564765212068174e-05,
      "loss": 0.3297,
      "step": 1775
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1556353440719845e-05,
      "loss": 0.3312,
      "step": 1776
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1547940540640039e-05,
      "loss": 0.3391,
      "step": 1777
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1539526517930125e-05,
      "loss": 0.3299,
      "step": 1778
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1531111378692305e-05,
      "loss": 0.3547,
      "step": 1779
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1522695129029581e-05,
      "loss": 0.333,
      "step": 1780
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1514277775045768e-05,
      "loss": 0.3234,
      "step": 1781
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1505859322845472e-05,
      "loss": 0.349,
      "step": 1782
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1497439778534104e-05,
      "loss": 0.3288,
      "step": 1783
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1489019148217862e-05,
      "loss": 0.3089,
      "step": 1784
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.148059743800373e-05,
      "loss": 0.3248,
      "step": 1785
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1472174653999484e-05,
      "loss": 0.3326,
      "step": 1786
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1463750802313668e-05,
      "loss": 0.354,
      "step": 1787
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1455325889055616e-05,
      "loss": 0.349,
      "step": 1788
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1446899920335407e-05,
      "loss": 0.3129,
      "step": 1789
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1438472902263906e-05,
      "loss": 0.3315,
      "step": 1790
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1430044840952734e-05,
      "loss": 0.3261,
      "step": 1791
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1421615742514267e-05,
      "loss": 0.3181,
      "step": 1792
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1413185613061632e-05,
      "loss": 0.3043,
      "step": 1793
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.140475445870871e-05,
      "loss": 0.3372,
      "step": 1794
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1396322285570119e-05,
      "loss": 0.3383,
      "step": 1795
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1387889099761211e-05,
      "loss": 0.3487,
      "step": 1796
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1379454907398084e-05,
      "loss": 0.3364,
      "step": 1797
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1371019714597562e-05,
      "loss": 0.3363,
      "step": 1798
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1362583527477195e-05,
      "loss": 0.3331,
      "step": 1799
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1354146352155248e-05,
      "loss": 0.3429,
      "step": 1800
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1345708194750709e-05,
      "loss": 0.314,
      "step": 1801
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1337269061383278e-05,
      "loss": 0.3252,
      "step": 1802
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1328828958173359e-05,
      "loss": 0.3518,
      "step": 1803
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1320387891242064e-05,
      "loss": 0.3404,
      "step": 1804
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1311945866711199e-05,
      "loss": 0.3169,
      "step": 1805
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1303502890703269e-05,
      "loss": 0.334,
      "step": 1806
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1295058969341469e-05,
      "loss": 0.3168,
      "step": 1807
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1286614108749672e-05,
      "loss": 0.329,
      "step": 1808
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1278168315052445e-05,
      "loss": 0.2925,
      "step": 1809
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1269721594375022e-05,
      "loss": 0.34,
      "step": 1810
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1261273952843312e-05,
      "loss": 0.3295,
      "step": 1811
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.125282539658389e-05,
      "loss": 0.3391,
      "step": 1812
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1244375931724e-05,
      "loss": 0.3483,
      "step": 1813
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1235925564391536e-05,
      "loss": 0.3216,
      "step": 1814
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1227474300715056e-05,
      "loss": 0.3425,
      "step": 1815
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1219022146823762e-05,
      "loss": 0.3153,
      "step": 1816
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1210569108847504e-05,
      "loss": 0.3451,
      "step": 1817
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1202115192916772e-05,
      "loss": 0.3454,
      "step": 1818
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1193660405162694e-05,
      "loss": 0.3238,
      "step": 1819
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.118520475171703e-05,
      "loss": 0.3457,
      "step": 1820
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1176748238712162e-05,
      "loss": 0.3283,
      "step": 1821
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1168290872281111e-05,
      "loss": 0.3362,
      "step": 1822
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1159832658557498e-05,
      "loss": 0.3428,
      "step": 1823
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1151373603675575e-05,
      "loss": 0.3143,
      "step": 1824
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1142913713770187e-05,
      "loss": 0.3151,
      "step": 1825
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1134452994976801e-05,
      "loss": 0.335,
      "step": 1826
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1125991453431474e-05,
      "loss": 0.3474,
      "step": 1827
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1117529095270867e-05,
      "loss": 0.3253,
      "step": 1828
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1109065926632227e-05,
      "loss": 0.3237,
      "step": 1829
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1100601953653393e-05,
      "loss": 0.3507,
      "step": 1830
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1092137182472787e-05,
      "loss": 0.3384,
      "step": 1831
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1083671619229409e-05,
      "loss": 0.3235,
      "step": 1832
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.107520527006283e-05,
      "loss": 0.3438,
      "step": 1833
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1066738141113198e-05,
      "loss": 0.3354,
      "step": 1834
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.105827023852122e-05,
      "loss": 0.3382,
      "step": 1835
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1049801568428171e-05,
      "loss": 0.3428,
      "step": 1836
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1041332136975874e-05,
      "loss": 0.3108,
      "step": 1837
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1032861950306715e-05,
      "loss": 0.3467,
      "step": 1838
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1024391014563618e-05,
      "loss": 0.3015,
      "step": 1839
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.101591933589005e-05,
      "loss": 0.3116,
      "step": 1840
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1007446920430026e-05,
      "loss": 0.3037,
      "step": 1841
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.0998973774328086e-05,
      "loss": 0.3011,
      "step": 1842
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.0990499903729307e-05,
      "loss": 0.3386,
      "step": 1843
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.0982025314779287e-05,
      "loss": 0.3531,
      "step": 1844
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.0973550013624148e-05,
      "loss": 0.341,
      "step": 1845
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0965074006410525e-05,
      "loss": 0.3489,
      "step": 1846
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0956597299285564e-05,
      "loss": 0.3205,
      "step": 1847
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0948119898396925e-05,
      "loss": 0.3342,
      "step": 1848
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0939641809892767e-05,
      "loss": 0.3393,
      "step": 1849
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0931163039921749e-05,
      "loss": 0.3609,
      "step": 1850
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.092268359463302e-05,
      "loss": 0.3323,
      "step": 1851
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0914203480176226e-05,
      "loss": 0.3322,
      "step": 1852
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0905722702701494e-05,
      "loss": 0.3486,
      "step": 1853
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0897241268359433e-05,
      "loss": 0.3406,
      "step": 1854
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0888759183301126e-05,
      "loss": 0.3421,
      "step": 1855
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0880276453678127e-05,
      "loss": 0.3468,
      "step": 1856
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0871793085642469e-05,
      "loss": 0.3597,
      "step": 1857
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.086330908534663e-05,
      "loss": 0.3249,
      "step": 1858
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0854824458943558e-05,
      "loss": 0.3385,
      "step": 1859
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0846339212586654e-05,
      "loss": 0.3329,
      "step": 1860
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0837853352429769e-05,
      "loss": 0.3296,
      "step": 1861
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0829366884627195e-05,
      "loss": 0.3172,
      "step": 1862
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0820879815333667e-05,
      "loss": 0.3261,
      "step": 1863
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0812392150704361e-05,
      "loss": 0.3343,
      "step": 1864
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0803903896894877e-05,
      "loss": 0.3259,
      "step": 1865
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0795415060061242e-05,
      "loss": 0.3413,
      "step": 1866
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0786925646359915e-05,
      "loss": 0.3442,
      "step": 1867
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0778435661947769e-05,
      "loss": 0.3543,
      "step": 1868
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.076994511298208e-05,
      "loss": 0.3356,
      "step": 1869
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0761454005620553e-05,
      "loss": 0.3316,
      "step": 1870
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0752962346021283e-05,
      "loss": 0.3494,
      "step": 1871
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0744470140342775e-05,
      "loss": 0.3358,
      "step": 1872
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0735977394743919e-05,
      "loss": 0.3528,
      "step": 1873
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0727484115384009e-05,
      "loss": 0.3425,
      "step": 1874
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0718990308422717e-05,
      "loss": 0.3464,
      "step": 1875
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.07104959800201e-05,
      "loss": 0.2968,
      "step": 1876
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0702001136336597e-05,
      "loss": 0.3195,
      "step": 1877
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0693505783533019e-05,
      "loss": 0.3107,
      "step": 1878
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0685009927770542e-05,
      "loss": 0.3345,
      "step": 1879
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.067651357521071e-05,
      "loss": 0.332,
      "step": 1880
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0668016732015426e-05,
      "loss": 0.3538,
      "step": 1881
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0659519404346955e-05,
      "loss": 0.327,
      "step": 1882
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0651021598367905e-05,
      "loss": 0.3637,
      "step": 1883
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.064252332024123e-05,
      "loss": 0.3355,
      "step": 1884
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0634024576130238e-05,
      "loss": 0.3043,
      "step": 1885
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0625525372198564e-05,
      "loss": 0.3527,
      "step": 1886
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0617025714610181e-05,
      "loss": 0.3459,
      "step": 1887
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0608525609529388e-05,
      "loss": 0.3452,
      "step": 1888
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0600025063120809e-05,
      "loss": 0.3459,
      "step": 1889
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0591524081549395e-05,
      "loss": 0.3424,
      "step": 1890
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0583022670980405e-05,
      "loss": 0.3184,
      "step": 1891
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0574520837579409e-05,
      "loss": 0.292,
      "step": 1892
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.056601858751229e-05,
      "loss": 0.3087,
      "step": 1893
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.055751592694523e-05,
      "loss": 0.3366,
      "step": 1894
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.05490128620447e-05,
      "loss": 0.3575,
      "step": 1895
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0540509398977482e-05,
      "loss": 0.3238,
      "step": 1896
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0532005543910629e-05,
      "loss": 0.3453,
      "step": 1897
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0523501303011491e-05,
      "loss": 0.3302,
      "step": 1898
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.051499668244769e-05,
      "loss": 0.3337,
      "step": 1899
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0506491688387128e-05,
      "loss": 0.3325,
      "step": 1900
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.049798632699798e-05,
      "loss": 0.3405,
      "step": 1901
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0489480604448676e-05,
      "loss": 0.3431,
      "step": 1902
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.048097452690792e-05,
      "loss": 0.3263,
      "step": 1903
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0472468100544667e-05,
      "loss": 0.3432,
      "step": 1904
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.046396133152813e-05,
      "loss": 0.3482,
      "step": 1905
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0455454226027766e-05,
      "loss": 0.3368,
      "step": 1906
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0446946790213275e-05,
      "loss": 0.3379,
      "step": 1907
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0438439030254604e-05,
      "loss": 0.3198,
      "step": 1908
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.042993095232193e-05,
      "loss": 0.3373,
      "step": 1909
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0421422562585657e-05,
      "loss": 0.3544,
      "step": 1910
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0412913867216418e-05,
      "loss": 0.3195,
      "step": 1911
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0404404872385076e-05,
      "loss": 0.315,
      "step": 1912
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0395895584262696e-05,
      "loss": 0.3364,
      "step": 1913
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0387386009020569e-05,
      "loss": 0.3226,
      "step": 1914
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0378876152830181e-05,
      "loss": 0.3024,
      "step": 1915
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.037036602186324e-05,
      "loss": 0.3626,
      "step": 1916
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0361855622291637e-05,
      "loss": 0.3296,
      "step": 1917
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0353344960287461e-05,
      "loss": 0.3028,
      "step": 1918
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0344834042022998e-05,
      "loss": 0.3478,
      "step": 1919
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0336322873670714e-05,
      "loss": 0.3509,
      "step": 1920
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.032781146140326e-05,
      "loss": 0.3334,
      "step": 1921
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.031929981139346e-05,
      "loss": 0.318,
      "step": 1922
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0310787929814315e-05,
      "loss": 0.3385,
      "step": 1923
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0302275822838987e-05,
      "loss": 0.3413,
      "step": 1924
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0293763496640808e-05,
      "loss": 0.3275,
      "step": 1925
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.028525095739327e-05,
      "loss": 0.3342,
      "step": 1926
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0276738211270012e-05,
      "loss": 0.3451,
      "step": 1927
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0268225264444829e-05,
      "loss": 0.3101,
      "step": 1928
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0259712123091659e-05,
      "loss": 0.3366,
      "step": 1929
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0251198793384583e-05,
      "loss": 0.3159,
      "step": 1930
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0242685281497817e-05,
      "loss": 0.3043,
      "step": 1931
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.023417159360571e-05,
      "loss": 0.3142,
      "step": 1932
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0225657735882741e-05,
      "loss": 0.3418,
      "step": 1933
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0217143714503508e-05,
      "loss": 0.3228,
      "step": 1934
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0208629535642726e-05,
      "loss": 0.3558,
      "step": 1935
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.020011520547523e-05,
      "loss": 0.3156,
      "step": 1936
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.019160073017596e-05,
      "loss": 0.3424,
      "step": 1937
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0183086115919971e-05,
      "loss": 0.3252,
      "step": 1938
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0174571368882403e-05,
      "loss": 0.3381,
      "step": 1939
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0166056495238501e-05,
      "loss": 0.3329,
      "step": 1940
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0157541501163604e-05,
      "loss": 0.3538,
      "step": 1941
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0149026392833137e-05,
      "loss": 0.3287,
      "step": 1942
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0140511176422605e-05,
      "loss": 0.3177,
      "step": 1943
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.013199585810759e-05,
      "loss": 0.314,
      "step": 1944
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0123480444063756e-05,
      "loss": 0.3512,
      "step": 1945
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0114964940466828e-05,
      "loss": 0.3171,
      "step": 1946
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0106449353492598e-05,
      "loss": 0.3349,
      "step": 1947
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0097933689316923e-05,
      "loss": 0.3433,
      "step": 1948
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0089417954115715e-05,
      "loss": 0.3269,
      "step": 1949
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0080902154064926e-05,
      "loss": 0.3688,
      "step": 1950
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0072386295340572e-05,
      "loss": 0.3132,
      "step": 1951
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0063870384118701e-05,
      "loss": 0.355,
      "step": 1952
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0055354426575406e-05,
      "loss": 0.3207,
      "step": 1953
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.00468384288868e-05,
      "loss": 0.3158,
      "step": 1954
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0038322397229044e-05,
      "loss": 0.3097,
      "step": 1955
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.002980633777831e-05,
      "loss": 0.3394,
      "step": 1956
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0021290256710793e-05,
      "loss": 0.3097,
      "step": 1957
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0012774160202708e-05,
      "loss": 0.3405,
      "step": 1958
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0004258054430276e-05,
      "loss": 0.317,
      "step": 1959
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.995741945569729e-06,
      "loss": 0.3351,
      "step": 1960
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.987225839797295e-06,
      "loss": 0.3356,
      "step": 1961
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.978709743289212e-06,
      "loss": 0.3211,
      "step": 1962
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.970193662221694e-06,
      "loss": 0.3523,
      "step": 1963
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.96167760277096e-06,
      "loss": 0.3301,
      "step": 1964
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.953161571113205e-06,
      "loss": 0.3188,
      "step": 1965
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.944645573424598e-06,
      "loss": 0.3703,
      "step": 1966
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.936129615881302e-06,
      "loss": 0.3462,
      "step": 1967
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.92761370465943e-06,
      "loss": 0.3525,
      "step": 1968
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.919097845935075e-06,
      "loss": 0.3156,
      "step": 1969
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.910582045884292e-06,
      "loss": 0.337,
      "step": 1970
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.902066310683077e-06,
      "loss": 0.3113,
      "step": 1971
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.893550646507405e-06,
      "loss": 0.3468,
      "step": 1972
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.885035059533177e-06,
      "loss": 0.3312,
      "step": 1973
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.876519555936248e-06,
      "loss": 0.3139,
      "step": 1974
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.868004141892412e-06,
      "loss": 0.3397,
      "step": 1975
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.8594888235774e-06,
      "loss": 0.3207,
      "step": 1976
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.850973607166865e-06,
      "loss": 0.3285,
      "step": 1977
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.8424584988364e-06,
      "loss": 0.3196,
      "step": 1978
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.8339435047615e-06,
      "loss": 0.2985,
      "step": 1979
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.825428631117602e-06,
      "loss": 0.3382,
      "step": 1980
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.816913884080035e-06,
      "loss": 0.3252,
      "step": 1981
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.808399269824039e-06,
      "loss": 0.3316,
      "step": 1982
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.799884794524774e-06,
      "loss": 0.3239,
      "step": 1983
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.791370464357279e-06,
      "loss": 0.3267,
      "step": 1984
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.782856285496494e-06,
      "loss": 0.3283,
      "step": 1985
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.774342264117262e-06,
      "loss": 0.3077,
      "step": 1986
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.765828406394291e-06,
      "loss": 0.3221,
      "step": 1987
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.757314718502185e-06,
      "loss": 0.3072,
      "step": 1988
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.748801206615419e-06,
      "loss": 0.3018,
      "step": 1989
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.740287876908341e-06,
      "loss": 0.3335,
      "step": 1990
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.731774735555174e-06,
      "loss": 0.3127,
      "step": 1991
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.723261788729993e-06,
      "loss": 0.3409,
      "step": 1992
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.714749042606732e-06,
      "loss": 0.3067,
      "step": 1993
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.706236503359194e-06,
      "loss": 0.3237,
      "step": 1994
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.697724177161018e-06,
      "loss": 0.3102,
      "step": 1995
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.68921207018569e-06,
      "loss": 0.3614,
      "step": 1996
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.680700188606545e-06,
      "loss": 0.3385,
      "step": 1997
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.672188538596746e-06,
      "loss": 0.3187,
      "step": 1998
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.66367712632929e-06,
      "loss": 0.3141,
      "step": 1999
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.655165957977005e-06,
      "loss": 0.3529,
      "step": 2000
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.646655039712542e-06,
      "loss": 0.3298,
      "step": 2001
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.638144377708366e-06,
      "loss": 0.3096,
      "step": 2002
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.629633978136765e-06,
      "loss": 0.3177,
      "step": 2003
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.621123847169819e-06,
      "loss": 0.3213,
      "step": 2004
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.612613990979436e-06,
      "loss": 0.3086,
      "step": 2005
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.604104415737309e-06,
      "loss": 0.3342,
      "step": 2006
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.595595127614925e-06,
      "loss": 0.3074,
      "step": 2007
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.587086132783584e-06,
      "loss": 0.3072,
      "step": 2008
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.578577437414348e-06,
      "loss": 0.3019,
      "step": 2009
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.570069047678072e-06,
      "loss": 0.3371,
      "step": 2010
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.561560969745398e-06,
      "loss": 0.3376,
      "step": 2011
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.553053209786725e-06,
      "loss": 0.3201,
      "step": 2012
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.544545773972239e-06,
      "loss": 0.3125,
      "step": 2013
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.536038668471874e-06,
      "loss": 0.3262,
      "step": 2014
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.527531899455335e-06,
      "loss": 0.3426,
      "step": 2015
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.519025473092084e-06,
      "loss": 0.2843,
      "step": 2016
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.51051939555133e-06,
      "loss": 0.338,
      "step": 2017
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.502013673002024e-06,
      "loss": 0.3248,
      "step": 2018
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.493508311612874e-06,
      "loss": 0.3379,
      "step": 2019
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.485003317552314e-06,
      "loss": 0.3285,
      "step": 2020
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.476498696988512e-06,
      "loss": 0.3028,
      "step": 2021
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.467994456089375e-06,
      "loss": 0.3033,
      "step": 2022
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.45949060102252e-06,
      "loss": 0.3195,
      "step": 2023
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.450987137955302e-06,
      "loss": 0.3045,
      "step": 2024
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.442484073054777e-06,
      "loss": 0.3205,
      "step": 2025
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.433981412487711e-06,
      "loss": 0.3428,
      "step": 2026
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.425479162420593e-06,
      "loss": 0.338,
      "step": 2027
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.416977329019598e-06,
      "loss": 0.3008,
      "step": 2028
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.408475918450606e-06,
      "loss": 0.3026,
      "step": 2029
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.399974936879193e-06,
      "loss": 0.3301,
      "step": 2030
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.391474390470619e-06,
      "loss": 0.347,
      "step": 2031
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.382974285389822e-06,
      "loss": 0.3244,
      "step": 2032
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.374474627801439e-06,
      "loss": 0.3088,
      "step": 2033
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.365975423869762e-06,
      "loss": 0.3031,
      "step": 2034
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.35747667975877e-06,
      "loss": 0.3203,
      "step": 2035
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.348978401632101e-06,
      "loss": 0.2938,
      "step": 2036
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.340480595653047e-06,
      "loss": 0.3064,
      "step": 2037
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.331983267984575e-06,
      "loss": 0.3292,
      "step": 2038
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.323486424789295e-06,
      "loss": 0.3267,
      "step": 2039
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.314990072229461e-06,
      "loss": 0.318,
      "step": 2040
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.306494216466984e-06,
      "loss": 0.2838,
      "step": 2041
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.297998863663405e-06,
      "loss": 0.3182,
      "step": 2042
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.289504019979902e-06,
      "loss": 0.3112,
      "step": 2043
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.281009691577288e-06,
      "loss": 0.3281,
      "step": 2044
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.272515884615995e-06,
      "loss": 0.2965,
      "step": 2045
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.264022605256083e-06,
      "loss": 0.3105,
      "step": 2046
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.25552985965723e-06,
      "loss": 0.3401,
      "step": 2047
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.247037653978718e-06,
      "loss": 0.2873,
      "step": 2048
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.238545994379449e-06,
      "loss": 0.3226,
      "step": 2049
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.230054887017924e-06,
      "loss": 0.3121,
      "step": 2050
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.221564338052236e-06,
      "loss": 0.329,
      "step": 2051
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.213074353640088e-06,
      "loss": 0.3214,
      "step": 2052
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.204584939938763e-06,
      "loss": 0.3191,
      "step": 2053
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.196096103105127e-06,
      "loss": 0.3495,
      "step": 2054
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.187607849295644e-06,
      "loss": 0.3285,
      "step": 2055
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.179120184666333e-06,
      "loss": 0.3292,
      "step": 2056
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.17063311537281e-06,
      "loss": 0.3207,
      "step": 2057
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.162146647570235e-06,
      "loss": 0.3162,
      "step": 2058
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.153660787413346e-06,
      "loss": 0.3481,
      "step": 2059
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.145175541056446e-06,
      "loss": 0.3389,
      "step": 2060
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.136690914653377e-06,
      "loss": 0.3306,
      "step": 2061
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.128206914357535e-06,
      "loss": 0.3524,
      "step": 2062
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.119723546321875e-06,
      "loss": 0.3107,
      "step": 2063
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.11124081669888e-06,
      "loss": 0.3018,
      "step": 2064
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.102758731640568e-06,
      "loss": 0.3121,
      "step": 2065
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.094277297298508e-06,
      "loss": 0.3057,
      "step": 2066
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.085796519823774e-06,
      "loss": 0.3122,
      "step": 2067
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.07731640536698e-06,
      "loss": 0.3524,
      "step": 2068
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.068836960078256e-06,
      "loss": 0.3483,
      "step": 2069
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.060358190107235e-06,
      "loss": 0.343,
      "step": 2070
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.051880101603077e-06,
      "loss": 0.3512,
      "step": 2071
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.043402700714439e-06,
      "loss": 0.3599,
      "step": 2072
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.034925993589478e-06,
      "loss": 0.3237,
      "step": 2073
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.026449986375855e-06,
      "loss": 0.3392,
      "step": 2074
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.017974685220716e-06,
      "loss": 0.3368,
      "step": 2075
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.009500096270694e-06,
      "loss": 0.3224,
      "step": 2076
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.001026225671917e-06,
      "loss": 0.2956,
      "step": 2077
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.992553079569976e-06,
      "loss": 0.324,
      "step": 2078
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.984080664109953e-06,
      "loss": 0.3063,
      "step": 2079
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.975608985436389e-06,
      "loss": 0.3276,
      "step": 2080
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.967138049693287e-06,
      "loss": 0.3529,
      "step": 2081
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.958667863024127e-06,
      "loss": 0.3362,
      "step": 2082
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.950198431571832e-06,
      "loss": 0.3424,
      "step": 2083
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.941729761478783e-06,
      "loss": 0.3417,
      "step": 2084
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.933261858886805e-06,
      "loss": 0.3259,
      "step": 2085
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.924794729937175e-06,
      "loss": 0.3216,
      "step": 2086
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.916328380770595e-06,
      "loss": 0.3151,
      "step": 2087
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.907862817527216e-06,
      "loss": 0.3045,
      "step": 2088
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.899398046346608e-06,
      "loss": 0.3254,
      "step": 2089
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.890934073367776e-06,
      "loss": 0.2942,
      "step": 2090
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.882470904729138e-06,
      "loss": 0.338,
      "step": 2091
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.874008546568528e-06,
      "loss": 0.3113,
      "step": 2092
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.865547005023202e-06,
      "loss": 0.3327,
      "step": 2093
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.857086286229818e-06,
      "loss": 0.318,
      "step": 2094
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.848626396324429e-06,
      "loss": 0.3311,
      "step": 2095
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.840167341442505e-06,
      "loss": 0.3261,
      "step": 2096
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.831709127718892e-06,
      "loss": 0.3323,
      "step": 2097
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.82325176128784e-06,
      "loss": 0.3417,
      "step": 2098
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.814795248282974e-06,
      "loss": 0.3314,
      "step": 2099
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.806339594837309e-06,
      "loss": 0.3137,
      "step": 2100
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.79788480708323e-06,
      "loss": 0.3236,
      "step": 2101
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.789430891152501e-06,
      "loss": 0.3395,
      "step": 2102
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.78097785317624e-06,
      "loss": 0.2936,
      "step": 2103
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.772525699284947e-06,
      "loss": 0.3536,
      "step": 2104
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.764074435608469e-06,
      "loss": 0.3534,
      "step": 2105
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.755624068276002e-06,
      "loss": 0.3109,
      "step": 2106
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.747174603416113e-06,
      "loss": 0.3154,
      "step": 2107
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.738726047156693e-06,
      "loss": 0.3054,
      "step": 2108
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.73027840562498e-06,
      "loss": 0.3433,
      "step": 2109
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.721831684947557e-06,
      "loss": 0.3195,
      "step": 2110
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.713385891250328e-06,
      "loss": 0.3261,
      "step": 2111
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.704941030658536e-06,
      "loss": 0.3005,
      "step": 2112
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.696497109296733e-06,
      "loss": 0.3185,
      "step": 2113
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.688054133288803e-06,
      "loss": 0.3201,
      "step": 2114
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.67961210875794e-06,
      "loss": 0.3318,
      "step": 2115
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.671171041826646e-06,
      "loss": 0.318,
      "step": 2116
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.662730938616724e-06,
      "loss": 0.3358,
      "step": 2117
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.654291805249295e-06,
      "loss": 0.3228,
      "step": 2118
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.645853647844757e-06,
      "loss": 0.3066,
      "step": 2119
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.637416472522807e-06,
      "loss": 0.3171,
      "step": 2120
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.62898028540244e-06,
      "loss": 0.3252,
      "step": 2121
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.620545092601916e-06,
      "loss": 0.3303,
      "step": 2122
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.612110900238792e-06,
      "loss": 0.3061,
      "step": 2123
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.603677714429888e-06,
      "loss": 0.3094,
      "step": 2124
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.59524554129129e-06,
      "loss": 0.2967,
      "step": 2125
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.58681438693837e-06,
      "loss": 0.3127,
      "step": 2126
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.578384257485736e-06,
      "loss": 0.3398,
      "step": 2127
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.56995515904727e-06,
      "loss": 0.3194,
      "step": 2128
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.561527097736098e-06,
      "loss": 0.323,
      "step": 2129
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.553100079664598e-06,
      "loss": 0.3256,
      "step": 2130
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.54467411094439e-06,
      "loss": 0.3058,
      "step": 2131
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.536249197686333e-06,
      "loss": 0.3131,
      "step": 2132
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.527825346000516e-06,
      "loss": 0.2919,
      "step": 2133
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.519402561996271e-06,
      "loss": 0.3076,
      "step": 2134
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.510980851782143e-06,
      "loss": 0.2842,
      "step": 2135
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.502560221465898e-06,
      "loss": 0.314,
      "step": 2136
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.494140677154532e-06,
      "loss": 0.3097,
      "step": 2137
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.485722224954237e-06,
      "loss": 0.3409,
      "step": 2138
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.47730487097042e-06,
      "loss": 0.3152,
      "step": 2139
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.468888621307698e-06,
      "loss": 0.3498,
      "step": 2140
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.46047348206988e-06,
      "loss": 0.2986,
      "step": 2141
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.452059459359965e-06,
      "loss": 0.3112,
      "step": 2142
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.443646559280157e-06,
      "loss": 0.3049,
      "step": 2143
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.435234787931828e-06,
      "loss": 0.3095,
      "step": 2144
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.426824151415548e-06,
      "loss": 0.2957,
      "step": 2145
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.418414655831054e-06,
      "loss": 0.2978,
      "step": 2146
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.410006307277252e-06,
      "loss": 0.3342,
      "step": 2147
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.40159911185223e-06,
      "loss": 0.311,
      "step": 2148
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.39319307565323e-06,
      "loss": 0.3132,
      "step": 2149
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.384788204776648e-06,
      "loss": 0.3174,
      "step": 2150
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.376384505318052e-06,
      "loss": 0.3317,
      "step": 2151
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.367981983372143e-06,
      "loss": 0.3092,
      "step": 2152
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.359580645032776e-06,
      "loss": 0.2962,
      "step": 2153
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.351180496392951e-06,
      "loss": 0.3252,
      "step": 2154
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.342781543544799e-06,
      "loss": 0.3268,
      "step": 2155
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.334383792579582e-06,
      "loss": 0.3222,
      "step": 2156
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.325987249587701e-06,
      "loss": 0.3197,
      "step": 2157
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.317591920658667e-06,
      "loss": 0.3138,
      "step": 2158
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.309197811881128e-06,
      "loss": 0.3181,
      "step": 2159
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.300804929342826e-06,
      "loss": 0.3143,
      "step": 2160
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.292413279130625e-06,
      "loss": 0.3227,
      "step": 2161
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.284022867330502e-06,
      "loss": 0.3533,
      "step": 2162
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.275633700027526e-06,
      "loss": 0.3219,
      "step": 2163
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.267245783305858e-06,
      "loss": 0.3052,
      "step": 2164
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.258859123248772e-06,
      "loss": 0.3198,
      "step": 2165
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.250473725938608e-06,
      "loss": 0.3377,
      "step": 2166
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.24208959745681e-06,
      "loss": 0.3191,
      "step": 2167
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.233706743883887e-06,
      "loss": 0.3223,
      "step": 2168
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.22532517129943e-06,
      "loss": 0.3144,
      "step": 2169
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.216944885782099e-06,
      "loss": 0.351,
      "step": 2170
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.208565893409625e-06,
      "loss": 0.345,
      "step": 2171
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.200188200258793e-06,
      "loss": 0.3134,
      "step": 2172
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.191811812405453e-06,
      "loss": 0.321,
      "step": 2173
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.183436735924509e-06,
      "loss": 0.2894,
      "step": 2174
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.1750629768899e-06,
      "loss": 0.3111,
      "step": 2175
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.166690541374634e-06,
      "loss": 0.304,
      "step": 2176
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.158319435450734e-06,
      "loss": 0.3352,
      "step": 2177
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.149949665189279e-06,
      "loss": 0.3178,
      "step": 2178
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.141581236660365e-06,
      "loss": 0.3007,
      "step": 2179
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.133214155933118e-06,
      "loss": 0.3437,
      "step": 2180
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.124848429075697e-06,
      "loss": 0.3503,
      "step": 2181
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.116484062155263e-06,
      "loss": 0.3266,
      "step": 2182
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.108121061238004e-06,
      "loss": 0.3333,
      "step": 2183
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.099759432389108e-06,
      "loss": 0.3218,
      "step": 2184
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.091399181672776e-06,
      "loss": 0.3287,
      "step": 2185
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.083040315152201e-06,
      "loss": 0.3295,
      "step": 2186
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.074682838889581e-06,
      "loss": 0.3252,
      "step": 2187
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.066326758946097e-06,
      "loss": 0.3296,
      "step": 2188
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.057972081381926e-06,
      "loss": 0.3238,
      "step": 2189
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.049618812256223e-06,
      "loss": 0.3086,
      "step": 2190
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.041266957627116e-06,
      "loss": 0.317,
      "step": 2191
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.03291652355172e-06,
      "loss": 0.3182,
      "step": 2192
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.024567516086113e-06,
      "loss": 0.3315,
      "step": 2193
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.01621994128533e-06,
      "loss": 0.3038,
      "step": 2194
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.007873805203388e-06,
      "loss": 0.3389,
      "step": 2195
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.999529113893237e-06,
      "loss": 0.3026,
      "step": 2196
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.991185873406797e-06,
      "loss": 0.3005,
      "step": 2197
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.982844089794923e-06,
      "loss": 0.3198,
      "step": 2198
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.974503769107422e-06,
      "loss": 0.3125,
      "step": 2199
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.966164917393043e-06,
      "loss": 0.2784,
      "step": 2200
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.95782754069946e-06,
      "loss": 0.3176,
      "step": 2201
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.949491645073273e-06,
      "loss": 0.3367,
      "step": 2202
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.941157236560029e-06,
      "loss": 0.3348,
      "step": 2203
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.93282432120418e-06,
      "loss": 0.2868,
      "step": 2204
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.92449290504909e-06,
      "loss": 0.3091,
      "step": 2205
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.916162994137056e-06,
      "loss": 0.3272,
      "step": 2206
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.907834594509268e-06,
      "loss": 0.3182,
      "step": 2207
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.899507712205818e-06,
      "loss": 0.339,
      "step": 2208
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.89118235326571e-06,
      "loss": 0.3024,
      "step": 2209
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.882858523726835e-06,
      "loss": 0.3223,
      "step": 2210
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.874536229625974e-06,
      "loss": 0.3095,
      "step": 2211
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.866215476998802e-06,
      "loss": 0.3254,
      "step": 2212
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.857896271879862e-06,
      "loss": 0.3264,
      "step": 2213
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.849578620302595e-06,
      "loss": 0.329,
      "step": 2214
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.841262528299296e-06,
      "loss": 0.3279,
      "step": 2215
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.832948001901134e-06,
      "loss": 0.3338,
      "step": 2216
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.824635047138153e-06,
      "loss": 0.3018,
      "step": 2217
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.816323670039248e-06,
      "loss": 0.3654,
      "step": 2218
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.808013876632164e-06,
      "loss": 0.2965,
      "step": 2219
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.799705672943515e-06,
      "loss": 0.2624,
      "step": 2220
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.791399064998742e-06,
      "loss": 0.3095,
      "step": 2221
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.783094058822147e-06,
      "loss": 0.2908,
      "step": 2222
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.774790660436857e-06,
      "loss": 0.3424,
      "step": 2223
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.766488875864839e-06,
      "loss": 0.3209,
      "step": 2224
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.758188711126889e-06,
      "loss": 0.3008,
      "step": 2225
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.749890172242626e-06,
      "loss": 0.3388,
      "step": 2226
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.74159326523049e-06,
      "loss": 0.3077,
      "step": 2227
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.733297996107747e-06,
      "loss": 0.3264,
      "step": 2228
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.72500437089046e-06,
      "loss": 0.3156,
      "step": 2229
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.716712395593504e-06,
      "loss": 0.3329,
      "step": 2230
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.708422076230569e-06,
      "loss": 0.3045,
      "step": 2231
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.700133418814127e-06,
      "loss": 0.3252,
      "step": 2232
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.691846429355462e-06,
      "loss": 0.3286,
      "step": 2233
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.683561113864634e-06,
      "loss": 0.3354,
      "step": 2234
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.67527747835049e-06,
      "loss": 0.338,
      "step": 2235
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.666995528820673e-06,
      "loss": 0.3334,
      "step": 2236
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.658715271281584e-06,
      "loss": 0.3213,
      "step": 2237
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.650436711738411e-06,
      "loss": 0.3109,
      "step": 2238
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.642159856195105e-06,
      "loss": 0.3116,
      "step": 2239
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.633884710654383e-06,
      "loss": 0.2963,
      "step": 2240
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.625611281117716e-06,
      "loss": 0.3284,
      "step": 2241
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.617339573585341e-06,
      "loss": 0.3239,
      "step": 2242
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.609069594056234e-06,
      "loss": 0.2964,
      "step": 2243
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.600801348528132e-06,
      "loss": 0.3305,
      "step": 2244
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.592534842997501e-06,
      "loss": 0.3117,
      "step": 2245
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.584270083459549e-06,
      "loss": 0.3066,
      "step": 2246
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.576007075908225e-06,
      "loss": 0.2933,
      "step": 2247
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.567745826336199e-06,
      "loss": 0.3253,
      "step": 2248
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.559486340734865e-06,
      "loss": 0.3278,
      "step": 2249
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.551228625094349e-06,
      "loss": 0.2782,
      "step": 2250
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.5429726854034794e-06,
      "loss": 0.3273,
      "step": 2251
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.534718527649805e-06,
      "loss": 0.3007,
      "step": 2252
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.5264661578195796e-06,
      "loss": 0.3332,
      "step": 2253
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.518215581897763e-06,
      "loss": 0.3247,
      "step": 2254
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.5099668058680096e-06,
      "loss": 0.3247,
      "step": 2255
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.5017198357126755e-06,
      "loss": 0.3186,
      "step": 2256
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.493474677412795e-06,
      "loss": 0.3029,
      "step": 2257
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.485231336948102e-06,
      "loss": 0.2892,
      "step": 2258
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.476989820297005e-06,
      "loss": 0.302,
      "step": 2259
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.468750133436585e-06,
      "loss": 0.3172,
      "step": 2260
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.460512282342609e-06,
      "loss": 0.2977,
      "step": 2261
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.452276272989504e-06,
      "loss": 0.3344,
      "step": 2262
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.444042111350355e-06,
      "loss": 0.2991,
      "step": 2263
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.435809803396923e-06,
      "loss": 0.298,
      "step": 2264
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.42757935509961e-06,
      "loss": 0.298,
      "step": 2265
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.4193507724274825e-06,
      "loss": 0.2951,
      "step": 2266
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.411124061348241e-06,
      "loss": 0.3163,
      "step": 2267
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.402899227828233e-06,
      "loss": 0.3024,
      "step": 2268
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.3946762778324554e-06,
      "loss": 0.3125,
      "step": 2269
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.386455217324522e-06,
      "loss": 0.3074,
      "step": 2270
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.37823605226668e-06,
      "loss": 0.3081,
      "step": 2271
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.370018788619815e-06,
      "loss": 0.297,
      "step": 2272
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.3618034323434195e-06,
      "loss": 0.332,
      "step": 2273
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.353589989395605e-06,
      "loss": 0.3206,
      "step": 2274
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.3453784657331055e-06,
      "loss": 0.3128,
      "step": 2275
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.337168867311246e-06,
      "loss": 0.3583,
      "step": 2276
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.3289612000839735e-06,
      "loss": 0.3285,
      "step": 2277
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.320755470003822e-06,
      "loss": 0.3016,
      "step": 2278
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.31255168302192e-06,
      "loss": 0.3266,
      "step": 2279
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.3043498450879965e-06,
      "loss": 0.3313,
      "step": 2280
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.296149962150363e-06,
      "loss": 0.3084,
      "step": 2281
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.287952040155903e-06,
      "loss": 0.2964,
      "step": 2282
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.279756085050098e-06,
      "loss": 0.3028,
      "step": 2283
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.271562102776983e-06,
      "loss": 0.3181,
      "step": 2284
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.263370099279173e-06,
      "loss": 0.3121,
      "step": 2285
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.2551800804978455e-06,
      "loss": 0.3217,
      "step": 2286
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.2469920523727375e-06,
      "loss": 0.3143,
      "step": 2287
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.238806020842149e-06,
      "loss": 0.3161,
      "step": 2288
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.230621991842923e-06,
      "loss": 0.3138,
      "step": 2289
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.2224399713104485e-06,
      "loss": 0.3064,
      "step": 2290
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.214259965178674e-06,
      "loss": 0.3225,
      "step": 2291
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.2060819793800665e-06,
      "loss": 0.3022,
      "step": 2292
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.1979060198456405e-06,
      "loss": 0.3109,
      "step": 2293
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.18973209250494e-06,
      "loss": 0.353,
      "step": 2294
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.181560203286032e-06,
      "loss": 0.2977,
      "step": 2295
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.173390358115503e-06,
      "loss": 0.3048,
      "step": 2296
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.1652225629184665e-06,
      "loss": 0.3135,
      "step": 2297
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.1570568236185376e-06,
      "loss": 0.2881,
      "step": 2298
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.148893146137852e-06,
      "loss": 0.3258,
      "step": 2299
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.140731536397041e-06,
      "loss": 0.3464,
      "step": 2300
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.1325720003152345e-06,
      "loss": 0.277,
      "step": 2301
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.124414543810073e-06,
      "loss": 0.3304,
      "step": 2302
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.116259172797675e-06,
      "loss": 0.333,
      "step": 2303
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.108105893192645e-06,
      "loss": 0.3084,
      "step": 2304
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.099954710908087e-06,
      "loss": 0.3027,
      "step": 2305
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.091805631855566e-06,
      "loss": 0.3098,
      "step": 2306
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.08365866194513e-06,
      "loss": 0.292,
      "step": 2307
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.075513807085299e-06,
      "loss": 0.3238,
      "step": 2308
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.067371073183054e-06,
      "loss": 0.2851,
      "step": 2309
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.059230466143842e-06,
      "loss": 0.3058,
      "step": 2310
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.051091991871567e-06,
      "loss": 0.286,
      "step": 2311
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.04295565626858e-06,
      "loss": 0.3024,
      "step": 2312
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.034821465235693e-06,
      "loss": 0.3112,
      "step": 2313
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.026689424672151e-06,
      "loss": 0.2921,
      "step": 2314
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.018559540475641e-06,
      "loss": 0.3192,
      "step": 2315
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.010431818542298e-06,
      "loss": 0.2842,
      "step": 2316
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.0023062647666755e-06,
      "loss": 0.3132,
      "step": 2317
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.994182885041754e-06,
      "loss": 0.3323,
      "step": 2318
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.986061685258951e-06,
      "loss": 0.312,
      "step": 2319
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.977942671308087e-06,
      "loss": 0.3053,
      "step": 2320
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.969825849077412e-06,
      "loss": 0.2862,
      "step": 2321
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.961711224453573e-06,
      "loss": 0.3326,
      "step": 2322
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.953598803321632e-06,
      "loss": 0.344,
      "step": 2323
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.9454885915650485e-06,
      "loss": 0.336,
      "step": 2324
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.937380595065686e-06,
      "loss": 0.3203,
      "step": 2325
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.929274819703789e-06,
      "loss": 0.3089,
      "step": 2326
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.921171271358007e-06,
      "loss": 0.3156,
      "step": 2327
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.9130699559053625e-06,
      "loss": 0.3022,
      "step": 2328
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.90497087922126e-06,
      "loss": 0.3011,
      "step": 2329
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.896874047179488e-06,
      "loss": 0.3462,
      "step": 2330
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.888779465652195e-06,
      "loss": 0.3206,
      "step": 2331
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.880687140509914e-06,
      "loss": 0.3223,
      "step": 2332
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.872597077621525e-06,
      "loss": 0.3275,
      "step": 2333
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.864509282854272e-06,
      "loss": 0.2927,
      "step": 2334
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.856423762073765e-06,
      "loss": 0.3143,
      "step": 2335
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.848340521143948e-06,
      "loss": 0.3418,
      "step": 2336
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.840259565927118e-06,
      "loss": 0.3057,
      "step": 2337
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.832180902283927e-06,
      "loss": 0.3378,
      "step": 2338
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.824104536073346e-06,
      "loss": 0.3353,
      "step": 2339
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.816030473152684e-06,
      "loss": 0.3228,
      "step": 2340
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.8079587193775935e-06,
      "loss": 0.2728,
      "step": 2341
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.799889280602031e-06,
      "loss": 0.3113,
      "step": 2342
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.791822162678295e-06,
      "loss": 0.2922,
      "step": 2343
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.783757371456987e-06,
      "loss": 0.3174,
      "step": 2344
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.775694912787021e-06,
      "loss": 0.3428,
      "step": 2345
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.7676347925156315e-06,
      "loss": 0.293,
      "step": 2346
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.759577016488343e-06,
      "loss": 0.3045,
      "step": 2347
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.751521590548986e-06,
      "loss": 0.3045,
      "step": 2348
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.743468520539689e-06,
      "loss": 0.3044,
      "step": 2349
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.735417812300871e-06,
      "loss": 0.3078,
      "step": 2350
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.727369471671228e-06,
      "loss": 0.325,
      "step": 2351
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.719323504487758e-06,
      "loss": 0.3446,
      "step": 2352
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.71127991658572e-06,
      "loss": 0.317,
      "step": 2353
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.7032387137986595e-06,
      "loss": 0.3214,
      "step": 2354
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.695199901958386e-06,
      "loss": 0.3067,
      "step": 2355
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.68716348689497e-06,
      "loss": 0.3391,
      "step": 2356
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.679129474436758e-06,
      "loss": 0.3038,
      "step": 2357
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.671097870410346e-06,
      "loss": 0.3199,
      "step": 2358
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.663068680640574e-06,
      "loss": 0.2899,
      "step": 2359
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.655041910950552e-06,
      "loss": 0.2961,
      "step": 2360
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.6470175671616155e-06,
      "loss": 0.3118,
      "step": 2361
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.638995655093351e-06,
      "loss": 0.3361,
      "step": 2362
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.630976180563579e-06,
      "loss": 0.3039,
      "step": 2363
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.622959149388353e-06,
      "loss": 0.3231,
      "step": 2364
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.614944567381951e-06,
      "loss": 0.3014,
      "step": 2365
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.606932440356881e-06,
      "loss": 0.2996,
      "step": 2366
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.59892277412386e-06,
      "loss": 0.3396,
      "step": 2367
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.590915574491837e-06,
      "loss": 0.2892,
      "step": 2368
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.582910847267957e-06,
      "loss": 0.3381,
      "step": 2369
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.574908598257572e-06,
      "loss": 0.3114,
      "step": 2370
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.566908833264251e-06,
      "loss": 0.2842,
      "step": 2371
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.558911558089751e-06,
      "loss": 0.2955,
      "step": 2372
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.550916778534015e-06,
      "loss": 0.3286,
      "step": 2373
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.542924500395199e-06,
      "loss": 0.2941,
      "step": 2374
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.53493472946962e-06,
      "loss": 0.3326,
      "step": 2375
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.526947471551799e-06,
      "loss": 0.3047,
      "step": 2376
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.518962732434415e-06,
      "loss": 0.3203,
      "step": 2377
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.510980517908334e-06,
      "loss": 0.2963,
      "step": 2378
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.503000833762583e-06,
      "loss": 0.2989,
      "step": 2379
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.495023685784361e-06,
      "loss": 0.3044,
      "step": 2380
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.487049079759017e-06,
      "loss": 0.3033,
      "step": 2381
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.47907702147007e-06,
      "loss": 0.2813,
      "step": 2382
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.471107516699183e-06,
      "loss": 0.3387,
      "step": 2383
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.463140571226163e-06,
      "loss": 0.2871,
      "step": 2384
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.455176190828972e-06,
      "loss": 0.305,
      "step": 2385
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.447214381283702e-06,
      "loss": 0.2863,
      "step": 2386
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.439255148364589e-06,
      "loss": 0.2625,
      "step": 2387
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.431298497843995e-06,
      "loss": 0.3073,
      "step": 2388
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.4233444354924035e-06,
      "loss": 0.2959,
      "step": 2389
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.415392967078438e-06,
      "loss": 0.2942,
      "step": 2390
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.4074440983688205e-06,
      "loss": 0.3015,
      "step": 2391
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.399497835128401e-06,
      "loss": 0.2907,
      "step": 2392
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.391554183120139e-06,
      "loss": 0.3036,
      "step": 2393
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.383613148105095e-06,
      "loss": 0.2903,
      "step": 2394
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.375674735842431e-06,
      "loss": 0.3109,
      "step": 2395
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.367738952089417e-06,
      "loss": 0.3036,
      "step": 2396
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.3598058026013995e-06,
      "loss": 0.2831,
      "step": 2397
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.3518752931318375e-06,
      "loss": 0.327,
      "step": 2398
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.343947429432253e-06,
      "loss": 0.3436,
      "step": 2399
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.336022217252258e-06,
      "loss": 0.3115,
      "step": 2400
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.328099662339549e-06,
      "loss": 0.3143,
      "step": 2401
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.320179770439885e-06,
      "loss": 0.3008,
      "step": 2402
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.312262547297092e-06,
      "loss": 0.3124,
      "step": 2403
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.304347998653074e-06,
      "loss": 0.3101,
      "step": 2404
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.2964361302477825e-06,
      "loss": 0.3279,
      "step": 2405
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.288526947819229e-06,
      "loss": 0.2986,
      "step": 2406
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.28062045710348e-06,
      "loss": 0.3065,
      "step": 2407
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.272716663834645e-06,
      "loss": 0.3127,
      "step": 2408
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.264815573744884e-06,
      "loss": 0.283,
      "step": 2409
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.2569171925643915e-06,
      "loss": 0.2856,
      "step": 2410
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.24902152602139e-06,
      "loss": 0.3111,
      "step": 2411
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.241128579842154e-06,
      "loss": 0.3321,
      "step": 2412
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.233238359750966e-06,
      "loss": 0.3059,
      "step": 2413
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.225350871470133e-06,
      "loss": 0.2897,
      "step": 2414
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.217466120719995e-06,
      "loss": 0.3032,
      "step": 2415
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.209584113218894e-06,
      "loss": 0.3154,
      "step": 2416
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.201704854683179e-06,
      "loss": 0.3189,
      "step": 2417
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.193828350827222e-06,
      "loss": 0.312,
      "step": 2418
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.185954607363382e-06,
      "loss": 0.2976,
      "step": 2419
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.178083630002023e-06,
      "loss": 0.3361,
      "step": 2420
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.170215424451505e-06,
      "loss": 0.2927,
      "step": 2421
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.162349996418164e-06,
      "loss": 0.3239,
      "step": 2422
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.154487351606342e-06,
      "loss": 0.3464,
      "step": 2423
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.146627495718348e-06,
      "loss": 0.3248,
      "step": 2424
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.1387704344544684e-06,
      "loss": 0.3215,
      "step": 2425
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.130916173512976e-06,
      "loss": 0.3363,
      "step": 2426
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.123064718590099e-06,
      "loss": 0.2902,
      "step": 2427
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.11521607538003e-06,
      "loss": 0.3064,
      "step": 2428
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.1073702495749365e-06,
      "loss": 0.3494,
      "step": 2429
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.099527246864927e-06,
      "loss": 0.3437,
      "step": 2430
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.091687072938075e-06,
      "loss": 0.3155,
      "step": 2431
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.083849733480394e-06,
      "loss": 0.3423,
      "step": 2432
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.076015234175843e-06,
      "loss": 0.3246,
      "step": 2433
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.068183580706326e-06,
      "loss": 0.3096,
      "step": 2434
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.0603547787516804e-06,
      "loss": 0.2969,
      "step": 2435
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.052528833989669e-06,
      "loss": 0.3032,
      "step": 2436
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.044705752095998e-06,
      "loss": 0.3067,
      "step": 2437
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.036885538744282e-06,
      "loss": 0.3239,
      "step": 2438
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.0290681996060605e-06,
      "loss": 0.3004,
      "step": 2439
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.021253740350793e-06,
      "loss": 0.3163,
      "step": 2440
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.013442166645841e-06,
      "loss": 0.3219,
      "step": 2441
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.005633484156486e-06,
      "loss": 0.3055,
      "step": 2442
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9978276985459024e-06,
      "loss": 0.3376,
      "step": 2443
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.990024815475162e-06,
      "loss": 0.3044,
      "step": 2444
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.982224840603244e-06,
      "loss": 0.3058,
      "step": 2445
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.974427779587004e-06,
      "loss": 0.3117,
      "step": 2446
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.966633638081197e-06,
      "loss": 0.2982,
      "step": 2447
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.95884242173845e-06,
      "loss": 0.3241,
      "step": 2448
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9510541362092774e-06,
      "loss": 0.3276,
      "step": 2449
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.943268787142059e-06,
      "loss": 0.3523,
      "step": 2450
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.935486380183057e-06,
      "loss": 0.3511,
      "step": 2451
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.927706920976384e-06,
      "loss": 0.3137,
      "step": 2452
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.919930415164033e-06,
      "loss": 0.3098,
      "step": 2453
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9121568683858385e-06,
      "loss": 0.2726,
      "step": 2454
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.9043862862794965e-06,
      "loss": 0.2762,
      "step": 2455
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.896618674480558e-06,
      "loss": 0.3414,
      "step": 2456
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.888854038622412e-06,
      "loss": 0.3057,
      "step": 2457
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.881092384336285e-06,
      "loss": 0.3286,
      "step": 2458
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.8733337172512575e-06,
      "loss": 0.3005,
      "step": 2459
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.865578042994227e-06,
      "loss": 0.3226,
      "step": 2460
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.857825367189931e-06,
      "loss": 0.3278,
      "step": 2461
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.850075695460926e-06,
      "loss": 0.2776,
      "step": 2462
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.842329033427594e-06,
      "loss": 0.3189,
      "step": 2463
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.8345853867081315e-06,
      "loss": 0.3052,
      "step": 2464
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.826844760918555e-06,
      "loss": 0.3147,
      "step": 2465
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.819107161672672e-06,
      "loss": 0.3031,
      "step": 2466
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.8113725945821245e-06,
      "loss": 0.2847,
      "step": 2467
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.803641065256323e-06,
      "loss": 0.2905,
      "step": 2468
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.795912579302498e-06,
      "loss": 0.2758,
      "step": 2469
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.788187142325663e-06,
      "loss": 0.3232,
      "step": 2470
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.780464759928623e-06,
      "loss": 0.3287,
      "step": 2471
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.772745437711969e-06,
      "loss": 0.3116,
      "step": 2472
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.7650291812740666e-06,
      "loss": 0.3084,
      "step": 2473
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.757315996211066e-06,
      "loss": 0.2842,
      "step": 2474
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.749605888116885e-06,
      "loss": 0.2817,
      "step": 2475
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.741898862583213e-06,
      "loss": 0.3056,
      "step": 2476
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.7341949251994926e-06,
      "loss": 0.3221,
      "step": 2477
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.726494081552949e-06,
      "loss": 0.2897,
      "step": 2478
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.7187963372285404e-06,
      "loss": 0.3046,
      "step": 2479
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.711101697808991e-06,
      "loss": 0.2929,
      "step": 2480
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.703410168874768e-06,
      "loss": 0.3175,
      "step": 2481
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.695721756004087e-06,
      "loss": 0.3136,
      "step": 2482
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.6880364647729e-06,
      "loss": 0.2881,
      "step": 2483
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.680354300754896e-06,
      "loss": 0.3124,
      "step": 2484
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.672675269521495e-06,
      "loss": 0.3236,
      "step": 2485
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.664999376641848e-06,
      "loss": 0.3204,
      "step": 2486
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.65732662768283e-06,
      "loss": 0.3055,
      "step": 2487
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.649657028209024e-06,
      "loss": 0.3257,
      "step": 2488
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.641990583782753e-06,
      "loss": 0.3067,
      "step": 2489
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.634327299964026e-06,
      "loss": 0.2897,
      "step": 2490
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.626667182310575e-06,
      "loss": 0.3251,
      "step": 2491
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.6190102363778345e-06,
      "loss": 0.3151,
      "step": 2492
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.6113564677189335e-06,
      "loss": 0.2919,
      "step": 2493
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.603705881884701e-06,
      "loss": 0.3205,
      "step": 2494
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.5960584844236565e-06,
      "loss": 0.2918,
      "step": 2495
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.588414280882006e-06,
      "loss": 0.3244,
      "step": 2496
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.580773276803641e-06,
      "loss": 0.3133,
      "step": 2497
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.573135477730136e-06,
      "loss": 0.3263,
      "step": 2498
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.565500889200724e-06,
      "loss": 0.2941,
      "step": 2499
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.557869516752339e-06,
      "loss": 0.2745,
      "step": 2500
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.550241365919553e-06,
      "loss": 0.3407,
      "step": 2501
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.542616442234618e-06,
      "loss": 0.2684,
      "step": 2502
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.534994751227445e-06,
      "loss": 0.2482,
      "step": 2503
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.527376298425598e-06,
      "loss": 0.3117,
      "step": 2504
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.519761089354283e-06,
      "loss": 0.3163,
      "step": 2505
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.512149129536377e-06,
      "loss": 0.2994,
      "step": 2506
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.504540424492371e-06,
      "loss": 0.3091,
      "step": 2507
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.496934979740425e-06,
      "loss": 0.3156,
      "step": 2508
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.48933280079631e-06,
      "loss": 0.2964,
      "step": 2509
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.481733893173442e-06,
      "loss": 0.3279,
      "step": 2510
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.47413826238286e-06,
      "loss": 0.318,
      "step": 2511
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.46654591393323e-06,
      "loss": 0.315,
      "step": 2512
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.458956853330833e-06,
      "loss": 0.2895,
      "step": 2513
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.451371086079569e-06,
      "loss": 0.2895,
      "step": 2514
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.443788617680952e-06,
      "loss": 0.3031,
      "step": 2515
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.436209453634087e-06,
      "loss": 0.2996,
      "step": 2516
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.428633599435711e-06,
      "loss": 0.3221,
      "step": 2517
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.421061060580131e-06,
      "loss": 0.2813,
      "step": 2518
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.413491842559278e-06,
      "loss": 0.3113,
      "step": 2519
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.405925950862646e-06,
      "loss": 0.3283,
      "step": 2520
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.398363390977338e-06,
      "loss": 0.3079,
      "step": 2521
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.390804168388032e-06,
      "loss": 0.358,
      "step": 2522
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.3832482885769855e-06,
      "loss": 0.3012,
      "step": 2523
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.375695757024035e-06,
      "loss": 0.327,
      "step": 2524
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.368146579206588e-06,
      "loss": 0.3309,
      "step": 2525
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.360600760599619e-06,
      "loss": 0.3538,
      "step": 2526
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.353058306675656e-06,
      "loss": 0.2871,
      "step": 2527
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.345519222904811e-06,
      "loss": 0.3197,
      "step": 2528
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.337983514754723e-06,
      "loss": 0.3613,
      "step": 2529
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.330451187690614e-06,
      "loss": 0.3183,
      "step": 2530
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.3229222471752225e-06,
      "loss": 0.288,
      "step": 2531
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.315396698668852e-06,
      "loss": 0.3157,
      "step": 2532
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.307874547629339e-06,
      "loss": 0.3038,
      "step": 2533
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.300355799512059e-06,
      "loss": 0.2984,
      "step": 2534
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.292840459769907e-06,
      "loss": 0.3069,
      "step": 2535
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.285328533853331e-06,
      "loss": 0.3091,
      "step": 2536
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.277820027210279e-06,
      "loss": 0.2336,
      "step": 2537
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.270314945286229e-06,
      "loss": 0.2264,
      "step": 2538
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.262813293524178e-06,
      "loss": 0.2182,
      "step": 2539
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.255315077364629e-06,
      "loss": 0.2251,
      "step": 2540
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.2478203022456e-06,
      "loss": 0.1947,
      "step": 2541
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.240328973602608e-06,
      "loss": 0.2349,
      "step": 2542
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.232841096868673e-06,
      "loss": 0.2239,
      "step": 2543
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.225356677474309e-06,
      "loss": 0.2304,
      "step": 2544
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.2178757208475315e-06,
      "loss": 0.2217,
      "step": 2545
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.210398232413824e-06,
      "loss": 0.2208,
      "step": 2546
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.202924217596186e-06,
      "loss": 0.218,
      "step": 2547
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.195453681815067e-06,
      "loss": 0.2235,
      "step": 2548
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.187986630488411e-06,
      "loss": 0.2246,
      "step": 2549
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1805230690316286e-06,
      "loss": 0.2186,
      "step": 2550
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1730630028576055e-06,
      "loss": 0.2058,
      "step": 2551
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1656064373766845e-06,
      "loss": 0.2081,
      "step": 2552
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1581533779966755e-06,
      "loss": 0.2066,
      "step": 2553
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.15070383012284e-06,
      "loss": 0.2096,
      "step": 2554
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1432577991579025e-06,
      "loss": 0.2191,
      "step": 2555
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.135815290502029e-06,
      "loss": 0.2015,
      "step": 2556
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.128376309552825e-06,
      "loss": 0.2138,
      "step": 2557
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.120940861705357e-06,
      "loss": 0.2357,
      "step": 2558
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.113508952352109e-06,
      "loss": 0.2138,
      "step": 2559
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.106080586883009e-06,
      "loss": 0.2154,
      "step": 2560
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0986557706854126e-06,
      "loss": 0.224,
      "step": 2561
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0912345091441025e-06,
      "loss": 0.1811,
      "step": 2562
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0838168076412844e-06,
      "loss": 0.1969,
      "step": 2563
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.076402671556578e-06,
      "loss": 0.2326,
      "step": 2564
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.068992106267021e-06,
      "loss": 0.2253,
      "step": 2565
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0615851171470585e-06,
      "loss": 0.1945,
      "step": 2566
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0541817095685474e-06,
      "loss": 0.2081,
      "step": 2567
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.046781888900732e-06,
      "loss": 0.2079,
      "step": 2568
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.039385660510283e-06,
      "loss": 0.2179,
      "step": 2569
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.0319930297612356e-06,
      "loss": 0.2004,
      "step": 2570
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.024604002015034e-06,
      "loss": 0.1902,
      "step": 2571
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.017218582630507e-06,
      "loss": 0.2093,
      "step": 2572
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.009836776963861e-06,
      "loss": 0.2017,
      "step": 2573
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.002458590368686e-06,
      "loss": 0.1939,
      "step": 2574
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.995084028195951e-06,
      "loss": 0.2155,
      "step": 2575
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.987713095793979e-06,
      "loss": 0.2094,
      "step": 2576
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.980345798508489e-06,
      "loss": 0.1851,
      "step": 2577
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.972982141682537e-06,
      "loss": 0.2156,
      "step": 2578
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.965622130656551e-06,
      "loss": 0.1989,
      "step": 2579
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.958265770768315e-06,
      "loss": 0.1996,
      "step": 2580
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.950913067352965e-06,
      "loss": 0.2151,
      "step": 2581
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.9435640257429815e-06,
      "loss": 0.2211,
      "step": 2582
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.936218651268192e-06,
      "loss": 0.2241,
      "step": 2583
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.9288769492557655e-06,
      "loss": 0.2193,
      "step": 2584
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.921538925030203e-06,
      "loss": 0.2009,
      "step": 2585
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.914204583913349e-06,
      "loss": 0.2146,
      "step": 2586
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.906873931224355e-06,
      "loss": 0.2167,
      "step": 2587
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.8995469722797275e-06,
      "loss": 0.1842,
      "step": 2588
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.892223712393267e-06,
      "loss": 0.2176,
      "step": 2589
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.8849041568761055e-06,
      "loss": 0.2137,
      "step": 2590
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.877588311036688e-06,
      "loss": 0.222,
      "step": 2591
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.870276180180762e-06,
      "loss": 0.2296,
      "step": 2592
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.862967769611389e-06,
      "loss": 0.2196,
      "step": 2593
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.85566308462893e-06,
      "loss": 0.2491,
      "step": 2594
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.848362130531039e-06,
      "loss": 0.1924,
      "step": 2595
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.84106491261267e-06,
      "loss": 0.2034,
      "step": 2596
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.83377143616607e-06,
      "loss": 0.2182,
      "step": 2597
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.826481706480754e-06,
      "loss": 0.2266,
      "step": 2598
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.81919572884355e-06,
      "loss": 0.2083,
      "step": 2599
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.8119135085385375e-06,
      "loss": 0.2139,
      "step": 2600
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.804635050847085e-06,
      "loss": 0.2169,
      "step": 2601
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.79736036104783e-06,
      "loss": 0.205,
      "step": 2602
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.79008944441668e-06,
      "loss": 0.2136,
      "step": 2603
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.782822306226791e-06,
      "loss": 0.2135,
      "step": 2604
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.7755589517486055e-06,
      "loss": 0.2126,
      "step": 2605
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.7682993862497915e-06,
      "loss": 0.2118,
      "step": 2606
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.7610436149953e-06,
      "loss": 0.2089,
      "step": 2607
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.753791643247303e-06,
      "loss": 0.1882,
      "step": 2608
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.746543476265232e-06,
      "loss": 0.2152,
      "step": 2609
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.739299119305755e-06,
      "loss": 0.2119,
      "step": 2610
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.732058577622779e-06,
      "loss": 0.2209,
      "step": 2611
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.724821856467441e-06,
      "loss": 0.1901,
      "step": 2612
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.717588961088109e-06,
      "loss": 0.2174,
      "step": 2613
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.710359896730379e-06,
      "loss": 0.232,
      "step": 2614
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.703134668637056e-06,
      "loss": 0.2047,
      "step": 2615
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.695913282048187e-06,
      "loss": 0.2027,
      "step": 2616
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.688695742201002e-06,
      "loss": 0.2168,
      "step": 2617
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.681482054329975e-06,
      "loss": 0.1989,
      "step": 2618
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.674272223666755e-06,
      "loss": 0.2182,
      "step": 2619
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.667066255440212e-06,
      "loss": 0.2367,
      "step": 2620
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.659864154876411e-06,
      "loss": 0.227,
      "step": 2621
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.6526659271986116e-06,
      "loss": 0.2365,
      "step": 2622
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.645471577627264e-06,
      "loss": 0.2089,
      "step": 2623
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.638281111380006e-06,
      "loss": 0.2047,
      "step": 2624
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.631094533671663e-06,
      "loss": 0.2011,
      "step": 2625
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.623911849714226e-06,
      "loss": 0.2075,
      "step": 2626
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.616733064716886e-06,
      "loss": 0.21,
      "step": 2627
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.609558183885979e-06,
      "loss": 0.2138,
      "step": 2628
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.602387212425039e-06,
      "loss": 0.2131,
      "step": 2629
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.595220155534738e-06,
      "loss": 0.1907,
      "step": 2630
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.588057018412922e-06,
      "loss": 0.1966,
      "step": 2631
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.580897806254594e-06,
      "loss": 0.2123,
      "step": 2632
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.57374252425191e-06,
      "loss": 0.1869,
      "step": 2633
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.566591177594172e-06,
      "loss": 0.2125,
      "step": 2634
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.559443771467833e-06,
      "loss": 0.2036,
      "step": 2635
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.5523003110564865e-06,
      "loss": 0.2,
      "step": 2636
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.5451608015408565e-06,
      "loss": 0.2135,
      "step": 2637
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.53802524809882e-06,
      "loss": 0.208,
      "step": 2638
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.530893655905361e-06,
      "loss": 0.2152,
      "step": 2639
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.523766030132616e-06,
      "loss": 0.2092,
      "step": 2640
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.516642375949823e-06,
      "loss": 0.2113,
      "step": 2641
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.509522698523352e-06,
      "loss": 0.2494,
      "step": 2642
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.502407003016685e-06,
      "loss": 0.2256,
      "step": 2643
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.495295294590422e-06,
      "loss": 0.187,
      "step": 2644
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4881875784022535e-06,
      "loss": 0.2037,
      "step": 2645
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.481083859607003e-06,
      "loss": 0.2133,
      "step": 2646
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4739841433565675e-06,
      "loss": 0.1934,
      "step": 2647
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.466888434799959e-06,
      "loss": 0.2086,
      "step": 2648
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4597967390832745e-06,
      "loss": 0.1984,
      "step": 2649
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.452709061349703e-06,
      "loss": 0.2267,
      "step": 2650
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.445625406739522e-06,
      "loss": 0.2223,
      "step": 2651
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.438545780390088e-06,
      "loss": 0.2154,
      "step": 2652
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.431470187435836e-06,
      "loss": 0.1863,
      "step": 2653
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.424398633008278e-06,
      "loss": 0.1892,
      "step": 2654
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.417331122235998e-06,
      "loss": 0.215,
      "step": 2655
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4102676602446375e-06,
      "loss": 0.2249,
      "step": 2656
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.403208252156921e-06,
      "loss": 0.1942,
      "step": 2657
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.396152903092611e-06,
      "loss": 0.2081,
      "step": 2658
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.389101618168541e-06,
      "loss": 0.2113,
      "step": 2659
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.382054402498594e-06,
      "loss": 0.2178,
      "step": 2660
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.375011261193697e-06,
      "loss": 0.2177,
      "step": 2661
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.367972199361827e-06,
      "loss": 0.1858,
      "step": 2662
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.360937222108002e-06,
      "loss": 0.1857,
      "step": 2663
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.353906334534278e-06,
      "loss": 0.2048,
      "step": 2664
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.34687954173974e-06,
      "loss": 0.2088,
      "step": 2665
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.339856848820511e-06,
      "loss": 0.2054,
      "step": 2666
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.33283826086973e-06,
      "loss": 0.2033,
      "step": 2667
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.325823782977575e-06,
      "loss": 0.2195,
      "step": 2668
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.318813420231225e-06,
      "loss": 0.2128,
      "step": 2669
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.3118071777148865e-06,
      "loss": 0.2077,
      "step": 2670
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.304805060509775e-06,
      "loss": 0.174,
      "step": 2671
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.297807073694112e-06,
      "loss": 0.2329,
      "step": 2672
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.290813222343127e-06,
      "loss": 0.2003,
      "step": 2673
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2838235115290515e-06,
      "loss": 0.2314,
      "step": 2674
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2768379463211005e-06,
      "loss": 0.2255,
      "step": 2675
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2698565317855065e-06,
      "loss": 0.1952,
      "step": 2676
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.262879272985468e-06,
      "loss": 0.2089,
      "step": 2677
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.255906174981185e-06,
      "loss": 0.2049,
      "step": 2678
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.248937242829832e-06,
      "loss": 0.207,
      "step": 2679
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.241972481585566e-06,
      "loss": 0.1828,
      "step": 2680
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.23501189629952e-06,
      "loss": 0.2048,
      "step": 2681
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.228055492019794e-06,
      "loss": 0.211,
      "step": 2682
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.2211032737914596e-06,
      "loss": 0.2,
      "step": 2683
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.21415524665655e-06,
      "loss": 0.2,
      "step": 2684
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.207211415654066e-06,
      "loss": 0.2056,
      "step": 2685
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.200271785819946e-06,
      "loss": 0.206,
      "step": 2686
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.19333636218711e-06,
      "loss": 0.2003,
      "step": 2687
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.186405149785403e-06,
      "loss": 0.2367,
      "step": 2688
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.179478153641626e-06,
      "loss": 0.2152,
      "step": 2689
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.172555378779524e-06,
      "loss": 0.1893,
      "step": 2690
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.165636830219776e-06,
      "loss": 0.2268,
      "step": 2691
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.158722512979998e-06,
      "loss": 0.1964,
      "step": 2692
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.151812432074738e-06,
      "loss": 0.2101,
      "step": 2693
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.144906592515469e-06,
      "loss": 0.2041,
      "step": 2694
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.138004999310591e-06,
      "loss": 0.2362,
      "step": 2695
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.131107657465426e-06,
      "loss": 0.2096,
      "step": 2696
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.1242145719822015e-06,
      "loss": 0.2147,
      "step": 2697
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.117325747860077e-06,
      "loss": 0.211,
      "step": 2698
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.110441190095102e-06,
      "loss": 0.1917,
      "step": 2699
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.103560903680244e-06,
      "loss": 0.1986,
      "step": 2700
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.0966848936053705e-06,
      "loss": 0.1804,
      "step": 2701
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.089813164857245e-06,
      "loss": 0.2216,
      "step": 2702
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.082945722419528e-06,
      "loss": 0.2151,
      "step": 2703
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.076082571272773e-06,
      "loss": 0.1982,
      "step": 2704
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.069223716394419e-06,
      "loss": 0.2176,
      "step": 2705
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.062369162758782e-06,
      "loss": 0.218,
      "step": 2706
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.05551891533708e-06,
      "loss": 0.2035,
      "step": 2707
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.0486729790973785e-06,
      "loss": 0.2205,
      "step": 2708
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.041831359004648e-06,
      "loss": 0.2026,
      "step": 2709
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.034994060020699e-06,
      "loss": 0.188,
      "step": 2710
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.028161087104227e-06,
      "loss": 0.2237,
      "step": 2711
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.021332445210785e-06,
      "loss": 0.2202,
      "step": 2712
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.014508139292787e-06,
      "loss": 0.2457,
      "step": 2713
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.0076881742994925e-06,
      "loss": 0.2301,
      "step": 2714
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.000872555177031e-06,
      "loss": 0.2074,
      "step": 2715
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.994061286868361e-06,
      "loss": 0.2175,
      "step": 2716
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.987254374313298e-06,
      "loss": 0.2105,
      "step": 2717
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.980451822448493e-06,
      "loss": 0.2045,
      "step": 2718
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.973653636207437e-06,
      "loss": 0.202,
      "step": 2719
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.966859820520456e-06,
      "loss": 0.2069,
      "step": 2720
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.960070380314701e-06,
      "loss": 0.1889,
      "step": 2721
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.953285320514157e-06,
      "loss": 0.2077,
      "step": 2722
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.946504646039624e-06,
      "loss": 0.2047,
      "step": 2723
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.939728361808732e-06,
      "loss": 0.2192,
      "step": 2724
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.932956472735909e-06,
      "loss": 0.2133,
      "step": 2725
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.9261889837324245e-06,
      "loss": 0.2197,
      "step": 2726
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.919425899706324e-06,
      "loss": 0.2183,
      "step": 2727
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.912667225562482e-06,
      "loss": 0.2164,
      "step": 2728
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.9059129662025645e-06,
      "loss": 0.2086,
      "step": 2729
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.899163126525038e-06,
      "loss": 0.2128,
      "step": 2730
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.892417711425164e-06,
      "loss": 0.2129,
      "step": 2731
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.8856767257949956e-06,
      "loss": 0.2236,
      "step": 2732
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.878940174523371e-06,
      "loss": 0.2347,
      "step": 2733
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8722080624959155e-06,
      "loss": 0.1998,
      "step": 2734
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.865480394595037e-06,
      "loss": 0.1901,
      "step": 2735
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.858757175699906e-06,
      "loss": 0.2107,
      "step": 2736
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8520384106864905e-06,
      "loss": 0.199,
      "step": 2737
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.845324104427505e-06,
      "loss": 0.2066,
      "step": 2738
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.838614261792444e-06,
      "loss": 0.2239,
      "step": 2739
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8319088876475595e-06,
      "loss": 0.1987,
      "step": 2740
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.825207986855868e-06,
      "loss": 0.2016,
      "step": 2741
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.818511564277134e-06,
      "loss": 0.197,
      "step": 2742
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8118196247678863e-06,
      "loss": 0.212,
      "step": 2743
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8051321731813793e-06,
      "loss": 0.2252,
      "step": 2744
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.7984492143676453e-06,
      "loss": 0.2101,
      "step": 2745
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7917707531734295e-06,
      "loss": 0.1916,
      "step": 2746
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.785096794442229e-06,
      "loss": 0.2187,
      "step": 2747
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7784273430142737e-06,
      "loss": 0.181,
      "step": 2748
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.771762403726523e-06,
      "loss": 0.2032,
      "step": 2749
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7651019814126656e-06,
      "loss": 0.187,
      "step": 2750
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.758446080903113e-06,
      "loss": 0.2047,
      "step": 2751
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7517947070249995e-06,
      "loss": 0.2116,
      "step": 2752
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.745147864602172e-06,
      "loss": 0.2053,
      "step": 2753
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7385055584552e-06,
      "loss": 0.2105,
      "step": 2754
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7318677934013447e-06,
      "loss": 0.2275,
      "step": 2755
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7252345742546005e-06,
      "loss": 0.2138,
      "step": 2756
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7186059058256407e-06,
      "loss": 0.2004,
      "step": 2757
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7119817929218503e-06,
      "loss": 0.2035,
      "step": 2758
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.70536224034731e-06,
      "loss": 0.225,
      "step": 2759
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6987472529027903e-06,
      "loss": 0.2121,
      "step": 2760
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6921368353857524e-06,
      "loss": 0.211,
      "step": 2761
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6855309925903416e-06,
      "loss": 0.1899,
      "step": 2762
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6789297293073876e-06,
      "loss": 0.2122,
      "step": 2763
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.672333050324398e-06,
      "loss": 0.1748,
      "step": 2764
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.66574096042556e-06,
      "loss": 0.2085,
      "step": 2765
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.659153464391715e-06,
      "loss": 0.214,
      "step": 2766
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.652570567000402e-06,
      "loss": 0.2148,
      "step": 2767
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.645992273025797e-06,
      "loss": 0.2128,
      "step": 2768
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.639418587238753e-06,
      "loss": 0.2246,
      "step": 2769
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6328495144067787e-06,
      "loss": 0.2159,
      "step": 2770
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6262850592940335e-06,
      "loss": 0.1903,
      "step": 2771
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6197252266613312e-06,
      "loss": 0.1907,
      "step": 2772
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6131700212661314e-06,
      "loss": 0.1946,
      "step": 2773
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6066194478625394e-06,
      "loss": 0.2284,
      "step": 2774
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6000735112012984e-06,
      "loss": 0.1975,
      "step": 2775
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.593532216029797e-06,
      "loss": 0.2381,
      "step": 2776
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.586995567092039e-06,
      "loss": 0.212,
      "step": 2777
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.580463569128685e-06,
      "loss": 0.2014,
      "step": 2778
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5739362268769985e-06,
      "loss": 0.2077,
      "step": 2779
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.567413545070879e-06,
      "loss": 0.1846,
      "step": 2780
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.560895528440844e-06,
      "loss": 0.1904,
      "step": 2781
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5543821817140313e-06,
      "loss": 0.1918,
      "step": 2782
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5478735096141746e-06,
      "loss": 0.2031,
      "step": 2783
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.541369516861648e-06,
      "loss": 0.2147,
      "step": 2784
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5348702081733976e-06,
      "loss": 0.198,
      "step": 2785
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5283755882630045e-06,
      "loss": 0.2212,
      "step": 2786
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5218856618406227e-06,
      "loss": 0.2118,
      "step": 2787
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5154004336130177e-06,
      "loss": 0.2069,
      "step": 2788
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5089199082835436e-06,
      "loss": 0.2131,
      "step": 2789
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.502444090552143e-06,
      "loss": 0.2355,
      "step": 2790
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.4959729851153466e-06,
      "loss": 0.212,
      "step": 2791
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.4895065966662643e-06,
      "loss": 0.225,
      "step": 2792
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.4830449298945902e-06,
      "loss": 0.2167,
      "step": 2793
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.476587989486583e-06,
      "loss": 0.2057,
      "step": 2794
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.4701357801250913e-06,
      "loss": 0.2069,
      "step": 2795
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.463688306489511e-06,
      "loss": 0.196,
      "step": 2796
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4572455732558287e-06,
      "loss": 0.2135,
      "step": 2797
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4508075850965685e-06,
      "loss": 0.193,
      "step": 2798
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4443743466808268e-06,
      "loss": 0.2168,
      "step": 2799
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4379458626742533e-06,
      "loss": 0.1809,
      "step": 2800
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4315221377390496e-06,
      "loss": 0.1957,
      "step": 2801
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4251031765339638e-06,
      "loss": 0.1967,
      "step": 2802
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.418688983714291e-06,
      "loss": 0.1926,
      "step": 2803
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4122795639318705e-06,
      "loss": 0.2194,
      "step": 2804
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4058749218350684e-06,
      "loss": 0.1926,
      "step": 2805
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.3994750620688043e-06,
      "loss": 0.2104,
      "step": 2806
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.39307998927451e-06,
      "loss": 0.1959,
      "step": 2807
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.3866897080901663e-06,
      "loss": 0.1848,
      "step": 2808
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.380304223150258e-06,
      "loss": 0.2077,
      "step": 2809
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.373923539085805e-06,
      "loss": 0.2121,
      "step": 2810
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3675476605243405e-06,
      "loss": 0.1776,
      "step": 2811
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.361176592089919e-06,
      "loss": 0.1901,
      "step": 2812
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3548103384030908e-06,
      "loss": 0.2144,
      "step": 2813
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.348448904080935e-06,
      "loss": 0.2254,
      "step": 2814
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3420922937370204e-06,
      "loss": 0.2067,
      "step": 2815
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3357405119814213e-06,
      "loss": 0.2126,
      "step": 2816
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.329393563420713e-06,
      "loss": 0.2077,
      "step": 2817
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3230514526579617e-06,
      "loss": 0.2173,
      "step": 2818
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.316714184292729e-06,
      "loss": 0.2213,
      "step": 2819
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.31038176292106e-06,
      "loss": 0.1757,
      "step": 2820
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3040541931354897e-06,
      "loss": 0.228,
      "step": 2821
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2977314795250293e-06,
      "loss": 0.2271,
      "step": 2822
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.291413626675175e-06,
      "loss": 0.215,
      "step": 2823
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.285100639167883e-06,
      "loss": 0.2014,
      "step": 2824
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.278792521581605e-06,
      "loss": 0.2023,
      "step": 2825
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2724892784912375e-06,
      "loss": 0.2087,
      "step": 2826
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2661909144681534e-06,
      "loss": 0.2017,
      "step": 2827
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.259897434080187e-06,
      "loss": 0.2196,
      "step": 2828
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.253608841891628e-06,
      "loss": 0.1945,
      "step": 2829
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2473251424632213e-06,
      "loss": 0.1851,
      "step": 2830
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2410463403521653e-06,
      "loss": 0.1904,
      "step": 2831
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.234772440112104e-06,
      "loss": 0.1908,
      "step": 2832
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2285034462931286e-06,
      "loss": 0.2275,
      "step": 2833
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2222393634417747e-06,
      "loss": 0.2026,
      "step": 2834
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.215980196101002e-06,
      "loss": 0.201,
      "step": 2835
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.2097259488102294e-06,
      "loss": 0.2017,
      "step": 2836
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.203476626105284e-06,
      "loss": 0.2058,
      "step": 2837
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1972322325184347e-06,
      "loss": 0.2187,
      "step": 2838
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.190992772578373e-06,
      "loss": 0.2211,
      "step": 2839
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1847582508102115e-06,
      "loss": 0.1961,
      "step": 2840
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.17852867173548e-06,
      "loss": 0.2128,
      "step": 2841
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1723040398721285e-06,
      "loss": 0.2059,
      "step": 2842
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1660843597345137e-06,
      "loss": 0.2047,
      "step": 2843
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1598696358334034e-06,
      "loss": 0.1754,
      "step": 2844
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1536598726759747e-06,
      "loss": 0.2132,
      "step": 2845
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.147455074765793e-06,
      "loss": 0.2166,
      "step": 2846
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1412552466028443e-06,
      "loss": 0.2036,
      "step": 2847
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.1350603926834922e-06,
      "loss": 0.2048,
      "step": 2848
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.1288705175005e-06,
      "loss": 0.2068,
      "step": 2849
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.1226856255430206e-06,
      "loss": 0.2133,
      "step": 2850
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.116505721296592e-06,
      "loss": 0.2207,
      "step": 2851
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.110330809243134e-06,
      "loss": 0.196,
      "step": 2852
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.104160893860951e-06,
      "loss": 0.205,
      "step": 2853
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.097995979624708e-06,
      "loss": 0.179,
      "step": 2854
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0918360710054696e-06,
      "loss": 0.1749,
      "step": 2855
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0856811724706427e-06,
      "loss": 0.2207,
      "step": 2856
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0795312884840166e-06,
      "loss": 0.1979,
      "step": 2857
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0733864235057397e-06,
      "loss": 0.2083,
      "step": 2858
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0672465819923215e-06,
      "loss": 0.2117,
      "step": 2859
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0611117683966263e-06,
      "loss": 0.2224,
      "step": 2860
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.054981987167873e-06,
      "loss": 0.1969,
      "step": 2861
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0488572427516327e-06,
      "loss": 0.2389,
      "step": 2862
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0427375395898195e-06,
      "loss": 0.1765,
      "step": 2863
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.036622882120699e-06,
      "loss": 0.2206,
      "step": 2864
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0305132747788613e-06,
      "loss": 0.1867,
      "step": 2865
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0244087219952565e-06,
      "loss": 0.1927,
      "step": 2866
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0183092281971503e-06,
      "loss": 0.2026,
      "step": 2867
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0122147978081473e-06,
      "loss": 0.2083,
      "step": 2868
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0061254352481806e-06,
      "loss": 0.2173,
      "step": 2869
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0000411449335042e-06,
      "loss": 0.205,
      "step": 2870
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.9939619312766966e-06,
      "loss": 0.2161,
      "step": 2871
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.987887798686654e-06,
      "loss": 0.2092,
      "step": 2872
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.981818751568586e-06,
      "loss": 0.2158,
      "step": 2873
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.975754794324015e-06,
      "loss": 0.2021,
      "step": 2874
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9696959313507757e-06,
      "loss": 0.2095,
      "step": 2875
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.963642167042995e-06,
      "loss": 0.1885,
      "step": 2876
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9575935057911244e-06,
      "loss": 0.1794,
      "step": 2877
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9515499519818913e-06,
      "loss": 0.2321,
      "step": 2878
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9455115099983335e-06,
      "loss": 0.2096,
      "step": 2879
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.939478184219777e-06,
      "loss": 0.2182,
      "step": 2880
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9334499790218406e-06,
      "loss": 0.2181,
      "step": 2881
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9274268987764176e-06,
      "loss": 0.218,
      "step": 2882
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.921408947851705e-06,
      "loss": 0.2065,
      "step": 2883
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.915396130612156e-06,
      "loss": 0.1951,
      "step": 2884
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9093884514185255e-06,
      "loss": 0.1871,
      "step": 2885
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.90338591462782e-06,
      "loss": 0.2133,
      "step": 2886
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8973885245933287e-06,
      "loss": 0.2187,
      "step": 2887
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.891396285664605e-06,
      "loss": 0.2164,
      "step": 2888
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.885409202187468e-06,
      "loss": 0.2107,
      "step": 2889
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.879427278503997e-06,
      "loss": 0.1829,
      "step": 2890
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.873450518952526e-06,
      "loss": 0.1897,
      "step": 2891
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.867478927867654e-06,
      "loss": 0.206,
      "step": 2892
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.861512509580212e-06,
      "loss": 0.2065,
      "step": 2893
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.855551268417305e-06,
      "loss": 0.1957,
      "step": 2894
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.849595208702257e-06,
      "loss": 0.2174,
      "step": 2895
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8436443347546605e-06,
      "loss": 0.1844,
      "step": 2896
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.837698650890326e-06,
      "loss": 0.2236,
      "step": 2897
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.8317581614213073e-06,
      "loss": 0.207,
      "step": 2898
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.8258228706558956e-06,
      "loss": 0.2217,
      "step": 2899
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.8198927828986043e-06,
      "loss": 0.2124,
      "step": 2900
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.813967902450179e-06,
      "loss": 0.2099,
      "step": 2901
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.808048233607585e-06,
      "loss": 0.2079,
      "step": 2902
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.8021337806640137e-06,
      "loss": 0.1929,
      "step": 2903
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7962245479088602e-06,
      "loss": 0.2123,
      "step": 2904
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7903205396277546e-06,
      "loss": 0.1993,
      "step": 2905
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7844217601025136e-06,
      "loss": 0.2081,
      "step": 2906
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.778528213611188e-06,
      "loss": 0.2143,
      "step": 2907
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7726399044280107e-06,
      "loss": 0.2182,
      "step": 2908
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7667568368234275e-06,
      "loss": 0.2196,
      "step": 2909
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7608790150640818e-06,
      "loss": 0.2122,
      "step": 2910
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.755006443412812e-06,
      "loss": 0.2089,
      "step": 2911
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7491391261286473e-06,
      "loss": 0.203,
      "step": 2912
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7432770674668075e-06,
      "loss": 0.2007,
      "step": 2913
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7374202716787014e-06,
      "loss": 0.198,
      "step": 2914
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7315687430119097e-06,
      "loss": 0.2004,
      "step": 2915
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.725722485710214e-06,
      "loss": 0.1772,
      "step": 2916
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7198815040135464e-06,
      "loss": 0.1995,
      "step": 2917
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7140458021580408e-06,
      "loss": 0.2102,
      "step": 2918
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7082153843759773e-06,
      "loss": 0.196,
      "step": 2919
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7023902548958193e-06,
      "loss": 0.2274,
      "step": 2920
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.6965704179421894e-06,
      "loss": 0.2229,
      "step": 2921
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.6907558777358756e-06,
      "loss": 0.2054,
      "step": 2922
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.6849466384938116e-06,
      "loss": 0.1803,
      "step": 2923
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6791427044291097e-06,
      "loss": 0.2113,
      "step": 2924
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6733440797510114e-06,
      "loss": 0.2263,
      "step": 2925
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6675507686649215e-06,
      "loss": 0.2482,
      "step": 2926
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6617627753723864e-06,
      "loss": 0.2271,
      "step": 2927
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6559801040710986e-06,
      "loss": 0.1945,
      "step": 2928
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.650202758954886e-06,
      "loss": 0.2206,
      "step": 2929
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6444307442137197e-06,
      "loss": 0.2267,
      "step": 2930
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6386640640337004e-06,
      "loss": 0.1993,
      "step": 2931
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6329027225970627e-06,
      "loss": 0.2079,
      "step": 2932
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6271467240821693e-06,
      "loss": 0.1878,
      "step": 2933
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.621396072663499e-06,
      "loss": 0.2094,
      "step": 2934
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6156507725116743e-06,
      "loss": 0.2198,
      "step": 2935
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.6099108277934105e-06,
      "loss": 0.2003,
      "step": 2936
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.6041762426715565e-06,
      "loss": 0.1964,
      "step": 2937
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5984470213050693e-06,
      "loss": 0.2166,
      "step": 2938
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5927231678490163e-06,
      "loss": 0.2112,
      "step": 2939
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5870046864545685e-06,
      "loss": 0.212,
      "step": 2940
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5812915812690067e-06,
      "loss": 0.2137,
      "step": 2941
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5755838564357085e-06,
      "loss": 0.2004,
      "step": 2942
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5698815160941494e-06,
      "loss": 0.1976,
      "step": 2943
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5641845643799047e-06,
      "loss": 0.1895,
      "step": 2944
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5584930054246304e-06,
      "loss": 0.1959,
      "step": 2945
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.552806843356089e-06,
      "loss": 0.2084,
      "step": 2946
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5471260822981112e-06,
      "loss": 0.226,
      "step": 2947
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.54145072637062e-06,
      "loss": 0.2016,
      "step": 2948
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5357807796896184e-06,
      "loss": 0.1868,
      "step": 2949
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5301162463671845e-06,
      "loss": 0.2128,
      "step": 2950
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5244571305114697e-06,
      "loss": 0.1998,
      "step": 2951
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5188034362267013e-06,
      "loss": 0.2039,
      "step": 2952
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5131551676131617e-06,
      "loss": 0.1805,
      "step": 2953
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5075123287672174e-06,
      "loss": 0.2018,
      "step": 2954
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5018749237812823e-06,
      "loss": 0.2236,
      "step": 2955
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.496242956743833e-06,
      "loss": 0.1942,
      "step": 2956
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4906164317394067e-06,
      "loss": 0.2079,
      "step": 2957
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.484995352848588e-06,
      "loss": 0.1995,
      "step": 2958
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4793797241480157e-06,
      "loss": 0.2046,
      "step": 2959
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4737695497103742e-06,
      "loss": 0.1997,
      "step": 2960
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.468164833604394e-06,
      "loss": 0.2214,
      "step": 2961
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.462565579894842e-06,
      "loss": 0.2048,
      "step": 2962
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.456971792642534e-06,
      "loss": 0.2333,
      "step": 2963
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.451383475904304e-06,
      "loss": 0.2196,
      "step": 2964
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4458006337330397e-06,
      "loss": 0.2066,
      "step": 2965
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.440223270177641e-06,
      "loss": 0.225,
      "step": 2966
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4346513892830427e-06,
      "loss": 0.1927,
      "step": 2967
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.429084995090202e-06,
      "loss": 0.1854,
      "step": 2968
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4235240916360968e-06,
      "loss": 0.2085,
      "step": 2969
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.417968682953723e-06,
      "loss": 0.1993,
      "step": 2970
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4124187730720916e-06,
      "loss": 0.2059,
      "step": 2971
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4068743660162242e-06,
      "loss": 0.2037,
      "step": 2972
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.401335465807154e-06,
      "loss": 0.1992,
      "step": 2973
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.3958020764619217e-06,
      "loss": 0.2124,
      "step": 2974
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.39027420199356e-06,
      "loss": 0.2025,
      "step": 2975
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3847518464111206e-06,
      "loss": 0.2084,
      "step": 2976
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.379235013719636e-06,
      "loss": 0.2179,
      "step": 2977
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3737237079201437e-06,
      "loss": 0.2117,
      "step": 2978
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3682179330096665e-06,
      "loss": 0.2294,
      "step": 2979
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3627176929812212e-06,
      "loss": 0.201,
      "step": 2980
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3572229918238066e-06,
      "loss": 0.2134,
      "step": 2981
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.351733833522405e-06,
      "loss": 0.1908,
      "step": 2982
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3462502220579806e-06,
      "loss": 0.203,
      "step": 2983
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.340772161407475e-06,
      "loss": 0.1979,
      "step": 2984
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3352996555438036e-06,
      "loss": 0.2111,
      "step": 2985
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3298327084358452e-06,
      "loss": 0.197,
      "step": 2986
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3243713240484667e-06,
      "loss": 0.2054,
      "step": 2987
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3189155063424784e-06,
      "loss": 0.1877,
      "step": 2988
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3134652592746664e-06,
      "loss": 0.1773,
      "step": 2989
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3080205867977755e-06,
      "loss": 0.2008,
      "step": 2990
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.302581492860507e-06,
      "loss": 0.1911,
      "step": 2991
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.297147981407509e-06,
      "loss": 0.1828,
      "step": 2992
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.291720056379396e-06,
      "loss": 0.2123,
      "step": 2993
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2862977217127136e-06,
      "loss": 0.1981,
      "step": 2994
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.280880981339971e-06,
      "loss": 0.1985,
      "step": 2995
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2754698391896037e-06,
      "loss": 0.2039,
      "step": 2996
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.270064299185997e-06,
      "loss": 0.1963,
      "step": 2997
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2646643652494693e-06,
      "loss": 0.2011,
      "step": 2998
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2592700412962775e-06,
      "loss": 0.2006,
      "step": 2999
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2538813312386044e-06,
      "loss": 0.2136,
      "step": 3000
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2484982389845656e-06,
      "loss": 0.2131,
      "step": 3001
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.243120768438203e-06,
      "loss": 0.1936,
      "step": 3002
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.23774892349947e-06,
      "loss": 0.2143,
      "step": 3003
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2323827080642622e-06,
      "loss": 0.2012,
      "step": 3004
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2270221260243675e-06,
      "loss": 0.1823,
      "step": 3005
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2216671812675118e-06,
      "loss": 0.214,
      "step": 3006
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.216317877677313e-06,
      "loss": 0.2238,
      "step": 3007
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.210974219133307e-06,
      "loss": 0.2316,
      "step": 3008
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2056362095109375e-06,
      "loss": 0.2263,
      "step": 3009
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2003038526815468e-06,
      "loss": 0.1883,
      "step": 3010
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.1949771525123786e-06,
      "loss": 0.2041,
      "step": 3011
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.189656112866576e-06,
      "loss": 0.2119,
      "step": 3012
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.184340737603178e-06,
      "loss": 0.1937,
      "step": 3013
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1790310305771046e-06,
      "loss": 0.1677,
      "step": 3014
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.173726995639186e-06,
      "loss": 0.2063,
      "step": 3015
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.168428636636113e-06,
      "loss": 0.1909,
      "step": 3016
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.163135957410486e-06,
      "loss": 0.2081,
      "step": 3017
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1578489618007646e-06,
      "loss": 0.2164,
      "step": 3018
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1525676536412986e-06,
      "loss": 0.2287,
      "step": 3019
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1472920367623094e-06,
      "loss": 0.1999,
      "step": 3020
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1420221149898935e-06,
      "loss": 0.2079,
      "step": 3021
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1367578921460075e-06,
      "loss": 0.2166,
      "step": 3022
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1314993720484923e-06,
      "loss": 0.212,
      "step": 3023
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.126246558511036e-06,
      "loss": 0.2054,
      "step": 3024
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.1209994553431957e-06,
      "loss": 0.1789,
      "step": 3025
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.1157580663503865e-06,
      "loss": 0.2178,
      "step": 3026
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.1105223953338805e-06,
      "loss": 0.2255,
      "step": 3027
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.105292446090801e-06,
      "loss": 0.1961,
      "step": 3028
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.100068222414121e-06,
      "loss": 0.177,
      "step": 3029
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.094849728092664e-06,
      "loss": 0.2052,
      "step": 3030
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.089636966911095e-06,
      "loss": 0.2033,
      "step": 3031
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.084429942649927e-06,
      "loss": 0.212,
      "step": 3032
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.079228659085498e-06,
      "loss": 0.2297,
      "step": 3033
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0740331199900053e-06,
      "loss": 0.1894,
      "step": 3034
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.068843329131458e-06,
      "loss": 0.2229,
      "step": 3035
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0636592902737095e-06,
      "loss": 0.2005,
      "step": 3036
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0584810071764362e-06,
      "loss": 0.1935,
      "step": 3037
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0533084835951434e-06,
      "loss": 0.2264,
      "step": 3038
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0481417232811575e-06,
      "loss": 0.1963,
      "step": 3039
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.042980729981624e-06,
      "loss": 0.2255,
      "step": 3040
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0378255074395094e-06,
      "loss": 0.2193,
      "step": 3041
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.032676059393592e-06,
      "loss": 0.1948,
      "step": 3042
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.027532389578466e-06,
      "loss": 0.1994,
      "step": 3043
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0223945017245238e-06,
      "loss": 0.1834,
      "step": 3044
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0172623995579845e-06,
      "loss": 0.204,
      "step": 3045
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0121360868008498e-06,
      "loss": 0.216,
      "step": 3046
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.007015567170937e-06,
      "loss": 0.1973,
      "step": 3047
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.001900844381857e-06,
      "loss": 0.1946,
      "step": 3048
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.9967919221430154e-06,
      "loss": 0.1939,
      "step": 3049
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.991688804159614e-06,
      "loss": 0.1947,
      "step": 3050
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9865914941326424e-06,
      "loss": 0.1964,
      "step": 3051
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.981499995758882e-06,
      "loss": 0.205,
      "step": 3052
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9764143127308942e-06,
      "loss": 0.2117,
      "step": 3053
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.971334448737029e-06,
      "loss": 0.2217,
      "step": 3054
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9662604074614044e-06,
      "loss": 0.2038,
      "step": 3055
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.961192192583934e-06,
      "loss": 0.1915,
      "step": 3056
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.956129807780288e-06,
      "loss": 0.2287,
      "step": 3057
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9510732567219192e-06,
      "loss": 0.1904,
      "step": 3058
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9460225430760447e-06,
      "loss": 0.2394,
      "step": 3059
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9409776705056514e-06,
      "loss": 0.1987,
      "step": 3060
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.935938642669487e-06,
      "loss": 0.2153,
      "step": 3061
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9309054632220645e-06,
      "loss": 0.218,
      "step": 3062
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9258781358136447e-06,
      "loss": 0.2099,
      "step": 3063
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9208566640902616e-06,
      "loss": 0.2124,
      "step": 3064
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9158410516936877e-06,
      "loss": 0.2125,
      "step": 3065
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.910831302261451e-06,
      "loss": 0.1993,
      "step": 3066
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9058274194268289e-06,
      "loss": 0.1886,
      "step": 3067
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9008294068188427e-06,
      "loss": 0.1836,
      "step": 3068
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.895837268062256e-06,
      "loss": 0.2125,
      "step": 3069
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8908510067775753e-06,
      "loss": 0.204,
      "step": 3070
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.88587062658104e-06,
      "loss": 0.2083,
      "step": 3071
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8808961310846263e-06,
      "loss": 0.217,
      "step": 3072
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8759275238960473e-06,
      "loss": 0.2137,
      "step": 3073
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8709648086187315e-06,
      "loss": 0.2161,
      "step": 3074
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8660079888518557e-06,
      "loss": 0.212,
      "step": 3075
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8610570681903018e-06,
      "loss": 0.2079,
      "step": 3076
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.856112050224681e-06,
      "loss": 0.2124,
      "step": 3077
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.851172938541327e-06,
      "loss": 0.1981,
      "step": 3078
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.846239736722284e-06,
      "loss": 0.2067,
      "step": 3079
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8413124483453138e-06,
      "loss": 0.2146,
      "step": 3080
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8363910769838878e-06,
      "loss": 0.2256,
      "step": 3081
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8314756262071875e-06,
      "loss": 0.2171,
      "step": 3082
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8265660995801004e-06,
      "loss": 0.2229,
      "step": 3083
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8216625006632183e-06,
      "loss": 0.1817,
      "step": 3084
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.816764833012825e-06,
      "loss": 0.1959,
      "step": 3085
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.811873100180921e-06,
      "loss": 0.203,
      "step": 3086
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8069873057151855e-06,
      "loss": 0.1736,
      "step": 3087
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8021074531589977e-06,
      "loss": 0.2179,
      "step": 3088
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.797233546051429e-06,
      "loss": 0.2134,
      "step": 3089
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7923655879272395e-06,
      "loss": 0.2102,
      "step": 3090
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7875035823168641e-06,
      "loss": 0.1935,
      "step": 3091
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7826475327464399e-06,
      "loss": 0.2003,
      "step": 3092
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7777974427377631e-06,
      "loss": 0.192,
      "step": 3093
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.772953315808329e-06,
      "loss": 0.1886,
      "step": 3094
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7681151554712894e-06,
      "loss": 0.1973,
      "step": 3095
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7632829652354799e-06,
      "loss": 0.1913,
      "step": 3096
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7584567486054039e-06,
      "loss": 0.207,
      "step": 3097
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7536365090812313e-06,
      "loss": 0.1934,
      "step": 3098
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.748822250158797e-06,
      "loss": 0.2033,
      "step": 3099
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7440139753296014e-06,
      "loss": 0.2056,
      "step": 3100
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7392116880808053e-06,
      "loss": 0.2186,
      "step": 3101
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7344153918952157e-06,
      "loss": 0.1754,
      "step": 3102
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7296250902513156e-06,
      "loss": 0.2042,
      "step": 3103
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7248407866232175e-06,
      "loss": 0.2064,
      "step": 3104
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7200624844807047e-06,
      "loss": 0.207,
      "step": 3105
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7152901872891925e-06,
      "loss": 0.1998,
      "step": 3106
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7105238985097472e-06,
      "loss": 0.2244,
      "step": 3107
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7057636215990792e-06,
      "loss": 0.2199,
      "step": 3108
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7010093600095356e-06,
      "loss": 0.1901,
      "step": 3109
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6962611171891042e-06,
      "loss": 0.1896,
      "step": 3110
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6915188965814034e-06,
      "loss": 0.2098,
      "step": 3111
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.686782701625691e-06,
      "loss": 0.2095,
      "step": 3112
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6820525357568418e-06,
      "loss": 0.1774,
      "step": 3113
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.677328402405376e-06,
      "loss": 0.1888,
      "step": 3114
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6726103049974207e-06,
      "loss": 0.225,
      "step": 3115
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6678982469547433e-06,
      "loss": 0.2246,
      "step": 3116
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6631922316947147e-06,
      "loss": 0.2174,
      "step": 3117
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6584922626303325e-06,
      "loss": 0.1941,
      "step": 3118
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6537983431702076e-06,
      "loss": 0.2099,
      "step": 3119
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.649110476718564e-06,
      "loss": 0.1826,
      "step": 3120
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.644428666675234e-06,
      "loss": 0.2037,
      "step": 3121
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6397529164356606e-06,
      "loss": 0.201,
      "step": 3122
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6350832293908903e-06,
      "loss": 0.2024,
      "step": 3123
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.630419608927566e-06,
      "loss": 0.2076,
      "step": 3124
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6257620584279454e-06,
      "loss": 0.2109,
      "step": 3125
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.621110581269868e-06,
      "loss": 0.2098,
      "step": 3126
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.616465180826784e-06,
      "loss": 0.2072,
      "step": 3127
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.6118258604677229e-06,
      "loss": 0.2104,
      "step": 3128
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.607192623557311e-06,
      "loss": 0.2287,
      "step": 3129
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.6025654734557638e-06,
      "loss": 0.1942,
      "step": 3130
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5979444135188837e-06,
      "loss": 0.1905,
      "step": 3131
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5933294470980443e-06,
      "loss": 0.1965,
      "step": 3132
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5887205775402193e-06,
      "loss": 0.1912,
      "step": 3133
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5841178081879427e-06,
      "loss": 0.1865,
      "step": 3134
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5795211423793356e-06,
      "loss": 0.1671,
      "step": 3135
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5749305834480877e-06,
      "loss": 0.2217,
      "step": 3136
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5703461347234617e-06,
      "loss": 0.2267,
      "step": 3137
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5657677995302889e-06,
      "loss": 0.215,
      "step": 3138
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5611955811889645e-06,
      "loss": 0.1729,
      "step": 3139
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5566294830154494e-06,
      "loss": 0.2106,
      "step": 3140
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5520695083212678e-06,
      "loss": 0.1772,
      "step": 3141
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5475156604135e-06,
      "loss": 0.2026,
      "step": 3142
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5429679425947796e-06,
      "loss": 0.2111,
      "step": 3143
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5384263581633063e-06,
      "loss": 0.2084,
      "step": 3144
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5338909104128164e-06,
      "loss": 0.2116,
      "step": 3145
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5293616026326053e-06,
      "loss": 0.1891,
      "step": 3146
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5248384381075155e-06,
      "loss": 0.1888,
      "step": 3147
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5203214201179284e-06,
      "loss": 0.2168,
      "step": 3148
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5158105519397737e-06,
      "loss": 0.204,
      "step": 3149
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5113058368445177e-06,
      "loss": 0.1687,
      "step": 3150
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.506807278099164e-06,
      "loss": 0.1991,
      "step": 3151
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.5023148789662534e-06,
      "loss": 0.1917,
      "step": 3152
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4978286427038602e-06,
      "loss": 0.209,
      "step": 3153
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4933485725655806e-06,
      "loss": 0.2043,
      "step": 3154
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4888746718005554e-06,
      "loss": 0.2017,
      "step": 3155
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4844069436534326e-06,
      "loss": 0.205,
      "step": 3156
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4799453913643968e-06,
      "loss": 0.1819,
      "step": 3157
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4754900181691467e-06,
      "loss": 0.1978,
      "step": 3158
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4710408272989052e-06,
      "loss": 0.1988,
      "step": 3159
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4665978219804056e-06,
      "loss": 0.2109,
      "step": 3160
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4621610054359014e-06,
      "loss": 0.1946,
      "step": 3161
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4577303808831467e-06,
      "loss": 0.2086,
      "step": 3162
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.453305951535423e-06,
      "loss": 0.2097,
      "step": 3163
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4488877206015007e-06,
      "loss": 0.195,
      "step": 3164
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.444475691285665e-06,
      "loss": 0.1965,
      "step": 3165
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4400698667877e-06,
      "loss": 0.1979,
      "step": 3166
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.435670250302892e-06,
      "loss": 0.1939,
      "step": 3167
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4312768450220228e-06,
      "loss": 0.2175,
      "step": 3168
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4268896541313714e-06,
      "loss": 0.1985,
      "step": 3169
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4225086808127074e-06,
      "loss": 0.2049,
      "step": 3170
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4181339282432948e-06,
      "loss": 0.2082,
      "step": 3171
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4137653995958844e-06,
      "loss": 0.1965,
      "step": 3172
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4094030980387063e-06,
      "loss": 0.218,
      "step": 3173
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.405047026735491e-06,
      "loss": 0.2097,
      "step": 3174
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4006971888454324e-06,
      "loss": 0.2254,
      "step": 3175
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.3963535875232148e-06,
      "loss": 0.2125,
      "step": 3176
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3920162259189963e-06,
      "loss": 0.1689,
      "step": 3177
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3876851071784092e-06,
      "loss": 0.1961,
      "step": 3178
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.383360234442559e-06,
      "loss": 0.1934,
      "step": 3179
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.379041610848022e-06,
      "loss": 0.1929,
      "step": 3180
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3747292395268407e-06,
      "loss": 0.192,
      "step": 3181
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3704231236065247e-06,
      "loss": 0.2048,
      "step": 3182
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3661232662100477e-06,
      "loss": 0.1817,
      "step": 3183
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3618296704558364e-06,
      "loss": 0.1947,
      "step": 3184
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3575423394577924e-06,
      "loss": 0.1951,
      "step": 3185
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3532612763252572e-06,
      "loss": 0.2011,
      "step": 3186
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3489864841630363e-06,
      "loss": 0.1972,
      "step": 3187
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.344717966071385e-06,
      "loss": 0.1962,
      "step": 3188
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.340455725146007e-06,
      "loss": 0.2043,
      "step": 3189
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3361997644780554e-06,
      "loss": 0.2045,
      "step": 3190
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.331950087154128e-06,
      "loss": 0.2049,
      "step": 3191
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3277066962562647e-06,
      "loss": 0.2155,
      "step": 3192
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.323469594861947e-06,
      "loss": 0.1999,
      "step": 3193
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3192387860440981e-06,
      "loss": 0.2201,
      "step": 3194
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3150142728710669e-06,
      "loss": 0.2043,
      "step": 3195
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3107960584066526e-06,
      "loss": 0.2122,
      "step": 3196
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3065841457100725e-06,
      "loss": 0.1843,
      "step": 3197
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.30237853783598e-06,
      "loss": 0.2065,
      "step": 3198
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2981792378344548e-06,
      "loss": 0.2156,
      "step": 3199
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2939862487510047e-06,
      "loss": 0.1864,
      "step": 3200
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.28979957362655e-06,
      "loss": 0.1995,
      "step": 3201
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2856192154974488e-06,
      "loss": 0.2107,
      "step": 3202
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2814451773954606e-06,
      "loss": 0.1943,
      "step": 3203
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.277277462347778e-06,
      "loss": 0.2013,
      "step": 3204
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.273116073376992e-06,
      "loss": 0.2069,
      "step": 3205
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2689610135011144e-06,
      "loss": 0.1717,
      "step": 3206
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2648122857335675e-06,
      "loss": 0.2419,
      "step": 3207
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2606698930831784e-06,
      "loss": 0.2121,
      "step": 3208
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2565338385541792e-06,
      "loss": 0.2067,
      "step": 3209
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2524041251462082e-06,
      "loss": 0.186,
      "step": 3210
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2482807558543042e-06,
      "loss": 0.2004,
      "step": 3211
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2441637336688995e-06,
      "loss": 0.2161,
      "step": 3212
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2400530615758345e-06,
      "loss": 0.2156,
      "step": 3213
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2359487425563299e-06,
      "loss": 0.2072,
      "step": 3214
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2318507795870138e-06,
      "loss": 0.1956,
      "step": 3215
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2277591756398933e-06,
      "loss": 0.2043,
      "step": 3216
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.223673933682369e-06,
      "loss": 0.2041,
      "step": 3217
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2195950566772253e-06,
      "loss": 0.2179,
      "step": 3218
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2155225475826316e-06,
      "loss": 0.2079,
      "step": 3219
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2114564093521397e-06,
      "loss": 0.1938,
      "step": 3220
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2073966449346807e-06,
      "loss": 0.2269,
      "step": 3221
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2033432572745629e-06,
      "loss": 0.1697,
      "step": 3222
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1992962493114645e-06,
      "loss": 0.1613,
      "step": 3223
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1952556239804502e-06,
      "loss": 0.1987,
      "step": 3224
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.191221384211939e-06,
      "loss": 0.1853,
      "step": 3225
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1871935329317363e-06,
      "loss": 0.2191,
      "step": 3226
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1831720730609997e-06,
      "loss": 0.215,
      "step": 3227
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1791570075162573e-06,
      "loss": 0.204,
      "step": 3228
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1751483392093998e-06,
      "loss": 0.1997,
      "step": 3229
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.171146071047683e-06,
      "loss": 0.1999,
      "step": 3230
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.167150205933707e-06,
      "loss": 0.1869,
      "step": 3231
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.163160746765447e-06,
      "loss": 0.2061,
      "step": 3232
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1591776964362167e-06,
      "loss": 0.2263,
      "step": 3233
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1552010578346894e-06,
      "loss": 0.2185,
      "step": 3234
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1512308338448885e-06,
      "loss": 0.2084,
      "step": 3235
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1472670273461806e-06,
      "loss": 0.2148,
      "step": 3236
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1433096412132838e-06,
      "loss": 0.1848,
      "step": 3237
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.139358678316257e-06,
      "loss": 0.2081,
      "step": 3238
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.135414141520499e-06,
      "loss": 0.2213,
      "step": 3239
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1314760336867513e-06,
      "loss": 0.2079,
      "step": 3240
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.127544357671093e-06,
      "loss": 0.2033,
      "step": 3241
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1236191163249322e-06,
      "loss": 0.2029,
      "step": 3242
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1197003124950223e-06,
      "loss": 0.2026,
      "step": 3243
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1157879490234346e-06,
      "loss": 0.1772,
      "step": 3244
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1118820287475796e-06,
      "loss": 0.1952,
      "step": 3245
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1079825545001887e-06,
      "loss": 0.2027,
      "step": 3246
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1040895291093233e-06,
      "loss": 0.2189,
      "step": 3247
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.100202955398364e-06,
      "loss": 0.202,
      "step": 3248
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0963228361860146e-06,
      "loss": 0.2082,
      "step": 3249
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0924491742862963e-06,
      "loss": 0.2077,
      "step": 3250
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.088581972508549e-06,
      "loss": 0.1987,
      "step": 3251
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0847212336574263e-06,
      "loss": 0.2047,
      "step": 3252
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0808669605328914e-06,
      "loss": 0.2386,
      "step": 3253
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0770191559302268e-06,
      "loss": 0.1923,
      "step": 3254
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0731778226400147e-06,
      "loss": 0.2066,
      "step": 3255
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.069342963448149e-06,
      "loss": 0.1986,
      "step": 3256
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0655145811358281e-06,
      "loss": 0.2201,
      "step": 3257
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0616926784795511e-06,
      "loss": 0.2031,
      "step": 3258
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0578772582511187e-06,
      "loss": 0.1733,
      "step": 3259
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0540683232176307e-06,
      "loss": 0.1972,
      "step": 3260
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0502658761414842e-06,
      "loss": 0.2247,
      "step": 3261
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0464699197803697e-06,
      "loss": 0.2183,
      "step": 3262
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0426804568872728e-06,
      "loss": 0.1975,
      "step": 3263
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.038897490210461e-06,
      "loss": 0.1971,
      "step": 3264
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.035121022493506e-06,
      "loss": 0.2144,
      "step": 3265
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0313510564752517e-06,
      "loss": 0.1887,
      "step": 3266
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.027587594889833e-06,
      "loss": 0.1946,
      "step": 3267
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0238306404666686e-06,
      "loss": 0.1916,
      "step": 3268
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0200801959304551e-06,
      "loss": 0.2137,
      "step": 3269
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0163362640011697e-06,
      "loss": 0.2017,
      "step": 3270
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0125988473940673e-06,
      "loss": 0.2204,
      "step": 3271
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0088679488196695e-06,
      "loss": 0.2147,
      "step": 3272
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0051435709837864e-06,
      "loss": 0.2102,
      "step": 3273
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.001425716587484e-06,
      "loss": 0.2324,
      "step": 3274
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.977143883271035e-07,
      "loss": 0.2176,
      "step": 3275
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.940095888942546e-07,
      "loss": 0.2206,
      "step": 3276
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.903113209758098e-07,
      "loss": 0.1764,
      "step": 3277
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.86619587253903e-07,
      "loss": 0.2063,
      "step": 3278
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.829343904059342e-07,
      "loss": 0.2209,
      "step": 3279
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.79255733104557e-07,
      "loss": 0.2157,
      "step": 3280
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.75583618017686e-07,
      "loss": 0.2086,
      "step": 3281
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.719180478084922e-07,
      "loss": 0.2251,
      "step": 3282
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.682590251353919e-07,
      "loss": 0.1972,
      "step": 3283
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.646065526520653e-07,
      "loss": 0.1813,
      "step": 3284
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.609606330074317e-07,
      "loss": 0.2116,
      "step": 3285
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.573212688456635e-07,
      "loss": 0.1974,
      "step": 3286
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.53688462806175e-07,
      "loss": 0.2133,
      "step": 3287
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.500622175236285e-07,
      "loss": 0.1928,
      "step": 3288
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.46442535627926e-07,
      "loss": 0.234,
      "step": 3289
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.42829419744209e-07,
      "loss": 0.1912,
      "step": 3290
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.392228724928576e-07,
      "loss": 0.205,
      "step": 3291
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.356228964894876e-07,
      "loss": 0.2135,
      "step": 3292
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.320294943449537e-07,
      "loss": 0.1801,
      "step": 3293
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.284426686653303e-07,
      "loss": 0.1945,
      "step": 3294
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.248624220519409e-07,
      "loss": 0.1919,
      "step": 3295
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.212887571013196e-07,
      "loss": 0.204,
      "step": 3296
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.177216764052377e-07,
      "loss": 0.2189,
      "step": 3297
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.141611825506902e-07,
      "loss": 0.1873,
      "step": 3298
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.106072781198938e-07,
      "loss": 0.2103,
      "step": 3299
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.070599656902801e-07,
      "loss": 0.2132,
      "step": 3300
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.035192478345157e-07,
      "loss": 0.1966,
      "step": 3301
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.999851271204651e-07,
      "loss": 0.2032,
      "step": 3302
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.964576061112251e-07,
      "loss": 0.1907,
      "step": 3303
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.929366873650957e-07,
      "loss": 0.2239,
      "step": 3304
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.894223734355911e-07,
      "loss": 0.2038,
      "step": 3305
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.85914666871438e-07,
      "loss": 0.2186,
      "step": 3306
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.824135702165693e-07,
      "loss": 0.2044,
      "step": 3307
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.789190860101226e-07,
      "loss": 0.186,
      "step": 3308
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.754312167864432e-07,
      "loss": 0.1967,
      "step": 3309
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.719499650750773e-07,
      "loss": 0.2262,
      "step": 3310
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.68475333400769e-07,
      "loss": 0.184,
      "step": 3311
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.650073242834689e-07,
      "loss": 0.2055,
      "step": 3312
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.615459402383131e-07,
      "loss": 0.2264,
      "step": 3313
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.580911837756467e-07,
      "loss": 0.2077,
      "step": 3314
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.546430574009956e-07,
      "loss": 0.2044,
      "step": 3315
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.51201563615085e-07,
      "loss": 0.2078,
      "step": 3316
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.477667049138283e-07,
      "loss": 0.1935,
      "step": 3317
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.44338483788325e-07,
      "loss": 0.1818,
      "step": 3318
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.409169027248621e-07,
      "loss": 0.2076,
      "step": 3319
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.375019642049109e-07,
      "loss": 0.185,
      "step": 3320
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.340936707051273e-07,
      "loss": 0.2189,
      "step": 3321
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.306920246973404e-07,
      "loss": 0.1933,
      "step": 3322
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.272970286485704e-07,
      "loss": 0.2082,
      "step": 3323
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.239086850210021e-07,
      "loss": 0.2081,
      "step": 3324
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.205269962720075e-07,
      "loss": 0.2127,
      "step": 3325
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.171519648541204e-07,
      "loss": 0.1911,
      "step": 3326
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.137835932150562e-07,
      "loss": 0.2012,
      "step": 3327
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.10421883797694e-07,
      "loss": 0.2119,
      "step": 3328
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.070668390400848e-07,
      "loss": 0.2206,
      "step": 3329
    },
    {
      "epoch": 2.63,
      "learning_rate": 8.037184613754456e-07,
      "loss": 0.1902,
      "step": 3330
    },
    {
      "epoch": 2.63,
      "learning_rate": 8.003767532321572e-07,
      "loss": 0.212,
      "step": 3331
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.970417170337641e-07,
      "loss": 0.2029,
      "step": 3332
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.937133551989673e-07,
      "loss": 0.183,
      "step": 3333
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.903916701416392e-07,
      "loss": 0.19,
      "step": 3334
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.87076664270795e-07,
      "loss": 0.2026,
      "step": 3335
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.837683399906193e-07,
      "loss": 0.2076,
      "step": 3336
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.804666997004407e-07,
      "loss": 0.2051,
      "step": 3337
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.771717457947459e-07,
      "loss": 0.2022,
      "step": 3338
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.738834806631712e-07,
      "loss": 0.2009,
      "step": 3339
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.706019066905024e-07,
      "loss": 0.2195,
      "step": 3340
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.673270262566668e-07,
      "loss": 0.2191,
      "step": 3341
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.6405884173675e-07,
      "loss": 0.1926,
      "step": 3342
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.607973555009684e-07,
      "loss": 0.1967,
      "step": 3343
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.575425699146877e-07,
      "loss": 0.2063,
      "step": 3344
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.542944873384106e-07,
      "loss": 0.1908,
      "step": 3345
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.510531101277819e-07,
      "loss": 0.1911,
      "step": 3346
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.478184406335809e-07,
      "loss": 0.2159,
      "step": 3347
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.445904812017235e-07,
      "loss": 0.1854,
      "step": 3348
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.413692341732582e-07,
      "loss": 0.2158,
      "step": 3349
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.381547018843648e-07,
      "loss": 0.1858,
      "step": 3350
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.349468866663567e-07,
      "loss": 0.2113,
      "step": 3351
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.317457908456682e-07,
      "loss": 0.2032,
      "step": 3352
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.285514167438723e-07,
      "loss": 0.1979,
      "step": 3353
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.253637666776558e-07,
      "loss": 0.1856,
      "step": 3354
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.221828429588351e-07,
      "loss": 0.1908,
      "step": 3355
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.190086478943459e-07,
      "loss": 0.1854,
      "step": 3356
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.158411837862456e-07,
      "loss": 0.1962,
      "step": 3357
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.126804529317088e-07,
      "loss": 0.2009,
      "step": 3358
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.095264576230287e-07,
      "loss": 0.1993,
      "step": 3359
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.063792001476122e-07,
      "loss": 0.1935,
      "step": 3360
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.032386827879778e-07,
      "loss": 0.1923,
      "step": 3361
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.001049078217614e-07,
      "loss": 0.1982,
      "step": 3362
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.969778775217007e-07,
      "loss": 0.1944,
      "step": 3363
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.938575941556514e-07,
      "loss": 0.184,
      "step": 3364
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.907440599865667e-07,
      "loss": 0.208,
      "step": 3365
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.876372772725126e-07,
      "loss": 0.192,
      "step": 3366
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.845372482666545e-07,
      "loss": 0.1961,
      "step": 3367
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.814439752172608e-07,
      "loss": 0.1967,
      "step": 3368
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.783574603677012e-07,
      "loss": 0.208,
      "step": 3369
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.752777059564431e-07,
      "loss": 0.2001,
      "step": 3370
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.722047142170484e-07,
      "loss": 0.1844,
      "step": 3371
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.691384873781814e-07,
      "loss": 0.2372,
      "step": 3372
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.66079027663592e-07,
      "loss": 0.1788,
      "step": 3373
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.630263372921276e-07,
      "loss": 0.1965,
      "step": 3374
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.599804184777237e-07,
      "loss": 0.1977,
      "step": 3375
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.569412734294078e-07,
      "loss": 0.1983,
      "step": 3376
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.539089043512914e-07,
      "loss": 0.2043,
      "step": 3377
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.508833134425741e-07,
      "loss": 0.2214,
      "step": 3378
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.478645028975372e-07,
      "loss": 0.2379,
      "step": 3379
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.448524749055485e-07,
      "loss": 0.2095,
      "step": 3380
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.418472316510538e-07,
      "loss": 0.1765,
      "step": 3381
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.388487753135764e-07,
      "loss": 0.2022,
      "step": 3382
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.358571080677245e-07,
      "loss": 0.2128,
      "step": 3383
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.328722320831737e-07,
      "loss": 0.1983,
      "step": 3384
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.298941495246813e-07,
      "loss": 0.1661,
      "step": 3385
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.269228625520751e-07,
      "loss": 0.2016,
      "step": 3386
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.239583733202537e-07,
      "loss": 0.2028,
      "step": 3387
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.210006839791871e-07,
      "loss": 0.2037,
      "step": 3388
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.180497966739129e-07,
      "loss": 0.217,
      "step": 3389
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.15105713544536e-07,
      "loss": 0.2092,
      "step": 3390
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.121684367262271e-07,
      "loss": 0.1851,
      "step": 3391
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.0923796834922e-07,
      "loss": 0.1916,
      "step": 3392
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.063143105388081e-07,
      "loss": 0.2039,
      "step": 3393
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.033974654153529e-07,
      "loss": 0.2188,
      "step": 3394
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.004874350942669e-07,
      "loss": 0.1928,
      "step": 3395
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.975842216860239e-07,
      "loss": 0.2135,
      "step": 3396
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.946878272961553e-07,
      "loss": 0.2,
      "step": 3397
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.917982540252442e-07,
      "loss": 0.1985,
      "step": 3398
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.889155039689276e-07,
      "loss": 0.2005,
      "step": 3399
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.860395792178952e-07,
      "loss": 0.2056,
      "step": 3400
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.831704818578842e-07,
      "loss": 0.2029,
      "step": 3401
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.803082139696847e-07,
      "loss": 0.1862,
      "step": 3402
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.774527776291305e-07,
      "loss": 0.2037,
      "step": 3403
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.746041749070974e-07,
      "loss": 0.2096,
      "step": 3404
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.71762407869515e-07,
      "loss": 0.1937,
      "step": 3405
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.689274785773457e-07,
      "loss": 0.1951,
      "step": 3406
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.660993890865973e-07,
      "loss": 0.1941,
      "step": 3407
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.632781414483179e-07,
      "loss": 0.2163,
      "step": 3408
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.60463737708592e-07,
      "loss": 0.1879,
      "step": 3409
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.576561799085379e-07,
      "loss": 0.18,
      "step": 3410
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.548554700843167e-07,
      "loss": 0.193,
      "step": 3411
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.520616102671128e-07,
      "loss": 0.1854,
      "step": 3412
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.492746024831541e-07,
      "loss": 0.2104,
      "step": 3413
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.464944487536894e-07,
      "loss": 0.1934,
      "step": 3414
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.437211510950002e-07,
      "loss": 0.2117,
      "step": 3415
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.409547115183977e-07,
      "loss": 0.217,
      "step": 3416
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.38195132030217e-07,
      "loss": 0.2152,
      "step": 3417
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.354424146318193e-07,
      "loss": 0.1932,
      "step": 3418
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.326965613195867e-07,
      "loss": 0.2068,
      "step": 3419
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.299575740849272e-07,
      "loss": 0.1925,
      "step": 3420
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.27225454914263e-07,
      "loss": 0.1865,
      "step": 3421
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.245002057890447e-07,
      "loss": 0.2024,
      "step": 3422
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.21781828685729e-07,
      "loss": 0.2023,
      "step": 3423
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.190703255758e-07,
      "loss": 0.2194,
      "step": 3424
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.163656984257481e-07,
      "loss": 0.2109,
      "step": 3425
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.136679491970809e-07,
      "loss": 0.1678,
      "step": 3426
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.109770798463165e-07,
      "loss": 0.2169,
      "step": 3427
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.082930923249852e-07,
      "loss": 0.22,
      "step": 3428
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.056159885796242e-07,
      "loss": 0.207,
      "step": 3429
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.029457705517793e-07,
      "loss": 0.198,
      "step": 3430
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.002824401780038e-07,
      "loss": 0.2033,
      "step": 3431
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.976259993898503e-07,
      "loss": 0.2244,
      "step": 3432
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.949764501138832e-07,
      "loss": 0.1935,
      "step": 3433
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.923337942716611e-07,
      "loss": 0.1916,
      "step": 3434
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.896980337797508e-07,
      "loss": 0.2192,
      "step": 3435
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.870691705497099e-07,
      "loss": 0.1953,
      "step": 3436
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.844472064881001e-07,
      "loss": 0.1904,
      "step": 3437
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.818321434964779e-07,
      "loss": 0.1938,
      "step": 3438
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.792239834713941e-07,
      "loss": 0.2013,
      "step": 3439
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.766227283043912e-07,
      "loss": 0.1962,
      "step": 3440
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.7402837988201113e-07,
      "loss": 0.1664,
      "step": 3441
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.7144094008577666e-07,
      "loss": 0.237,
      "step": 3442
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.688604107922101e-07,
      "loss": 0.2017,
      "step": 3443
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.662867938728144e-07,
      "loss": 0.1932,
      "step": 3444
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.6372009119408313e-07,
      "loss": 0.198,
      "step": 3445
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.6116030461749504e-07,
      "loss": 0.1855,
      "step": 3446
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.5860743599951186e-07,
      "loss": 0.192,
      "step": 3447
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.560614871915792e-07,
      "loss": 0.1723,
      "step": 3448
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.5352246004012446e-07,
      "loss": 0.1693,
      "step": 3449
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.5099035638655477e-07,
      "loss": 0.2037,
      "step": 3450
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.484651780672522e-07,
      "loss": 0.1919,
      "step": 3451
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.459469269135852e-07,
      "loss": 0.2146,
      "step": 3452
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.4343560475188843e-07,
      "loss": 0.2141,
      "step": 3453
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.4093121340347824e-07,
      "loss": 0.2031,
      "step": 3454
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.3843375468464175e-07,
      "loss": 0.2001,
      "step": 3455
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.359432304066391e-07,
      "loss": 0.2083,
      "step": 3456
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.334596423756998e-07,
      "loss": 0.1949,
      "step": 3457
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.309829923930242e-07,
      "loss": 0.1945,
      "step": 3458
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.285132822547822e-07,
      "loss": 0.1987,
      "step": 3459
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.260505137521076e-07,
      "loss": 0.205,
      "step": 3460
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.235946886711018e-07,
      "loss": 0.1935,
      "step": 3461
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.2114580879282884e-07,
      "loss": 0.2194,
      "step": 3462
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.187038758933204e-07,
      "loss": 0.2072,
      "step": 3463
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.162688917435631e-07,
      "loss": 0.2305,
      "step": 3464
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.138408581095088e-07,
      "loss": 0.2004,
      "step": 3465
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.114197767520667e-07,
      "loss": 0.205,
      "step": 3466
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.0900564942710554e-07,
      "loss": 0.2134,
      "step": 3467
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.0659847788544926e-07,
      "loss": 0.2122,
      "step": 3468
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.0419826387287807e-07,
      "loss": 0.1998,
      "step": 3469
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.01805009130124e-07,
      "loss": 0.1903,
      "step": 3470
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.9941871539287634e-07,
      "loss": 0.1888,
      "step": 3471
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.97039384391773e-07,
      "loss": 0.1921,
      "step": 3472
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.946670178523992e-07,
      "loss": 0.2159,
      "step": 3473
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.923016174952976e-07,
      "loss": 0.1737,
      "step": 3474
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.899431850359503e-07,
      "loss": 0.2026,
      "step": 3475
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.8759172218479135e-07,
      "loss": 0.2075,
      "step": 3476
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.852472306471966e-07,
      "loss": 0.2026,
      "step": 3477
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.8290971212348924e-07,
      "loss": 0.2075,
      "step": 3478
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.8057916830893106e-07,
      "loss": 0.1909,
      "step": 3479
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.7825560089373324e-07,
      "loss": 0.1924,
      "step": 3480
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.759390115630357e-07,
      "loss": 0.2018,
      "step": 3481
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.736294019969311e-07,
      "loss": 0.2131,
      "step": 3482
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.7132677387043845e-07,
      "loss": 0.2101,
      "step": 3483
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.690311288535209e-07,
      "loss": 0.2099,
      "step": 3484
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.667424686110721e-07,
      "loss": 0.2029,
      "step": 3485
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.644607948029244e-07,
      "loss": 0.2047,
      "step": 3486
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.6218610908384187e-07,
      "loss": 0.1874,
      "step": 3487
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5991841310351825e-07,
      "loss": 0.2253,
      "step": 3488
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5765770850658244e-07,
      "loss": 0.2117,
      "step": 3489
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.554039969325862e-07,
      "loss": 0.2187,
      "step": 3490
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5315728001601877e-07,
      "loss": 0.202,
      "step": 3491
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.509175593862868e-07,
      "loss": 0.1775,
      "step": 3492
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.48684836667732e-07,
      "loss": 0.2071,
      "step": 3493
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.4645911347961357e-07,
      "loss": 0.1925,
      "step": 3494
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.442403914361192e-07,
      "loss": 0.1905,
      "step": 3495
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.420286721463562e-07,
      "loss": 0.1702,
      "step": 3496
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.398239572143558e-07,
      "loss": 0.2068,
      "step": 3497
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.3762624823906577e-07,
      "loss": 0.1939,
      "step": 3498
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.354355468143566e-07,
      "loss": 0.1922,
      "step": 3499
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.332518545290153e-07,
      "loss": 0.2031,
      "step": 3500
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.310751729667427e-07,
      "loss": 0.2053,
      "step": 3501
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.289055037061617e-07,
      "loss": 0.2158,
      "step": 3502
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2674284832080127e-07,
      "loss": 0.1832,
      "step": 3503
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.245872083791124e-07,
      "loss": 0.214,
      "step": 3504
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2243858544445007e-07,
      "loss": 0.1764,
      "step": 3505
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2029698107508554e-07,
      "loss": 0.209,
      "step": 3506
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.1816239682419757e-07,
      "loss": 0.2083,
      "step": 3507
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.160348342398778e-07,
      "loss": 0.201,
      "step": 3508
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.1391429486511524e-07,
      "loss": 0.1866,
      "step": 3509
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.118007802378198e-07,
      "loss": 0.2314,
      "step": 3510
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0969429189079413e-07,
      "loss": 0.1885,
      "step": 3511
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0759483135175184e-07,
      "loss": 0.2161,
      "step": 3512
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.055024001433071e-07,
      "loss": 0.21,
      "step": 3513
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.034169997829783e-07,
      "loss": 0.2242,
      "step": 3514
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0133863178318237e-07,
      "loss": 0.2033,
      "step": 3515
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.9926729765123783e-07,
      "loss": 0.1981,
      "step": 3516
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.972029988893621e-07,
      "loss": 0.2013,
      "step": 3517
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.9514573699466754e-07,
      "loss": 0.195,
      "step": 3518
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.930955134591673e-07,
      "loss": 0.1878,
      "step": 3519
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.9105232976976536e-07,
      "loss": 0.2313,
      "step": 3520
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.890161874082653e-07,
      "loss": 0.2041,
      "step": 3521
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.869870878513581e-07,
      "loss": 0.215,
      "step": 3522
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.8496503257063325e-07,
      "loss": 0.1983,
      "step": 3523
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.8295002303256546e-07,
      "loss": 0.1929,
      "step": 3524
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.809420606985236e-07,
      "loss": 0.1965,
      "step": 3525
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7894114702476604e-07,
      "loss": 0.1974,
      "step": 3526
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7694728346243535e-07,
      "loss": 0.1927,
      "step": 3527
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.749604714575649e-07,
      "loss": 0.1956,
      "step": 3528
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.729807124510708e-07,
      "loss": 0.213,
      "step": 3529
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7100800787875916e-07,
      "loss": 0.2095,
      "step": 3530
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.6904235917131094e-07,
      "loss": 0.2241,
      "step": 3531
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.6708376775430035e-07,
      "loss": 0.2195,
      "step": 3532
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.6513223504817553e-07,
      "loss": 0.2039,
      "step": 3533
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.631877624682688e-07,
      "loss": 0.2133,
      "step": 3534
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.6125035142479215e-07,
      "loss": 0.21,
      "step": 3535
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.593200033228349e-07,
      "loss": 0.1856,
      "step": 3536
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.573967195623661e-07,
      "loss": 0.2021,
      "step": 3537
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.554805015382289e-07,
      "loss": 0.184,
      "step": 3538
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.535713506401438e-07,
      "loss": 0.1793,
      "step": 3539
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.516692682527055e-07,
      "loss": 0.1994,
      "step": 3540
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.4977425575538395e-07,
      "loss": 0.2224,
      "step": 3541
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.478863145225163e-07,
      "loss": 0.2249,
      "step": 3542
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.4600544592331966e-07,
      "loss": 0.1778,
      "step": 3543
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.441316513218739e-07,
      "loss": 0.2056,
      "step": 3544
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.422649320771331e-07,
      "loss": 0.2255,
      "step": 3545
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.4040528954291766e-07,
      "loss": 0.1763,
      "step": 3546
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.385527250679187e-07,
      "loss": 0.2025,
      "step": 3547
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.3670723999569157e-07,
      "loss": 0.2049,
      "step": 3548
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.348688356646578e-07,
      "loss": 0.2125,
      "step": 3549
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.3303751340810088e-07,
      "loss": 0.1965,
      "step": 3550
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.3121327455417618e-07,
      "loss": 0.2149,
      "step": 3551
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.293961204258932e-07,
      "loss": 0.2007,
      "step": 3552
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2758605234112774e-07,
      "loss": 0.2052,
      "step": 3553
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2578307161261525e-07,
      "loss": 0.175,
      "step": 3554
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.23987179547952e-07,
      "loss": 0.1965,
      "step": 3555
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2219837744959284e-07,
      "loss": 0.1934,
      "step": 3556
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2041666661484884e-07,
      "loss": 0.2176,
      "step": 3557
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1864204833589197e-07,
      "loss": 0.2296,
      "step": 3558
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1687452389974829e-07,
      "loss": 0.2199,
      "step": 3559
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.15114094588299e-07,
      "loss": 0.2318,
      "step": 3560
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1336076167827736e-07,
      "loss": 0.221,
      "step": 3561
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1161452644127611e-07,
      "loss": 0.1915,
      "step": 3562
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0987539014373448e-07,
      "loss": 0.2201,
      "step": 3563
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0814335404694685e-07,
      "loss": 0.2026,
      "step": 3564
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0641841940705622e-07,
      "loss": 0.1816,
      "step": 3565
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0470058747505516e-07,
      "loss": 0.1921,
      "step": 3566
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0298985949678828e-07,
      "loss": 0.21,
      "step": 3567
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.012862367129431e-07,
      "loss": 0.1859,
      "step": 3568
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.9958972035906022e-07,
      "loss": 0.2054,
      "step": 3569
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.9790031166551983e-07,
      "loss": 0.2066,
      "step": 3570
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9621801185755408e-07,
      "loss": 0.2363,
      "step": 3571
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9454282215523145e-07,
      "loss": 0.1942,
      "step": 3572
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9287474377347238e-07,
      "loss": 0.1828,
      "step": 3573
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9121377792203244e-07,
      "loss": 0.2019,
      "step": 3574
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8955992580551584e-07,
      "loss": 0.2096,
      "step": 3575
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8791318862336094e-07,
      "loss": 0.2103,
      "step": 3576
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8627356756985238e-07,
      "loss": 0.1762,
      "step": 3577
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8464106383410673e-07,
      "loss": 0.1976,
      "step": 3578
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8301567860008695e-07,
      "loss": 0.2059,
      "step": 3579
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8139741304658566e-07,
      "loss": 0.1975,
      "step": 3580
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.7978626834723955e-07,
      "loss": 0.1865,
      "step": 3581
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.781822456705129e-07,
      "loss": 0.21,
      "step": 3582
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.765853461797107e-07,
      "loss": 0.2133,
      "step": 3583
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7499557103297094e-07,
      "loss": 0.2044,
      "step": 3584
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.734129213832625e-07,
      "loss": 0.1909,
      "step": 3585
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.718373983783883e-07,
      "loss": 0.1925,
      "step": 3586
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7026900316098217e-07,
      "loss": 0.2004,
      "step": 3587
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6870773686851082e-07,
      "loss": 0.2104,
      "step": 3588
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6715360063326525e-07,
      "loss": 0.1926,
      "step": 3589
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6560659558237157e-07,
      "loss": 0.189,
      "step": 3590
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6406672283778012e-07,
      "loss": 0.195,
      "step": 3591
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6253398351627093e-07,
      "loss": 0.1933,
      "step": 3592
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6100837872944919e-07,
      "loss": 0.2276,
      "step": 3593
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.5948990958374543e-07,
      "loss": 0.1897,
      "step": 3594
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.5797857718041652e-07,
      "loss": 0.205,
      "step": 3595
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5647438261554237e-07,
      "loss": 0.1977,
      "step": 3596
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.549773269800281e-07,
      "loss": 0.2157,
      "step": 3597
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5348741135959856e-07,
      "loss": 0.1977,
      "step": 3598
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.520046368348027e-07,
      "loss": 0.1792,
      "step": 3599
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5052900448100815e-07,
      "loss": 0.2236,
      "step": 3600
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.490605153684066e-07,
      "loss": 0.205,
      "step": 3601
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.475991705620028e-07,
      "loss": 0.2044,
      "step": 3602
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4614497112162895e-07,
      "loss": 0.2178,
      "step": 3603
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.44697918101927e-07,
      "loss": 0.1833,
      "step": 3604
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.432580125523597e-07,
      "loss": 0.209,
      "step": 3605
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4182525551720505e-07,
      "loss": 0.1913,
      "step": 3606
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4039964803555961e-07,
      "loss": 0.1877,
      "step": 3607
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.3898119114133192e-07,
      "loss": 0.1904,
      "step": 3608
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.375698858632435e-07,
      "loss": 0.2061,
      "step": 3609
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3616573322483228e-07,
      "loss": 0.1919,
      "step": 3610
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3476873424444703e-07,
      "loss": 0.2189,
      "step": 3611
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.333788899352495e-07,
      "loss": 0.1894,
      "step": 3612
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3199620130521117e-07,
      "loss": 0.173,
      "step": 3613
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3062066935711548e-07,
      "loss": 0.185,
      "step": 3614
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.292522950885533e-07,
      "loss": 0.182,
      "step": 3615
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2789107949192747e-07,
      "loss": 0.2125,
      "step": 3616
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2653702355444608e-07,
      "loss": 0.1825,
      "step": 3617
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2519012825812804e-07,
      "loss": 0.1925,
      "step": 3618
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.238503945797942e-07,
      "loss": 0.2059,
      "step": 3619
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2251782349107734e-07,
      "loss": 0.1995,
      "step": 3620
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.2119241595841213e-07,
      "loss": 0.1877,
      "step": 3621
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1987417294303748e-07,
      "loss": 0.195,
      "step": 3622
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1856309540099754e-07,
      "loss": 0.2098,
      "step": 3623
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1725918428314053e-07,
      "loss": 0.1964,
      "step": 3624
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1596244053511674e-07,
      "loss": 0.1871,
      "step": 3625
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1467286509737718e-07,
      "loss": 0.2034,
      "step": 3626
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1339045890517486e-07,
      "loss": 0.2024,
      "step": 3627
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1211522288856357e-07,
      "loss": 0.2158,
      "step": 3628
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1084715797239798e-07,
      "loss": 0.2281,
      "step": 3629
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0958626507632797e-07,
      "loss": 0.1962,
      "step": 3630
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0833254511480872e-07,
      "loss": 0.1998,
      "step": 3631
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0708599899708738e-07,
      "loss": 0.199,
      "step": 3632
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0584662762721076e-07,
      "loss": 0.1981,
      "step": 3633
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0461443190402098e-07,
      "loss": 0.2067,
      "step": 3634
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0338941272115766e-07,
      "loss": 0.2162,
      "step": 3635
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0217157096705676e-07,
      "loss": 0.2041,
      "step": 3636
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0096090752494403e-07,
      "loss": 0.2066,
      "step": 3637
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.975742327284487e-08,
      "loss": 0.2176,
      "step": 3638
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.856111908357446e-08,
      "loss": 0.1927,
      "step": 3639
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.737199582474321e-08,
      "loss": 0.1992,
      "step": 3640
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.619005435874907e-08,
      "loss": 0.1971,
      "step": 3641
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.501529554278744e-08,
      "loss": 0.1991,
      "step": 3642
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.384772022884015e-08,
      "loss": 0.2113,
      "step": 3643
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.268732926368207e-08,
      "loss": 0.1915,
      "step": 3644
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.153412348887559e-08,
      "loss": 0.1885,
      "step": 3645
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.038810374077277e-08,
      "loss": 0.2079,
      "step": 3646
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.924927085051438e-08,
      "loss": 0.1931,
      "step": 3647
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.811762564402971e-08,
      "loss": 0.1918,
      "step": 3648
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.699316894203225e-08,
      "loss": 0.2111,
      "step": 3649
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.587590156002635e-08,
      "loss": 0.1849,
      "step": 3650
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.476582430830049e-08,
      "loss": 0.1956,
      "step": 3651
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.366293799192627e-08,
      "loss": 0.1855,
      "step": 3652
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.256724341076272e-08,
      "loss": 0.2063,
      "step": 3653
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.147874135945422e-08,
      "loss": 0.209,
      "step": 3654
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.039743262742594e-08,
      "loss": 0.1906,
      "step": 3655
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.932331799888727e-08,
      "loss": 0.1972,
      "step": 3656
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.825639825282949e-08,
      "loss": 0.2117,
      "step": 3657
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.719667416302923e-08,
      "loss": 0.2201,
      "step": 3658
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.614414649804059e-08,
      "loss": 0.1792,
      "step": 3659
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.50988160211985e-08,
      "loss": 0.2022,
      "step": 3660
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.406068349062212e-08,
      "loss": 0.2054,
      "step": 3661
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.302974965920584e-08,
      "loss": 0.1937,
      "step": 3662
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.200601527462603e-08,
      "loss": 0.1856,
      "step": 3663
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.098948107933656e-08,
      "loss": 0.2178,
      "step": 3664
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.998014781057105e-08,
      "loss": 0.2038,
      "step": 3665
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.89780162003384e-08,
      "loss": 0.1857,
      "step": 3666
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.7983086975425e-08,
      "loss": 0.1941,
      "step": 3667
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.699536085739589e-08,
      "loss": 0.2003,
      "step": 3668
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.601483856258917e-08,
      "loss": 0.1971,
      "step": 3669
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.504152080212157e-08,
      "loss": 0.1914,
      "step": 3670
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.407540828188175e-08,
      "loss": 0.2085,
      "step": 3671
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.31165017025348e-08,
      "loss": 0.2022,
      "step": 3672
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.216480175952e-08,
      "loss": 0.1981,
      "step": 3673
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.122030914304744e-08,
      "loss": 0.1877,
      "step": 3674
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.028302453810253e-08,
      "loss": 0.2075,
      "step": 3675
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.935294862444374e-08,
      "loss": 0.192,
      "step": 3676
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.843008207659928e-08,
      "loss": 0.1975,
      "step": 3677
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.7514425563870436e-08,
      "loss": 0.194,
      "step": 3678
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.660597975032822e-08,
      "loss": 0.1763,
      "step": 3679
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.5704745294815624e-08,
      "loss": 0.1991,
      "step": 3680
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.481072285094535e-08,
      "loss": 0.1883,
      "step": 3681
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.392391306709766e-08,
      "loss": 0.2334,
      "step": 3682
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.304431658642473e-08,
      "loss": 0.1821,
      "step": 3683
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.2171934046846286e-08,
      "loss": 0.2237,
      "step": 3684
    },
    {
      "epoch": 2.91,
      "learning_rate": 5.1306766081048456e-08,
      "loss": 0.2184,
      "step": 3685
    },
    {
      "epoch": 2.91,
      "learning_rate": 5.044881331648821e-08,
      "loss": 0.2029,
      "step": 3686
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.9598076375387826e-08,
      "loss": 0.1794,
      "step": 3687
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.875455587473488e-08,
      "loss": 0.1869,
      "step": 3688
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.791825242628778e-08,
      "loss": 0.2211,
      "step": 3689
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.708916663656693e-08,
      "loss": 0.1739,
      "step": 3690
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.626729910685912e-08,
      "loss": 0.182,
      "step": 3691
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.545265043321645e-08,
      "loss": 0.2069,
      "step": 3692
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.464522120645742e-08,
      "loss": 0.2166,
      "step": 3693
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.384501201216029e-08,
      "loss": 0.1958,
      "step": 3694
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.305202343067083e-08,
      "loss": 0.2015,
      "step": 3695
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.226625603709677e-08,
      "loss": 0.2019,
      "step": 3696
    },
    {
      "epoch": 2.92,
      "learning_rate": 4.1487710401310057e-08,
      "loss": 0.1807,
      "step": 3697
    },
    {
      "epoch": 2.92,
      "learning_rate": 4.071638708794234e-08,
      "loss": 0.2069,
      "step": 3698
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.9952286656389506e-08,
      "loss": 0.1962,
      "step": 3699
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.919540966080937e-08,
      "loss": 0.2017,
      "step": 3700
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.844575665011951e-08,
      "loss": 0.2206,
      "step": 3701
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.7703328167999485e-08,
      "loss": 0.2157,
      "step": 3702
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.696812475288858e-08,
      "loss": 0.1996,
      "step": 3703
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.624014693798583e-08,
      "loss": 0.2001,
      "step": 3704
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.551939525125225e-08,
      "loss": 0.2021,
      "step": 3705
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.480587021540527e-08,
      "loss": 0.2055,
      "step": 3706
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.409957234792316e-08,
      "loss": 0.21,
      "step": 3707
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.340050216104285e-08,
      "loss": 0.1947,
      "step": 3708
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.2708660161757666e-08,
      "loss": 0.1998,
      "step": 3709
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.202404685181959e-08,
      "loss": 0.2163,
      "step": 3710
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.134666272774034e-08,
      "loss": 0.2063,
      "step": 3711
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.067650828078361e-08,
      "loss": 0.2123,
      "step": 3712
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.001358399697618e-08,
      "loss": 0.1697,
      "step": 3713
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.9357890357096797e-08,
      "loss": 0.2094,
      "step": 3714
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.8709427836679514e-08,
      "loss": 0.1863,
      "step": 3715
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.806819690601814e-08,
      "loss": 0.1813,
      "step": 3716
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.7434198030160676e-08,
      "loss": 0.2188,
      "step": 3717
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.680743166890598e-08,
      "loss": 0.1942,
      "step": 3718
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.6187898276813785e-08,
      "loss": 0.1994,
      "step": 3719
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.557559830319245e-08,
      "loss": 0.2072,
      "step": 3720
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.4970532192108987e-08,
      "loss": 0.2232,
      "step": 3721
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.4372700382382376e-08,
      "loss": 0.1883,
      "step": 3722
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.378210330758357e-08,
      "loss": 0.2079,
      "step": 3723
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.319874139603884e-08,
      "loss": 0.2039,
      "step": 3724
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.2622615070825305e-08,
      "loss": 0.2181,
      "step": 3725
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.2053724749774296e-08,
      "loss": 0.222,
      "step": 3726
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.1492070845468005e-08,
      "loss": 0.1936,
      "step": 3727
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.0937653765241706e-08,
      "loss": 0.1952,
      "step": 3728
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.039047391117932e-08,
      "loss": 0.2094,
      "step": 3729
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.9850531680121187e-08,
      "loss": 0.1762,
      "step": 3730
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.931782746365296e-08,
      "loss": 0.2005,
      "step": 3731
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.8792361648115597e-08,
      "loss": 0.1636,
      "step": 3732
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.8274134614597592e-08,
      "loss": 0.1931,
      "step": 3733
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.7763146738938307e-08,
      "loss": 0.1937,
      "step": 3734
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.7259398391726854e-08,
      "loss": 0.2017,
      "step": 3735
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.6762889938303216e-08,
      "loss": 0.2133,
      "step": 3736
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.6273621738754908e-08,
      "loss": 0.2077,
      "step": 3737
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.5791594147920308e-08,
      "loss": 0.187,
      "step": 3738
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.5316807515385335e-08,
      "loss": 0.2066,
      "step": 3739
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4849262185483437e-08,
      "loss": 0.2043,
      "step": 3740
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4388958497300043e-08,
      "loss": 0.2178,
      "step": 3741
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.3935896784663671e-08,
      "loss": 0.2101,
      "step": 3742
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.349007737615482e-08,
      "loss": 0.1953,
      "step": 3743
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.3051500595100409e-08,
      "loss": 0.2044,
      "step": 3744
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.2620166759573783e-08,
      "loss": 0.1918,
      "step": 3745
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.2196076182395821e-08,
      "loss": 0.1901,
      "step": 3746
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.1779229171134943e-08,
      "loss": 0.2174,
      "step": 3747
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.1369626028104874e-08,
      "loss": 0.1752,
      "step": 3748
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0967267050367991e-08,
      "loss": 0.185,
      "step": 3749
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0572152529729762e-08,
      "loss": 0.2117,
      "step": 3750
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0184282752745411e-08,
      "loss": 0.2085,
      "step": 3751
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.803658000712146e-09,
      "loss": 0.2062,
      "step": 3752
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.430278549675819e-09,
      "loss": 0.2351,
      "step": 3753
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.064144670426489e-09,
      "loss": 0.1942,
      "step": 3754
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.705256628499525e-09,
      "loss": 0.1921,
      "step": 3755
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.353614684175614e-09,
      "loss": 0.2071,
      "step": 3756
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.009219092478537e-09,
      "loss": 0.2065,
      "step": 3757
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.672070103179608e-09,
      "loss": 0.2015,
      "step": 3758
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.342167960792124e-09,
      "loss": 0.1735,
      "step": 3759
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.019512904575809e-09,
      "loss": 0.2254,
      "step": 3760
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.704105168532371e-09,
      "loss": 0.1923,
      "step": 3761
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.39594498140883e-09,
      "loss": 0.1914,
      "step": 3762
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.095032566695303e-09,
      "loss": 0.2134,
      "step": 3763
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.801368142626107e-09,
      "loss": 0.1942,
      "step": 3764
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.514951922178657e-09,
      "loss": 0.1978,
      "step": 3765
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.235784113074571e-09,
      "loss": 0.1991,
      "step": 3766
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.9638649177763394e-09,
      "loss": 0.188,
      "step": 3767
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.699194533491769e-09,
      "loss": 0.2104,
      "step": 3768
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.4417731521717576e-09,
      "loss": 0.2136,
      "step": 3769
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.1916009605058594e-09,
      "loss": 0.1859,
      "step": 3770
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.948678139932272e-09,
      "loss": 0.208,
      "step": 3771
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.713004866627845e-09,
      "loss": 0.1966,
      "step": 3772
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.4845813115114147e-09,
      "loss": 0.2372,
      "step": 3773
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.263407640246019e-09,
      "loss": 0.2077,
      "step": 3774
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.0494840132366808e-09,
      "loss": 0.1959,
      "step": 3775
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.842810585627076e-09,
      "loss": 0.2026,
      "step": 3776
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.643387507308415e-09,
      "loss": 0.1866,
      "step": 3777
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.4512149229083404e-09,
      "loss": 0.2171,
      "step": 3778
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.2662929717987003e-09,
      "loss": 0.2016,
      "step": 3779
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.0886217880922155e-09,
      "loss": 0.1799,
      "step": 3780
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.9182015006447008e-09,
      "loss": 0.2095,
      "step": 3781
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.7550322330506242e-09,
      "loss": 0.2023,
      "step": 3782
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.5991141036475478e-09,
      "loss": 0.222,
      "step": 3783
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.4504472255139069e-09,
      "loss": 0.1793,
      "step": 3784
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.3090317064690105e-09,
      "loss": 0.2037,
      "step": 3785
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.1748676490730415e-09,
      "loss": 0.2208,
      "step": 3786
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.0479551506259456e-09,
      "loss": 0.1962,
      "step": 3787
    },
    {
      "epoch": 2.99,
      "learning_rate": 9.28294303172983e-10,
      "loss": 0.2171,
      "step": 3788
    },
    {
      "epoch": 2.99,
      "learning_rate": 8.158851934936263e-10,
      "loss": 0.2065,
      "step": 3789
    },
    {
      "epoch": 2.99,
      "learning_rate": 7.107279031148828e-10,
      "loss": 0.2149,
      "step": 3790
    },
    {
      "epoch": 2.99,
      "learning_rate": 6.128225082979722e-10,
      "loss": 0.1849,
      "step": 3791
    },
    {
      "epoch": 2.99,
      "learning_rate": 5.221690800505386e-10,
      "loss": 0.1945,
      "step": 3792
    },
    {
      "epoch": 2.99,
      "learning_rate": 4.3876768411665884e-10,
      "loss": 0.1894,
      "step": 3793
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.626183809835038e-10,
      "loss": 0.208,
      "step": 3794
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.937212258757871e-10,
      "loss": 0.184,
      "step": 3795
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.3207626876242672e-10,
      "loss": 0.2065,
      "step": 3796
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.776835543509936e-10,
      "loss": 0.1959,
      "step": 3797
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.305431220877118e-10,
      "loss": 0.1794,
      "step": 3798
    },
    {
      "epoch": 3.0,
      "learning_rate": 9.065500616189937e-11,
      "loss": 0.184,
      "step": 3799
    },
    {
      "epoch": 3.0,
      "learning_rate": 5.801923550263766e-11,
      "loss": 0.2106,
      "step": 3800
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.2635833776550886e-11,
      "loss": 0.2186,
      "step": 3801
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.4504819395577686e-11,
      "loss": 0.1864,
      "step": 3802
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.626205505868896e-12,
      "loss": 0.1886,
      "step": 3803
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0,
      "loss": 0.2286,
      "step": 3804
    },
    {
      "epoch": 3.0,
      "step": 3804,
      "total_flos": 9.662968504994759e+18,
      "train_loss": 0.3583733306807925,
      "train_runtime": 137988.234,
      "train_samples_per_second": 3.529,
      "train_steps_per_second": 0.028
    }
  ],
  "max_steps": 3804,
  "num_train_epochs": 3,
  "total_flos": 9.662968504994759e+18,
  "trial_name": null,
  "trial_params": null
}