{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 4443,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.4925373134328358e-06,
      "loss": 1.3317,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.9850746268656716e-06,
      "loss": 1.2834,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.477611940298508e-06,
      "loss": 1.2884,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.970149253731343e-06,
      "loss": 1.2797,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.4626865671641785e-06,
      "loss": 1.2744,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.955223880597016e-06,
      "loss": 1.2664,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.0447761194029851e-05,
      "loss": 1.2561,
      "step": 7
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1940298507462686e-05,
      "loss": 1.2025,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3432835820895523e-05,
      "loss": 1.2404,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4925373134328357e-05,
      "loss": 1.2051,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.6417910447761194e-05,
      "loss": 1.1355,
      "step": 11
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.791044776119403e-05,
      "loss": 1.1003,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9402985074626868e-05,
      "loss": 1.0716,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.0895522388059702e-05,
      "loss": 1.0376,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.238805970149254e-05,
      "loss": 0.9945,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.3880597014925373e-05,
      "loss": 0.9669,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.537313432835821e-05,
      "loss": 0.959,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6865671641791047e-05,
      "loss": 0.9157,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.835820895522388e-05,
      "loss": 0.9042,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.9850746268656714e-05,
      "loss": 0.8601,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.1343283582089554e-05,
      "loss": 0.8522,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.283582089552239e-05,
      "loss": 0.8284,
      "step": 22
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.432835820895522e-05,
      "loss": 0.8485,
      "step": 23
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.582089552238806e-05,
      "loss": 0.837,
      "step": 24
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.73134328358209e-05,
      "loss": 0.8241,
      "step": 25
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.8805970149253736e-05,
      "loss": 0.8086,
      "step": 26
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.029850746268657e-05,
      "loss": 0.7947,
      "step": 27
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.1791044776119404e-05,
      "loss": 0.808,
      "step": 28
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.328358208955224e-05,
      "loss": 0.7924,
      "step": 29
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.477611940298508e-05,
      "loss": 0.795,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.626865671641791e-05,
      "loss": 0.7825,
      "step": 31
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.7761194029850745e-05,
      "loss": 0.7776,
      "step": 32
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9253731343283586e-05,
      "loss": 0.7623,
      "step": 33
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.074626865671642e-05,
      "loss": 0.7757,
      "step": 34
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.223880597014925e-05,
      "loss": 0.7685,
      "step": 35
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.373134328358209e-05,
      "loss": 0.754,
      "step": 36
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.5223880597014934e-05,
      "loss": 0.763,
      "step": 37
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.671641791044776e-05,
      "loss": 0.7489,
      "step": 38
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.82089552238806e-05,
      "loss": 0.7388,
      "step": 39
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.970149253731343e-05,
      "loss": 0.7374,
      "step": 40
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.119402985074628e-05,
      "loss": 0.7463,
      "step": 41
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.268656716417911e-05,
      "loss": 0.7431,
      "step": 42
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.417910447761194e-05,
      "loss": 0.7415,
      "step": 43
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.567164179104478e-05,
      "loss": 0.7343,
      "step": 44
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.716417910447762e-05,
      "loss": 0.7406,
      "step": 45
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.865671641791044e-05,
      "loss": 0.733,
      "step": 46
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.014925373134329e-05,
      "loss": 0.7043,
      "step": 47
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.164179104477612e-05,
      "loss": 0.7431,
      "step": 48
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.313432835820896e-05,
      "loss": 0.7085,
      "step": 49
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.46268656716418e-05,
      "loss": 0.7295,
      "step": 50
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.611940298507463e-05,
      "loss": 0.716,
      "step": 51
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.761194029850747e-05,
      "loss": 0.7171,
      "step": 52
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.910447761194029e-05,
      "loss": 0.713,
      "step": 53
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.059701492537314e-05,
      "loss": 0.707,
      "step": 54
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.208955223880597e-05,
      "loss": 0.7118,
      "step": 55
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.358208955223881e-05,
      "loss": 0.712,
      "step": 56
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.507462686567164e-05,
      "loss": 0.7042,
      "step": 57
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.656716417910447e-05,
      "loss": 0.6983,
      "step": 58
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.805970149253732e-05,
      "loss": 0.6967,
      "step": 59
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.955223880597016e-05,
      "loss": 0.716,
      "step": 60
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.104477611940299e-05,
      "loss": 0.6991,
      "step": 61
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.253731343283582e-05,
      "loss": 0.6937,
      "step": 62
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.402985074626867e-05,
      "loss": 0.6773,
      "step": 63
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.552238805970149e-05,
      "loss": 0.7036,
      "step": 64
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.701492537313434e-05,
      "loss": 0.6797,
      "step": 65
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.850746268656717e-05,
      "loss": 0.671,
      "step": 66
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001,
      "loss": 0.6788,
      "step": 67
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00010149253731343284,
      "loss": 0.7005,
      "step": 68
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00010298507462686569,
      "loss": 0.6892,
      "step": 69
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001044776119402985,
      "loss": 0.6898,
      "step": 70
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00010597014925373134,
      "loss": 0.6768,
      "step": 71
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00010746268656716419,
      "loss": 0.6985,
      "step": 72
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00010895522388059702,
      "loss": 0.6898,
      "step": 73
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00011044776119402987,
      "loss": 0.6902,
      "step": 74
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00011194029850746269,
      "loss": 0.6706,
      "step": 75
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00011343283582089552,
      "loss": 0.6781,
      "step": 76
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00011492537313432837,
      "loss": 0.6765,
      "step": 77
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001164179104477612,
      "loss": 0.669,
      "step": 78
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00011791044776119405,
      "loss": 0.674,
      "step": 79
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00011940298507462686,
      "loss": 0.6674,
      "step": 80
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001208955223880597,
      "loss": 0.6724,
      "step": 81
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00012238805970149255,
      "loss": 0.6723,
      "step": 82
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00012388059701492538,
      "loss": 0.6664,
      "step": 83
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00012537313432835822,
      "loss": 0.6893,
      "step": 84
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00012686567164179105,
      "loss": 0.6928,
      "step": 85
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00012835820895522389,
      "loss": 0.6756,
      "step": 86
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00012985074626865672,
      "loss": 0.6908,
      "step": 87
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00013134328358208955,
      "loss": 0.6719,
      "step": 88
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001328358208955224,
      "loss": 0.6757,
      "step": 89
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00013432835820895525,
      "loss": 0.6776,
      "step": 90
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00013582089552238805,
      "loss": 0.6807,
      "step": 91
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001373134328358209,
      "loss": 0.6649,
      "step": 92
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00013880597014925375,
      "loss": 0.6705,
      "step": 93
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00014029850746268658,
      "loss": 0.6742,
      "step": 94
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00014179104477611942,
      "loss": 0.6618,
      "step": 95
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00014328358208955225,
      "loss": 0.6564,
      "step": 96
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00014477611940298508,
      "loss": 0.6552,
      "step": 97
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00014626865671641792,
      "loss": 0.6553,
      "step": 98
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00014776119402985075,
      "loss": 0.6708,
      "step": 99
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001492537313432836,
      "loss": 0.648,
      "step": 100
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00015074626865671642,
      "loss": 0.6546,
      "step": 101
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00015223880597014925,
      "loss": 0.6698,
      "step": 102
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00015373134328358208,
      "loss": 0.6577,
      "step": 103
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00015522388059701495,
      "loss": 0.6673,
      "step": 104
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00015671641791044778,
      "loss": 0.6632,
      "step": 105
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00015820895522388059,
      "loss": 0.6555,
      "step": 106
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00015970149253731345,
      "loss": 0.6616,
      "step": 107
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00016119402985074628,
      "loss": 0.6562,
      "step": 108
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00016268656716417911,
      "loss": 0.6601,
      "step": 109
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00016417910447761195,
      "loss": 0.6406,
      "step": 110
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00016567164179104478,
      "loss": 0.6587,
      "step": 111
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00016716417910447761,
      "loss": 0.6593,
      "step": 112
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00016865671641791045,
      "loss": 0.6416,
      "step": 113
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00017014925373134328,
      "loss": 0.6498,
      "step": 114
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00017164179104477614,
      "loss": 0.6435,
      "step": 115
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00017313432835820895,
      "loss": 0.6458,
      "step": 116
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00017462686567164178,
      "loss": 0.6571,
      "step": 117
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00017611940298507464,
      "loss": 0.6424,
      "step": 118
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00017761194029850748,
      "loss": 0.6466,
      "step": 119
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001791044776119403,
      "loss": 0.6612,
      "step": 120
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018059701492537314,
      "loss": 0.6297,
      "step": 121
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018208955223880598,
      "loss": 0.6401,
      "step": 122
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001835820895522388,
      "loss": 0.6524,
      "step": 123
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018507462686567165,
      "loss": 0.6331,
      "step": 124
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018656716417910448,
      "loss": 0.6435,
      "step": 125
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018805970149253734,
      "loss": 0.6557,
      "step": 126
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018955223880597015,
      "loss": 0.653,
      "step": 127
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019104477611940298,
      "loss": 0.6376,
      "step": 128
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019253731343283584,
      "loss": 0.6326,
      "step": 129
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019402985074626867,
      "loss": 0.6464,
      "step": 130
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001955223880597015,
      "loss": 0.6347,
      "step": 131
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019701492537313434,
      "loss": 0.6405,
      "step": 132
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019850746268656718,
      "loss": 0.6351,
      "step": 133
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 0.6391,
      "step": 134
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019999997342233993,
      "loss": 0.6448,
      "step": 135
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019999989368937388,
      "loss": 0.6391,
      "step": 136
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019999976080114422,
      "loss": 0.6529,
      "step": 137
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001999995747577216,
      "loss": 0.6379,
      "step": 138
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019999933555920487,
      "loss": 0.6335,
      "step": 139
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019999904320572122,
      "loss": 0.6358,
      "step": 140
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999869769742604,
      "loss": 0.6332,
      "step": 141
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999829903450295,
      "loss": 0.6357,
      "step": 142
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999784721716393,
      "loss": 0.6485,
      "step": 143
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001999973422456491,
      "loss": 0.6348,
      "step": 144
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999678412022687,
      "loss": 0.6545,
      "step": 145
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999617284119394,
      "loss": 0.6472,
      "step": 146
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999550840887523,
      "loss": 0.6402,
      "step": 147
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999479082362392,
      "loss": 0.6415,
      "step": 148
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999402008582143,
      "loss": 0.6335,
      "step": 149
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999319619587748,
      "loss": 0.6571,
      "step": 150
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999231915423,
      "loss": 0.6285,
      "step": 151
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999138896134518,
      "loss": 0.626,
      "step": 152
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019999040561771743,
      "loss": 0.6434,
      "step": 153
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001999893691238695,
      "loss": 0.6507,
      "step": 154
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019998827948035234,
      "loss": 0.6531,
      "step": 155
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001999871366877451,
      "loss": 0.634,
      "step": 156
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019998594074665534,
      "loss": 0.6147,
      "step": 157
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019998469165771865,
      "loss": 0.6364,
      "step": 158
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019998338942159904,
      "loss": 0.6298,
      "step": 159
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019998203403898872,
      "loss": 0.6283,
      "step": 160
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019998062551060817,
      "loss": 0.6314,
      "step": 161
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019997916383720604,
      "loss": 0.6432,
      "step": 162
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019997764901955934,
      "loss": 0.6305,
      "step": 163
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001999760810584732,
      "loss": 0.629,
      "step": 164
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001999744599547812,
      "loss": 0.628,
      "step": 165
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019997278570934492,
      "loss": 0.6202,
      "step": 166
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001999710583230544,
      "loss": 0.6531,
      "step": 167
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001999692777968278,
      "loss": 0.6485,
      "step": 168
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019996744413161154,
      "loss": 0.6267,
      "step": 169
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019996555732838033,
      "loss": 0.6193,
      "step": 170
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019996361738813714,
      "loss": 0.6239,
      "step": 171
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001999616243119131,
      "loss": 0.6375,
      "step": 172
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019995957810076768,
      "loss": 0.6255,
      "step": 173
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019995747875578852,
      "loss": 0.637,
      "step": 174
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019995532627809155,
      "loss": 0.6333,
      "step": 175
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001999531206688209,
      "loss": 0.6231,
      "step": 176
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019995086192914902,
      "loss": 0.6348,
      "step": 177
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001999485500602765,
      "loss": 0.635,
      "step": 178
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019994618506343224,
      "loss": 0.622,
      "step": 179
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019994376693987332,
      "loss": 0.6224,
      "step": 180
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001999412956908852,
      "loss": 0.6308,
      "step": 181
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001999387713177814,
      "loss": 0.6456,
      "step": 182
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019993619382190382,
      "loss": 0.633,
      "step": 183
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001999335632046224,
      "loss": 0.6225,
      "step": 184
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019993087946733565,
      "loss": 0.6183,
      "step": 185
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019992814261146998,
      "loss": 0.6311,
      "step": 186
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001999253526384802,
      "loss": 0.6496,
      "step": 187
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019992250954984936,
      "loss": 0.6381,
      "step": 188
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001999196133470887,
      "loss": 0.6179,
      "step": 189
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019991666403173773,
      "loss": 0.6237,
      "step": 190
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019991366160536408,
      "loss": 0.6242,
      "step": 191
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001999106060695638,
      "loss": 0.6202,
      "step": 192
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019990749742596103,
      "loss": 0.6422,
      "step": 193
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019990433567620815,
      "loss": 0.6308,
      "step": 194
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019990112082198582,
      "loss": 0.6384,
      "step": 195
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019989785286500295,
      "loss": 0.6215,
      "step": 196
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019989453180699657,
      "loss": 0.6129,
      "step": 197
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019989115764973203,
      "loss": 0.629,
      "step": 198
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019988773039500288,
      "loss": 0.6045,
      "step": 199
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019988425004463086,
      "loss": 0.6439,
      "step": 200
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019988071660046602,
      "loss": 0.6135,
      "step": 201
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019987713006438648,
      "loss": 0.6264,
      "step": 202
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019987349043829874,
      "loss": 0.6279,
      "step": 203
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019986979772413746,
      "loss": 0.6185,
      "step": 204
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001998660519238655,
      "loss": 0.6172,
      "step": 205
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019986225303947396,
      "loss": 0.6186,
      "step": 206
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001998584010729821,
      "loss": 0.6385,
      "step": 207
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019985449602643747,
      "loss": 0.6191,
      "step": 208
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019985053790191586,
      "loss": 0.6167,
      "step": 209
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019984652670152118,
      "loss": 0.6266,
      "step": 210
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019984246242738557,
      "loss": 0.6318,
      "step": 211
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019983834508166946,
      "loss": 0.6349,
      "step": 212
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019983417466656143,
      "loss": 0.6275,
      "step": 213
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001998299511842783,
      "loss": 0.6093,
      "step": 214
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019982567463706495,
      "loss": 0.6227,
      "step": 215
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019982134502719476,
      "loss": 0.6305,
      "step": 216
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019981696235696906,
      "loss": 0.6281,
      "step": 217
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019981252662871744,
      "loss": 0.6195,
      "step": 218
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019980803784479782,
      "loss": 0.6157,
      "step": 219
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019980349600759616,
      "loss": 0.6296,
      "step": 220
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001997989011195267,
      "loss": 0.6221,
      "step": 221
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019979425318303188,
      "loss": 0.6149,
      "step": 222
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019978955220058232,
      "loss": 0.6174,
      "step": 223
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019978479817467682,
      "loss": 0.6112,
      "step": 224
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019977999110784244,
      "loss": 0.6173,
      "step": 225
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019977513100263438,
      "loss": 0.6208,
      "step": 226
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019977021786163598,
      "loss": 0.6231,
      "step": 227
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019976525168745896,
      "loss": 0.6165,
      "step": 228
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019976023248274302,
      "loss": 0.6206,
      "step": 229
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019975516025015614,
      "loss": 0.615,
      "step": 230
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019975003499239446,
      "loss": 0.6193,
      "step": 231
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001997448567121824,
      "loss": 0.6157,
      "step": 232
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019973962541227245,
      "loss": 0.6107,
      "step": 233
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019973434109544531,
      "loss": 0.6066,
      "step": 234
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019972900376450986,
      "loss": 0.617,
      "step": 235
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019972361342230328,
      "loss": 0.6262,
      "step": 236
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019971817007169069,
      "loss": 0.6087,
      "step": 237
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001997126737155656,
      "loss": 0.6072,
      "step": 238
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019970712435684958,
      "loss": 0.6237,
      "step": 239
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019970152199849246,
      "loss": 0.6172,
      "step": 240
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019969586664347216,
      "loss": 0.6195,
      "step": 241
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019969015829479476,
      "loss": 0.6275,
      "step": 242
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001996843969554946,
      "loss": 0.6041,
      "step": 243
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019967858262863412,
      "loss": 0.6134,
      "step": 244
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019967271531730395,
      "loss": 0.6208,
      "step": 245
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019966679502462294,
      "loss": 0.6233,
      "step": 246
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019966082175373791,
      "loss": 0.6115,
      "step": 247
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019965479550782407,
      "loss": 0.6191,
      "step": 248
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019964871629008467,
      "loss": 0.6213,
      "step": 249
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001996425841037511,
      "loss": 0.6064,
      "step": 250
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019963639895208298,
      "loss": 0.6076,
      "step": 251
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019963016083836808,
      "loss": 0.6086,
      "step": 252
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001996238697659222,
      "loss": 0.6051,
      "step": 253
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019961752573808945,
      "loss": 0.6147,
      "step": 254
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000199611128758242,
      "loss": 0.6083,
      "step": 255
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001996046788297802,
      "loss": 0.6032,
      "step": 256
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019959817595613247,
      "loss": 0.6243,
      "step": 257
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019959162014075553,
      "loss": 0.6275,
      "step": 258
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019958501138713403,
      "loss": 0.6132,
      "step": 259
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000199578349698781,
      "loss": 0.6285,
      "step": 260
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019957163507923739,
      "loss": 0.6273,
      "step": 261
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019956486753207239,
      "loss": 0.6093,
      "step": 262
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019955804706088334,
      "loss": 0.609,
      "step": 263
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019955117366929565,
      "loss": 0.6083,
      "step": 264
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019954424736096292,
      "loss": 0.6097,
      "step": 265
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019953726813956682,
      "loss": 0.6012,
      "step": 266
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019953023600881723,
      "loss": 0.6011,
      "step": 267
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019952315097245208,
      "loss": 0.6249,
      "step": 268
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019951601303423743,
      "loss": 0.6063,
      "step": 269
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019950882219796747,
      "loss": 0.6033,
      "step": 270
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019950157846746453,
      "loss": 0.6145,
      "step": 271
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019949428184657905,
      "loss": 0.6038,
      "step": 272
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019948693233918952,
      "loss": 0.5937,
      "step": 273
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019947952994920265,
      "loss": 0.6087,
      "step": 274
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019947207468055318,
      "loss": 0.6095,
      "step": 275
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019946456653720395,
      "loss": 0.6033,
      "step": 276
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019945700552314598,
      "loss": 0.6175,
      "step": 277
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019944939164239837,
      "loss": 0.6126,
      "step": 278
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019944172489900827,
      "loss": 0.6067,
      "step": 279
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019943400529705095,
      "loss": 0.599,
      "step": 280
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019942623284062983,
      "loss": 0.6132,
      "step": 281
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019941840753387634,
      "loss": 0.6045,
      "step": 282
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019941052938095006,
      "loss": 0.6007,
      "step": 283
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019940259838603866,
      "loss": 0.5995,
      "step": 284
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001993946145533579,
      "loss": 0.5852,
      "step": 285
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019938657788715153,
      "loss": 0.6274,
      "step": 286
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001993784883916916,
      "loss": 0.6117,
      "step": 287
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000199370346071278,
      "loss": 0.6027,
      "step": 288
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019936215093023884,
      "loss": 0.6177,
      "step": 289
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019935390297293027,
      "loss": 0.6088,
      "step": 290
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019934560220373655,
      "loss": 0.6006,
      "step": 291
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019933724862706992,
      "loss": 0.6135,
      "step": 292
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019932884224737078,
      "loss": 0.6188,
      "step": 293
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001993203830691076,
      "loss": 0.6045,
      "step": 294
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019931187109677685,
      "loss": 0.6139,
      "step": 295
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001993033063349031,
      "loss": 0.5977,
      "step": 296
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019929468878803895,
      "loss": 0.5967,
      "step": 297
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019928601846076513,
      "loss": 0.6082,
      "step": 298
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019927729535769038,
      "loss": 0.6121,
      "step": 299
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019926851948345146,
      "loss": 0.6052,
      "step": 300
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019925969084271322,
      "loss": 0.5896,
      "step": 301
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019925080944016856,
      "loss": 0.6019,
      "step": 302
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019924187528053846,
      "loss": 0.5948,
      "step": 303
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001992328883685718,
      "loss": 0.5981,
      "step": 304
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019922384870904572,
      "loss": 0.6261,
      "step": 305
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019921475630676517,
      "loss": 0.5982,
      "step": 306
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019920561116656336,
      "loss": 0.6011,
      "step": 307
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001991964132933013,
      "loss": 0.6264,
      "step": 308
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019918716269186823,
      "loss": 0.6064,
      "step": 309
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019917785936718132,
      "loss": 0.5949,
      "step": 310
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019916850332418574,
      "loss": 0.6032,
      "step": 311
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001991590945678548,
      "loss": 0.6053,
      "step": 312
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001991496331031897,
      "loss": 0.6039,
      "step": 313
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019914011893521974,
      "loss": 0.5951,
      "step": 314
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019913055206900214,
      "loss": 0.5914,
      "step": 315
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019912093250962229,
      "loss": 0.6002,
      "step": 316
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019911126026219344,
      "loss": 0.602,
      "step": 317
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019910153533185697,
      "loss": 0.6124,
      "step": 318
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001990917577237821,
      "loss": 0.6038,
      "step": 319
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001990819274431662,
      "loss": 0.6052,
      "step": 320
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019907204449523463,
      "loss": 0.6075,
      "step": 321
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001990621088852406,
      "loss": 0.6002,
      "step": 322
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001990521206184655,
      "loss": 0.6105,
      "step": 323
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001990420797002186,
      "loss": 0.6145,
      "step": 324
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019903198613583717,
      "loss": 0.6028,
      "step": 325
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019902183993068652,
      "loss": 0.6034,
      "step": 326
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019901164109015985,
      "loss": 0.6005,
      "step": 327
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019900138961967837,
      "loss": 0.6056,
      "step": 328
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019899108552469134,
      "loss": 0.6003,
      "step": 329
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019898072881067592,
      "loss": 0.6129,
      "step": 330
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019897031948313724,
      "loss": 0.61,
      "step": 331
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001989598575476084,
      "loss": 0.5978,
      "step": 332
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001989493430096505,
      "loss": 0.603,
      "step": 333
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019893877587485258,
      "loss": 0.6101,
      "step": 334
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019892815614883157,
      "loss": 0.607,
      "step": 335
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001989174838372325,
      "loss": 0.5858,
      "step": 336
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019890675894572825,
      "loss": 0.5956,
      "step": 337
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019889598148001964,
      "loss": 0.5811,
      "step": 338
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019888515144583552,
      "loss": 0.6003,
      "step": 339
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019887426884893257,
      "loss": 0.6089,
      "step": 340
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019886333369509548,
      "loss": 0.599,
      "step": 341
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019885234599013694,
      "loss": 0.5894,
      "step": 342
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019884130573989738,
      "loss": 0.6032,
      "step": 343
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019883021295024537,
      "loss": 0.5975,
      "step": 344
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019881906762707731,
      "loss": 0.5874,
      "step": 345
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019880786977631747,
      "loss": 0.6222,
      "step": 346
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019879661940391818,
      "loss": 0.5979,
      "step": 347
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019878531651585957,
      "loss": 0.6064,
      "step": 348
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019877396111814974,
      "loss": 0.5835,
      "step": 349
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019876255321682468,
      "loss": 0.5919,
      "step": 350
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019875109281794825,
      "loss": 0.5957,
      "step": 351
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019873957992761238,
      "loss": 0.6203,
      "step": 352
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019872801455193665,
      "loss": 0.6129,
      "step": 353
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001987163966970688,
      "loss": 0.6271,
      "step": 354
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019870472636918422,
      "loss": 0.6084,
      "step": 355
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001986930035744864,
      "loss": 0.587,
      "step": 356
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001986812283192066,
      "loss": 0.6021,
      "step": 357
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019866940060960398,
      "loss": 0.5926,
      "step": 358
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019865752045196557,
      "loss": 0.6003,
      "step": 359
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019864558785260636,
      "loss": 0.5977,
      "step": 360
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019863360281786917,
      "loss": 0.6097,
      "step": 361
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001986215653541246,
      "loss": 0.6032,
      "step": 362
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019860947546777128,
      "loss": 0.601,
      "step": 363
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019859733316523564,
      "loss": 0.581,
      "step": 364
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019858513845297192,
      "loss": 0.5794,
      "step": 365
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019857289133746222,
      "loss": 0.5876,
      "step": 366
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001985605918252166,
      "loss": 0.5925,
      "step": 367
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001985482399227729,
      "loss": 0.5913,
      "step": 368
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019853583563669678,
      "loss": 0.5975,
      "step": 369
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019852337897358177,
      "loss": 0.6163,
      "step": 370
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001985108699400493,
      "loss": 0.5944,
      "step": 371
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019849830854274858,
      "loss": 0.5798,
      "step": 372
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001984856947883566,
      "loss": 0.5917,
      "step": 373
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001984730286835783,
      "loss": 0.6027,
      "step": 374
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019846031023514638,
      "loss": 0.5872,
      "step": 375
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019844753944982135,
      "loss": 0.5833,
      "step": 376
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001984347163343916,
      "loss": 0.6033,
      "step": 377
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019842184089567325,
      "loss": 0.5985,
      "step": 378
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001984089131405103,
      "loss": 0.5902,
      "step": 379
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019839593307577452,
      "loss": 0.62,
      "step": 380
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019838290070836557,
      "loss": 0.6039,
      "step": 381
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019836981604521076,
      "loss": 0.5879,
      "step": 382
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019835667909326538,
      "loss": 0.602,
      "step": 383
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019834348985951232,
      "loss": 0.6023,
      "step": 384
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019833024835096244,
      "loss": 0.5941,
      "step": 385
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001983169545746543,
      "loss": 0.5997,
      "step": 386
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019830360853765418,
      "loss": 0.6022,
      "step": 387
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019829021024705626,
      "loss": 0.6061,
      "step": 388
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019827675970998242,
      "loss": 0.5902,
      "step": 389
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019826325693358236,
      "loss": 0.6029,
      "step": 390
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019824970192503354,
      "loss": 0.5874,
      "step": 391
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001982360946915411,
      "loss": 0.5997,
      "step": 392
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001982224352403381,
      "loss": 0.6034,
      "step": 393
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019820872357868518,
      "loss": 0.583,
      "step": 394
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001981949597138709,
      "loss": 0.589,
      "step": 395
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019818114365321142,
      "loss": 0.6027,
      "step": 396
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001981672754040507,
      "loss": 0.5913,
      "step": 397
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019815335497376052,
      "loss": 0.5896,
      "step": 398
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001981393823697403,
      "loss": 0.5975,
      "step": 399
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001981253575994172,
      "loss": 0.5905,
      "step": 400
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019811128067024616,
      "loss": 0.5918,
      "step": 401
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019809715158970978,
      "loss": 0.6002,
      "step": 402
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019808297036531846,
      "loss": 0.5877,
      "step": 403
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019806873700461028,
      "loss": 0.5847,
      "step": 404
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019805445151515095,
      "loss": 0.5962,
      "step": 405
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019804011390453407,
      "loss": 0.6044,
      "step": 406
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001980257241803808,
      "loss": 0.5913,
      "step": 407
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019801128235034002,
      "loss": 0.6149,
      "step": 408
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019799678842208836,
      "loss": 0.5884,
      "step": 409
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019798224240333013,
      "loss": 0.5896,
      "step": 410
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019796764430179726,
      "loss": 0.6068,
      "step": 411
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019795299412524945,
      "loss": 0.5992,
      "step": 412
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019793829188147406,
      "loss": 0.5813,
      "step": 413
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001979235375782861,
      "loss": 0.5869,
      "step": 414
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019790873122352824,
      "loss": 0.5708,
      "step": 415
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001978938728250709,
      "loss": 0.5939,
      "step": 416
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019787896239081208,
      "loss": 0.5995,
      "step": 417
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019786399992867743,
      "loss": 0.5924,
      "step": 418
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001978489854466204,
      "loss": 0.5896,
      "step": 419
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019783391895262183,
      "loss": 0.5693,
      "step": 420
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019781880045469054,
      "loss": 0.5919,
      "step": 421
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019780362996086267,
      "loss": 0.5938,
      "step": 422
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019778840747920225,
      "loss": 0.5846,
      "step": 423
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019777313301780079,
      "loss": 0.5901,
      "step": 424
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019775780658477747,
      "loss": 0.5874,
      "step": 425
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001977424281882791,
      "loss": 0.5984,
      "step": 426
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019772699783648015,
      "loss": 0.6004,
      "step": 427
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019771151553758264,
      "loss": 0.5845,
      "step": 428
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001976959812998163,
      "loss": 0.5921,
      "step": 429
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001976803951314383,
      "loss": 0.5992,
      "step": 430
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001976647570407336,
      "loss": 0.5907,
      "step": 431
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019764906703601464,
      "loss": 0.6098,
      "step": 432
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001976333251256215,
      "loss": 0.6026,
      "step": 433
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019761753131792187,
      "loss": 0.598,
      "step": 434
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019760168562131094,
      "loss": 0.586,
      "step": 435
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001975857880442116,
      "loss": 0.5996,
      "step": 436
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019756983859507422,
      "loss": 0.6117,
      "step": 437
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019755383728237678,
      "loss": 0.5777,
      "step": 438
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019753778411462486,
      "loss": 0.5907,
      "step": 439
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019752167910035156,
      "loss": 0.5836,
      "step": 440
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019750552224811753,
      "loss": 0.5972,
      "step": 441
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019748931356651105,
      "loss": 0.6016,
      "step": 442
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019747305306414781,
      "loss": 0.5869,
      "step": 443
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001974567407496712,
      "loss": 0.6042,
      "step": 444
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019744037663175208,
      "loss": 0.5852,
      "step": 445
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019742396071908882,
      "loss": 0.5932,
      "step": 446
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019740749302040738,
      "loss": 0.5903,
      "step": 447
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001973909735444612,
      "loss": 0.5807,
      "step": 448
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019737440230003126,
      "loss": 0.598,
      "step": 449
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00019735777929592608,
      "loss": 0.5867,
      "step": 450
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001973411045409816,
      "loss": 0.5781,
      "step": 451
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019732437804406144,
      "loss": 0.5844,
      "step": 452
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019730759981405655,
      "loss": 0.5949,
      "step": 453
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001972907698598855,
      "loss": 0.597,
      "step": 454
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019727388819049425,
      "loss": 0.5949,
      "step": 455
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019725695481485637,
      "loss": 0.5988,
      "step": 456
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001972399697419728,
      "loss": 0.5886,
      "step": 457
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019722293298087205,
      "loss": 0.5838,
      "step": 458
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019720584454061002,
      "loss": 0.6099,
      "step": 459
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001971887044302702,
      "loss": 0.594,
      "step": 460
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019717151265896335,
      "loss": 0.5913,
      "step": 461
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019715426923582792,
      "loss": 0.5955,
      "step": 462
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019713697417002963,
      "loss": 0.582,
      "step": 463
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001971196274707618,
      "loss": 0.5938,
      "step": 464
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019710222914724507,
      "loss": 0.5898,
      "step": 465
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00019708477920872758,
      "loss": 0.5857,
      "step": 466
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019706727766448493,
      "loss": 0.5862,
      "step": 467
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019704972452382015,
      "loss": 0.5912,
      "step": 468
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019703211979606357,
      "loss": 0.596,
      "step": 469
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001970144634905731,
      "loss": 0.587,
      "step": 470
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000196996755616734,
      "loss": 0.5975,
      "step": 471
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019697899618395894,
      "loss": 0.5823,
      "step": 472
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000196961185201688,
      "loss": 0.5893,
      "step": 473
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019694332267938868,
      "loss": 0.581,
      "step": 474
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019692540862655585,
      "loss": 0.5858,
      "step": 475
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001969074430527118,
      "loss": 0.6096,
      "step": 476
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001968894259674062,
      "loss": 0.5942,
      "step": 477
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019687135738021602,
      "loss": 0.5943,
      "step": 478
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019685323730074574,
      "loss": 0.5973,
      "step": 479
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019683506573862713,
      "loss": 0.5871,
      "step": 480
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00019681684270351932,
      "loss": 0.5916,
      "step": 481
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019679856820510884,
      "loss": 0.5855,
      "step": 482
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019678024225310958,
      "loss": 0.5856,
      "step": 483
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001967618648572627,
      "loss": 0.5754,
      "step": 484
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019674343602733686,
      "loss": 0.5913,
      "step": 485
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019672495577312783,
      "loss": 0.588,
      "step": 486
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019670642410445899,
      "loss": 0.5987,
      "step": 487
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001966878410311808,
      "loss": 0.5848,
      "step": 488
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019666920656317117,
      "loss": 0.5888,
      "step": 489
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019665052071033537,
      "loss": 0.5971,
      "step": 490
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019663178348260592,
      "loss": 0.5967,
      "step": 491
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019661299488994257,
      "loss": 0.5874,
      "step": 492
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001965941549423325,
      "loss": 0.5844,
      "step": 493
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001965752636497902,
      "loss": 0.5822,
      "step": 494
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00019655632102235732,
      "loss": 0.5846,
      "step": 495
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001965373270701029,
      "loss": 0.5788,
      "step": 496
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019651828180312325,
      "loss": 0.595,
      "step": 497
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019649918523154193,
      "loss": 0.593,
      "step": 498
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001964800373655098,
      "loss": 0.5784,
      "step": 499
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019646083821520493,
      "loss": 0.5967,
      "step": 500
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019644158779083272,
      "loss": 0.5817,
      "step": 501
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019642228610262583,
      "loss": 0.5743,
      "step": 502
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019640293316084404,
      "loss": 0.5779,
      "step": 503
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019638352897577454,
      "loss": 0.5803,
      "step": 504
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001963640735577317,
      "loss": 0.589,
      "step": 505
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019634456691705702,
      "loss": 0.5927,
      "step": 506
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019632500906411942,
      "loss": 0.585,
      "step": 507
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001963054000093149,
      "loss": 0.5975,
      "step": 508
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00019628573976306668,
      "loss": 0.5955,
      "step": 509
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001962660283358253,
      "loss": 0.5821,
      "step": 510
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019624626573806837,
      "loss": 0.5877,
      "step": 511
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019622645198030076,
      "loss": 0.5993,
      "step": 512
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001962065870730546,
      "loss": 0.5926,
      "step": 513
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019618667102688905,
      "loss": 0.5737,
      "step": 514
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019616670385239066,
      "loss": 0.5962,
      "step": 515
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019614668556017294,
      "loss": 0.5923,
      "step": 516
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019612661616087674,
      "loss": 0.5887,
      "step": 517
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019610649566517,
      "loss": 0.5899,
      "step": 518
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019608632408374787,
      "loss": 0.5843,
      "step": 519
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019606610142733252,
      "loss": 0.5899,
      "step": 520
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001960458277066735,
      "loss": 0.5874,
      "step": 521
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019602550293254723,
      "loss": 0.5924,
      "step": 522
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019600512711575753,
      "loss": 0.5968,
      "step": 523
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019598470026713514,
      "loss": 0.5834,
      "step": 524
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00019596422239753807,
      "loss": 0.5798,
      "step": 525
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00019594369351785137,
      "loss": 0.5798,
      "step": 526
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00019592311363898727,
      "loss": 0.5974,
      "step": 527
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00019590248277188502,
      "loss": 0.579,
      "step": 528
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001958818009275111,
      "loss": 0.5675,
      "step": 529
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001958610681168589,
      "loss": 0.5893,
      "step": 530
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00019584028435094909,
      "loss": 0.5885,
      "step": 531
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00019581944964082933,
      "loss": 0.5874,
      "step": 532
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001957985639975744,
      "loss": 0.6013,
      "step": 533
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001957776274322861,
      "loss": 0.582,
      "step": 534
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001957566399560933,
      "loss": 0.5966,
      "step": 535
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000195735601580152,
      "loss": 0.5763,
      "step": 536
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00019571451231564525,
      "loss": 0.5987,
      "step": 537
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00019569337217378304,
      "loss": 0.5913,
      "step": 538
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00019567218116580253,
      "loss": 0.5746,
      "step": 539
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00019565093930296786,
      "loss": 0.5958,
      "step": 540
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019562964659657018,
      "loss": 0.5903,
      "step": 541
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019560830305792774,
      "loss": 0.5877,
      "step": 542
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019558690869838575,
      "loss": 0.5939,
      "step": 543
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019556546352931644,
      "loss": 0.5953,
      "step": 544
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019554396756211908,
      "loss": 0.5963,
      "step": 545
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001955224208082199,
      "loss": 0.5776,
      "step": 546
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019550082327907212,
      "loss": 0.5871,
      "step": 547
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019547917498615608,
      "loss": 0.5757,
      "step": 548
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019545747594097886,
      "loss": 0.5979,
      "step": 549
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019543572615507475,
      "loss": 0.5884,
      "step": 550
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019541392564000488,
      "loss": 0.5863,
      "step": 551
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019539207440735743,
      "loss": 0.5799,
      "step": 552
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019537017246874742,
      "loss": 0.5816,
      "step": 553
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019534821983581698,
      "loss": 0.5758,
      "step": 554
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00019532621652023502,
      "loss": 0.607,
      "step": 555
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001953041625336975,
      "loss": 0.5822,
      "step": 556
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019528205788792733,
      "loss": 0.5823,
      "step": 557
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019525990259467425,
      "loss": 0.5808,
      "step": 558
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000195237696665715,
      "loss": 0.5768,
      "step": 559
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019521544011285322,
      "loss": 0.5957,
      "step": 560
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019519313294791944,
      "loss": 0.5892,
      "step": 561
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019517077518277113,
      "loss": 0.5799,
      "step": 562
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019514836682929258,
      "loss": 0.5803,
      "step": 563
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019512590789939505,
      "loss": 0.5947,
      "step": 564
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019510339840501668,
      "loss": 0.5819,
      "step": 565
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019508083835812243,
      "loss": 0.5782,
      "step": 566
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019505822777070417,
      "loss": 0.5896,
      "step": 567
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019503556665478067,
      "loss": 0.5902,
      "step": 568
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00019501285502239744,
      "loss": 0.5911,
      "step": 569
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000194990092885627,
      "loss": 0.5763,
      "step": 570
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001949672802565686,
      "loss": 0.5741,
      "step": 571
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019494441714734836,
      "loss": 0.5738,
      "step": 572
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019492150357011925,
      "loss": 0.5834,
      "step": 573
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019489853953706104,
      "loss": 0.5811,
      "step": 574
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019487552506038038,
      "loss": 0.6025,
      "step": 575
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019485246015231065,
      "loss": 0.5881,
      "step": 576
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019482934482511208,
      "loss": 0.593,
      "step": 577
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001948061790910717,
      "loss": 0.5743,
      "step": 578
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019478296296250334,
      "loss": 0.5822,
      "step": 579
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019475969645174756,
      "loss": 0.5832,
      "step": 580
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019473637957117182,
      "loss": 0.5837,
      "step": 581
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019471301233317024,
      "loss": 0.59,
      "step": 582
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019468959475016378,
      "loss": 0.5876,
      "step": 583
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00019466612683460008,
      "loss": 0.5804,
      "step": 584
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00019464260859895362,
      "loss": 0.5774,
      "step": 585
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00019461904005572557,
      "loss": 0.5785,
      "step": 586
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001945954212174439,
      "loss": 0.5939,
      "step": 587
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00019457175209666329,
      "loss": 0.5873,
      "step": 588
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00019454803270596506,
      "loss": 0.5949,
      "step": 589
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00019452426305795743,
      "loss": 0.6005,
      "step": 590
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00019450044316527513,
      "loss": 0.5853,
      "step": 591
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001944765730405798,
      "loss": 0.5799,
      "step": 592
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001944526526965596,
      "loss": 0.5876,
      "step": 593
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001944286821459295,
      "loss": 0.5905,
      "step": 594
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001944046614014311,
      "loss": 0.5898,
      "step": 595
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00019438059047583274,
      "loss": 0.57,
      "step": 596
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001943564693819294,
      "loss": 0.5793,
      "step": 597
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00019433229813254268,
      "loss": 0.5739,
      "step": 598
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00019430807674052092,
      "loss": 0.5889,
      "step": 599
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00019428380521873905,
      "loss": 0.5968,
      "step": 600
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00019425948358009875,
      "loss": 0.5897,
      "step": 601
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00019423511183752816,
      "loss": 0.5816,
      "step": 602
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001942106900039822,
      "loss": 0.5803,
      "step": 603
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001941862180924424,
      "loss": 0.5995,
      "step": 604
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00019416169611591685,
      "loss": 0.5955,
      "step": 605
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00019413712408744034,
      "loss": 0.5773,
      "step": 606
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001941125020200741,
      "loss": 0.584,
      "step": 607
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00019408782992690618,
      "loss": 0.5936,
      "step": 608
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001940631078210511,
      "loss": 0.5732,
      "step": 609
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001940383357156499,
      "loss": 0.5829,
      "step": 610
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00019401351362387034,
      "loss": 0.5773,
      "step": 611
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00019398864155890659,
      "loss": 0.6002,
      "step": 612
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001939637195339796,
      "loss": 0.5723,
      "step": 613
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00019393874756233669,
      "loss": 0.5957,
      "step": 614
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00019391372565725177,
      "loss": 0.5692,
      "step": 615
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00019388865383202538,
      "loss": 0.5888,
      "step": 616
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00019386353209998447,
      "loss": 0.5795,
      "step": 617
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001938383604744826,
      "loss": 0.5696,
      "step": 618
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00019381313896889982,
      "loss": 0.579,
      "step": 619
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001937878675966427,
      "loss": 0.5823,
      "step": 620
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00019376254637114434,
      "loss": 0.5702,
      "step": 621
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00019373717530586426,
      "loss": 0.5926,
      "step": 622
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001937117544142886,
      "loss": 0.5807,
      "step": 623
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001936862837099299,
      "loss": 0.58,
      "step": 624
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001936607632063272,
      "loss": 0.5704,
      "step": 625
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00019363519291704594,
      "loss": 0.6036,
      "step": 626
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00019360957285567817,
      "loss": 0.588,
      "step": 627
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001935839030358423,
      "loss": 0.5765,
      "step": 628
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00019355818347118318,
      "loss": 0.5918,
      "step": 629
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00019353241417537214,
      "loss": 0.602,
      "step": 630
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001935065951621069,
      "loss": 0.5614,
      "step": 631
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001934807264451117,
      "loss": 0.5785,
      "step": 632
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001934548080381371,
      "loss": 0.5705,
      "step": 633
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001934288399549601,
      "loss": 0.5936,
      "step": 634
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00019340282220938412,
      "loss": 0.5917,
      "step": 635
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00019337675481523905,
      "loss": 0.5804,
      "step": 636
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000193350637786381,
      "loss": 0.5759,
      "step": 637
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001933244711366926,
      "loss": 0.5884,
      "step": 638
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00019329825488008284,
      "loss": 0.5863,
      "step": 639
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00019327198903048703,
      "loss": 0.5839,
      "step": 640
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00019324567360186684,
      "loss": 0.5886,
      "step": 641
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001932193086082104,
      "loss": 0.5697,
      "step": 642
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000193192894063532,
      "loss": 0.5838,
      "step": 643
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00019316642998187247,
      "loss": 0.5819,
      "step": 644
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001931399163772988,
      "loss": 0.5802,
      "step": 645
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019311335326390445,
      "loss": 0.5702,
      "step": 646
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019308674065580906,
      "loss": 0.5703,
      "step": 647
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019306007856715872,
      "loss": 0.5747,
      "step": 648
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019303336701212572,
      "loss": 0.5721,
      "step": 649
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019300660600490862,
      "loss": 0.5756,
      "step": 650
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019297979555973243,
      "loss": 0.5791,
      "step": 651
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001929529356908482,
      "loss": 0.5824,
      "step": 652
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019292602641253349,
      "loss": 0.5742,
      "step": 653
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001928990677390919,
      "loss": 0.5729,
      "step": 654
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019287205968485354,
      "loss": 0.5867,
      "step": 655
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019284500226417453,
      "loss": 0.5872,
      "step": 656
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019281789549143734,
      "loss": 0.572,
      "step": 657
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001927907393810507,
      "loss": 0.5824,
      "step": 658
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00019276353394744944,
      "loss": 0.5791,
      "step": 659
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019273627920509477,
      "loss": 0.57,
      "step": 660
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019270897516847403,
      "loss": 0.5877,
      "step": 661
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019268162185210076,
      "loss": 0.6029,
      "step": 662
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019265421927051475,
      "loss": 0.5687,
      "step": 663
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019262676743828184,
      "loss": 0.5703,
      "step": 664
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019259926636999418,
      "loss": 0.5676,
      "step": 665
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019257171608027003,
      "loss": 0.5811,
      "step": 666
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019254411658375386,
      "loss": 0.5725,
      "step": 667
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001925164678951163,
      "loss": 0.5903,
      "step": 668
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019248877002905405,
      "loss": 0.5707,
      "step": 669
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019246102300029004,
      "loss": 0.5813,
      "step": 670
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001924332268235732,
      "loss": 0.5779,
      "step": 671
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019240538151367882,
      "loss": 0.5742,
      "step": 672
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00019237748708540804,
      "loss": 0.5896,
      "step": 673
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019234954355358829,
      "loss": 0.5731,
      "step": 674
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019232155093307305,
      "loss": 0.5852,
      "step": 675
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019229350923874184,
      "loss": 0.5768,
      "step": 676
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019226541848550038,
      "loss": 0.5807,
      "step": 677
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019223727868828035,
      "loss": 0.5744,
      "step": 678
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019220908986203956,
      "loss": 0.5721,
      "step": 679
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001921808520217619,
      "loss": 0.59,
      "step": 680
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019215256518245722,
      "loss": 0.5828,
      "step": 681
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001921242293591615,
      "loss": 0.5742,
      "step": 682
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001920958445669368,
      "loss": 0.5815,
      "step": 683
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019206741082087109,
      "loss": 0.5773,
      "step": 684
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019203892813607842,
      "loss": 0.5919,
      "step": 685
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019201039652769889,
      "loss": 0.5692,
      "step": 686
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001919818160108985,
      "loss": 0.5893,
      "step": 687
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00019195318660086938,
      "loss": 0.5721,
      "step": 688
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019192450831282955,
      "loss": 0.5746,
      "step": 689
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019189578116202307,
      "loss": 0.5748,
      "step": 690
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019186700516371993,
      "loss": 0.574,
      "step": 691
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019183818033321614,
      "loss": 0.5746,
      "step": 692
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019180930668583355,
      "loss": 0.5775,
      "step": 693
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019178038423692012,
      "loss": 0.587,
      "step": 694
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001917514130018496,
      "loss": 0.5852,
      "step": 695
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019172239299602187,
      "loss": 0.5968,
      "step": 696
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019169332423486245,
      "loss": 0.5739,
      "step": 697
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000191664206733823,
      "loss": 0.5803,
      "step": 698
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019163504050838105,
      "loss": 0.5784,
      "step": 699
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019160582557403998,
      "loss": 0.5869,
      "step": 700
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001915765619463291,
      "loss": 0.5689,
      "step": 701
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001915472496408035,
      "loss": 0.5717,
      "step": 702
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00019151788867304435,
      "loss": 0.584,
      "step": 703
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019148847905865846,
      "loss": 0.5663,
      "step": 704
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019145902081327867,
      "loss": 0.5857,
      "step": 705
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001914295139525636,
      "loss": 0.5824,
      "step": 706
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001913999584921977,
      "loss": 0.5823,
      "step": 707
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019137035444789126,
      "loss": 0.5854,
      "step": 708
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019134070183538042,
      "loss": 0.5671,
      "step": 709
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019131100067042712,
      "loss": 0.5905,
      "step": 710
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019128125096881912,
      "loss": 0.5679,
      "step": 711
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019125145274636996,
      "loss": 0.5805,
      "step": 712
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019122160601891896,
      "loss": 0.5773,
      "step": 713
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019119171080233128,
      "loss": 0.5862,
      "step": 714
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001911617671124978,
      "loss": 0.5726,
      "step": 715
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019113177496533513,
      "loss": 0.5791,
      "step": 716
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001911017343767858,
      "loss": 0.568,
      "step": 717
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00019107164536281794,
      "loss": 0.5786,
      "step": 718
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001910415079394254,
      "loss": 0.5631,
      "step": 719
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001910113221226279,
      "loss": 0.585,
      "step": 720
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019098108792847077,
      "loss": 0.5805,
      "step": 721
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001909508053730251,
      "loss": 0.5882,
      "step": 722
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019092047447238773,
      "loss": 0.5624,
      "step": 723
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019089009524268105,
      "loss": 0.5836,
      "step": 724
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019085966770005327,
      "loss": 0.5648,
      "step": 725
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001908291918606783,
      "loss": 0.5665,
      "step": 726
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019079866774075562,
      "loss": 0.5949,
      "step": 727
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019076809535651044,
      "loss": 0.5854,
      "step": 728
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001907374747241936,
      "loss": 0.5793,
      "step": 729
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019070680586008161,
      "loss": 0.5649,
      "step": 730
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019067608878047657,
      "loss": 0.5782,
      "step": 731
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019064532350170627,
      "loss": 0.5689,
      "step": 732
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00019061451004012412,
      "loss": 0.5893,
      "step": 733
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019058364841210908,
      "loss": 0.5701,
      "step": 734
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001905527386340657,
      "loss": 0.5852,
      "step": 735
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019052178072242426,
      "loss": 0.5824,
      "step": 736
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019049077469364048,
      "loss": 0.5773,
      "step": 737
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019045972056419577,
      "loss": 0.5726,
      "step": 738
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000190428618350597,
      "loss": 0.5834,
      "step": 739
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019039746806937663,
      "loss": 0.5847,
      "step": 740
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019036626973709274,
      "loss": 0.5692,
      "step": 741
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019033502337032895,
      "loss": 0.5882,
      "step": 742
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019030372898569425,
      "loss": 0.5868,
      "step": 743
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019027238659982332,
      "loss": 0.5828,
      "step": 744
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001902409962293763,
      "loss": 0.5802,
      "step": 745
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001902095578910389,
      "loss": 0.5883,
      "step": 746
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00019017807160152218,
      "loss": 0.5713,
      "step": 747
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00019014653737756286,
      "loss": 0.578,
      "step": 748
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000190114955235923,
      "loss": 0.5754,
      "step": 749
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00019008332519339022,
      "loss": 0.5751,
      "step": 750
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00019005164726677753,
      "loss": 0.5825,
      "step": 751
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001900199214729235,
      "loss": 0.5677,
      "step": 752
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000189988147828692,
      "loss": 0.576,
      "step": 753
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001899563263509725,
      "loss": 0.5706,
      "step": 754
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00018992445705667973,
      "loss": 0.5692,
      "step": 755
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00018989253996275394,
      "loss": 0.5667,
      "step": 756
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00018986057508616074,
      "loss": 0.5859,
      "step": 757
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00018982856244389125,
      "loss": 0.5816,
      "step": 758
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001897965020529618,
      "loss": 0.5781,
      "step": 759
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00018976439393041423,
      "loss": 0.5746,
      "step": 760
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001897322380933157,
      "loss": 0.5744,
      "step": 761
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00018970003455875877,
      "loss": 0.5769,
      "step": 762
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018966778334386135,
      "loss": 0.5634,
      "step": 763
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001896354844657666,
      "loss": 0.5718,
      "step": 764
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018960313794164318,
      "loss": 0.568,
      "step": 765
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018957074378868493,
      "loss": 0.5637,
      "step": 766
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001895383020241111,
      "loss": 0.5762,
      "step": 767
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001895058126651662,
      "loss": 0.586,
      "step": 768
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018947327572912004,
      "loss": 0.5753,
      "step": 769
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018944069123326773,
      "loss": 0.5629,
      "step": 770
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018940805919492968,
      "loss": 0.5691,
      "step": 771
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018937537963145155,
      "loss": 0.578,
      "step": 772
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018934265256020428,
      "loss": 0.5724,
      "step": 773
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018930987799858407,
      "loss": 0.5785,
      "step": 774
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018927705596401224,
      "loss": 0.5766,
      "step": 775
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018924418647393556,
      "loss": 0.5713,
      "step": 776
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00018921126954582587,
      "loss": 0.564,
      "step": 777
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018917830519718027,
      "loss": 0.5757,
      "step": 778
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018914529344552105,
      "loss": 0.5628,
      "step": 779
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018911223430839573,
      "loss": 0.5704,
      "step": 780
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000189079127803377,
      "loss": 0.5782,
      "step": 781
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018904597394806273,
      "loss": 0.5937,
      "step": 782
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018901277276007594,
      "loss": 0.5752,
      "step": 783
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018897952425706488,
      "loss": 0.5823,
      "step": 784
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018894622845670283,
      "loss": 0.5877,
      "step": 785
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0001889128853766883,
      "loss": 0.5737,
      "step": 786
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018887949503474492,
      "loss": 0.5839,
      "step": 787
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018884605744862146,
      "loss": 0.5857,
      "step": 788
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018881257263609168,
      "loss": 0.572,
      "step": 789
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018877904061495468,
      "loss": 0.5826,
      "step": 790
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018874546140303441,
      "loss": 0.5681,
      "step": 791
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018871183501818005,
      "loss": 0.5763,
      "step": 792
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0001886781614782658,
      "loss": 0.59,
      "step": 793
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018864444080119092,
      "loss": 0.5632,
      "step": 794
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0001886106730048798,
      "loss": 0.5621,
      "step": 795
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018857685810728177,
      "loss": 0.5829,
      "step": 796
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018854299612637126,
      "loss": 0.5612,
      "step": 797
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018850908708014768,
      "loss": 0.5733,
      "step": 798
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018847513098663556,
      "loss": 0.5792,
      "step": 799
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018844112786388432,
      "loss": 0.5834,
      "step": 800
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018840707772996846,
      "loss": 0.5689,
      "step": 801
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018837298060298742,
      "loss": 0.5767,
      "step": 802
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018833883650106565,
      "loss": 0.5758,
      "step": 803
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018830464544235254,
      "loss": 0.5633,
      "step": 804
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018827040744502248,
      "loss": 0.5744,
      "step": 805
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018823612252727472,
      "loss": 0.5824,
      "step": 806
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0001882017907073336,
      "loss": 0.5751,
      "step": 807
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00018816741200344827,
      "loss": 0.5677,
      "step": 808
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00018813298643389286,
      "loss": 0.5793,
      "step": 809
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00018809851401696637,
      "loss": 0.5694,
      "step": 810
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00018806399477099274,
      "loss": 0.585,
      "step": 811
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00018802942871432076,
      "loss": 0.5907,
      "step": 812
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001879948158653242,
      "loss": 0.5865,
      "step": 813
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001879601562424015,
      "loss": 0.5809,
      "step": 814
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00018792544986397623,
      "loss": 0.5858,
      "step": 815
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001878906967484966,
      "loss": 0.5685,
      "step": 816
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00018785589691443574,
      "loss": 0.5792,
      "step": 817
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00018782105038029164,
      "loss": 0.5873,
      "step": 818
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001877861571645871,
      "loss": 0.5726,
      "step": 819
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00018775121728586966,
      "loss": 0.585,
      "step": 820
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001877162307627118,
      "loss": 0.5837,
      "step": 821
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018768119761371064,
      "loss": 0.5532,
      "step": 822
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018764611785748823,
      "loss": 0.5764,
      "step": 823
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018761099151269131,
      "loss": 0.5606,
      "step": 824
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018757581859799136,
      "loss": 0.5823,
      "step": 825
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001875405991320847,
      "loss": 0.5639,
      "step": 826
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018750533313369232,
      "loss": 0.5705,
      "step": 827
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018747002062156,
      "loss": 0.57,
      "step": 828
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018743466161445823,
      "loss": 0.5666,
      "step": 829
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018739925613118214,
      "loss": 0.5875,
      "step": 830
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001873638041905517,
      "loss": 0.5715,
      "step": 831
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018732830581141147,
      "loss": 0.5675,
      "step": 832
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018729276101263074,
      "loss": 0.5943,
      "step": 833
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018725716981310344,
      "loss": 0.5765,
      "step": 834
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018722153223174818,
      "loss": 0.5648,
      "step": 835
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018718584828750828,
      "loss": 0.58,
      "step": 836
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001871501179993516,
      "loss": 0.5851,
      "step": 837
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018711434138627073,
      "loss": 0.5801,
      "step": 838
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001870785184672828,
      "loss": 0.5812,
      "step": 839
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018704264926142967,
      "loss": 0.5657,
      "step": 840
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018700673378777762,
      "loss": 0.5755,
      "step": 841
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018697077206541774,
      "loss": 0.5724,
      "step": 842
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018693476411346554,
      "loss": 0.5783,
      "step": 843
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001868987099510612,
      "loss": 0.5919,
      "step": 844
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001868626095973694,
      "loss": 0.562,
      "step": 845
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001868264630715794,
      "loss": 0.579,
      "step": 846
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018679027039290497,
      "loss": 0.5678,
      "step": 847
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018675403158058452,
      "loss": 0.5623,
      "step": 848
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018671774665388085,
      "loss": 0.5601,
      "step": 849
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018668141563208133,
      "loss": 0.5886,
      "step": 850
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00018664503853449787,
      "loss": 0.5678,
      "step": 851
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018660861538046683,
      "loss": 0.5631,
      "step": 852
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018657214618934899,
      "loss": 0.5661,
      "step": 853
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018653563098052972,
      "loss": 0.5753,
      "step": 854
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001864990697734188,
      "loss": 0.5777,
      "step": 855
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018646246258745042,
      "loss": 0.5717,
      "step": 856
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018642580944208328,
      "loss": 0.5572,
      "step": 857
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018638911035680043,
      "loss": 0.5613,
      "step": 858
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018635236535110948,
      "loss": 0.5856,
      "step": 859
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018631557444454228,
      "loss": 0.5827,
      "step": 860
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018627873765665514,
      "loss": 0.5663,
      "step": 861
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001862418550070288,
      "loss": 0.5867,
      "step": 862
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018620492651526838,
      "loss": 0.574,
      "step": 863
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001861679522010033,
      "loss": 0.5692,
      "step": 864
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018613093208388738,
      "loss": 0.568,
      "step": 865
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018609386618359877,
      "loss": 0.5784,
      "step": 866
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018605675451984004,
      "loss": 0.5726,
      "step": 867
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018601959711233792,
      "loss": 0.5912,
      "step": 868
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018598239398084362,
      "loss": 0.5693,
      "step": 869
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018594514514513254,
      "loss": 0.5739,
      "step": 870
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001859078506250044,
      "loss": 0.587,
      "step": 871
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018587051044028328,
      "loss": 0.5682,
      "step": 872
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001858331246108174,
      "loss": 0.5766,
      "step": 873
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018579569315647942,
      "loss": 0.5796,
      "step": 874
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018575821609716607,
      "loss": 0.5746,
      "step": 875
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018572069345279844,
      "loss": 0.573,
      "step": 876
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018568312524332178,
      "loss": 0.5871,
      "step": 877
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018564551148870563,
      "loss": 0.5662,
      "step": 878
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018560785220894365,
      "loss": 0.5798,
      "step": 879
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018557014742405379,
      "loss": 0.5537,
      "step": 880
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00018553239715407812,
      "loss": 0.5785,
      "step": 881
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018549460141908296,
      "loss": 0.5758,
      "step": 882
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018545676023915873,
      "loss": 0.5709,
      "step": 883
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018541887363442,
      "loss": 0.5761,
      "step": 884
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018538094162500558,
      "loss": 0.5738,
      "step": 885
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001853429642310783,
      "loss": 0.567,
      "step": 886
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018530494147282522,
      "loss": 0.5674,
      "step": 887
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018526687337045737,
      "loss": 0.5906,
      "step": 888
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018522875994421005,
      "loss": 0.5784,
      "step": 889
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018519060121434256,
      "loss": 0.5696,
      "step": 890
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018515239720113825,
      "loss": 0.5812,
      "step": 891
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018511414792490464,
      "loss": 0.5954,
      "step": 892
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001850758534059732,
      "loss": 0.5735,
      "step": 893
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018503751366469957,
      "loss": 0.577,
      "step": 894
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018499912872146332,
      "loss": 0.5798,
      "step": 895
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00018496069859666808,
      "loss": 0.5645,
      "step": 896
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018492222331074155,
      "loss": 0.5728,
      "step": 897
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018488370288413536,
      "loss": 0.5814,
      "step": 898
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018484513733732518,
      "loss": 0.5641,
      "step": 899
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018480652669081063,
      "loss": 0.5668,
      "step": 900
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001847678709651153,
      "loss": 0.5776,
      "step": 901
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018472917018078685,
      "loss": 0.5601,
      "step": 902
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018469042435839675,
      "loss": 0.5754,
      "step": 903
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018465163351854047,
      "loss": 0.5727,
      "step": 904
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001846127976818374,
      "loss": 0.5803,
      "step": 905
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018457391686893084,
      "loss": 0.5598,
      "step": 906
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000184534991100488,
      "loss": 0.5756,
      "step": 907
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001844960203972001,
      "loss": 0.5701,
      "step": 908
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018445700477978205,
      "loss": 0.5814,
      "step": 909
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00018441794426897273,
      "loss": 0.5755,
      "step": 910
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001843788388855349,
      "loss": 0.5669,
      "step": 911
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00018433968865025512,
      "loss": 0.5575,
      "step": 912
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001843004935839439,
      "loss": 0.5685,
      "step": 913
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00018426125370743542,
      "loss": 0.5773,
      "step": 914
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001842219690415878,
      "loss": 0.5613,
      "step": 915
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001841826396072829,
      "loss": 0.5777,
      "step": 916
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00018414326542542649,
      "loss": 0.5653,
      "step": 917
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00018410384651694792,
      "loss": 0.565,
      "step": 918
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00018406438290280054,
      "loss": 0.5555,
      "step": 919
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001840248746039613,
      "loss": 0.5601,
      "step": 920
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00018398532164143097,
      "loss": 0.5796,
      "step": 921
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001839457240362341,
      "loss": 0.5726,
      "step": 922
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001839060818094189,
      "loss": 0.5678,
      "step": 923
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00018386639498205727,
      "loss": 0.5846,
      "step": 924
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00018382666357524493,
      "loss": 0.569,
      "step": 925
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018378688761010124,
      "loss": 0.566,
      "step": 926
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018374706710776919,
      "loss": 0.5653,
      "step": 927
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018370720208941554,
      "loss": 0.5755,
      "step": 928
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018366729257623067,
      "loss": 0.5755,
      "step": 929
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018362733858942858,
      "loss": 0.5758,
      "step": 930
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018358734015024697,
      "loss": 0.5645,
      "step": 931
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0001835472972799471,
      "loss": 0.5764,
      "step": 932
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018350720999981391,
      "loss": 0.5633,
      "step": 933
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018346707833115594,
      "loss": 0.5792,
      "step": 934
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018342690229530527,
      "loss": 0.5845,
      "step": 935
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018338668191361762,
      "loss": 0.5725,
      "step": 936
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018334641720747227,
      "loss": 0.5588,
      "step": 937
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018330610819827206,
      "loss": 0.5788,
      "step": 938
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018326575490744333,
      "loss": 0.5354,
      "step": 939
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00018322535735643605,
      "loss": 0.5722,
      "step": 940
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018318491556672362,
      "loss": 0.5668,
      "step": 941
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018314442955980303,
      "loss": 0.5799,
      "step": 942
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018310389935719478,
      "loss": 0.5694,
      "step": 943
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018306332498044278,
      "loss": 0.5706,
      "step": 944
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018302270645111446,
      "loss": 0.5711,
      "step": 945
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018298204379080076,
      "loss": 0.5577,
      "step": 946
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018294133702111603,
      "loss": 0.5861,
      "step": 947
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018290058616369815,
      "loss": 0.5669,
      "step": 948
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018285979124020823,
      "loss": 0.5681,
      "step": 949
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001828189522723311,
      "loss": 0.5918,
      "step": 950
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001827780692817747,
      "loss": 0.5627,
      "step": 951
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018273714229027067,
      "loss": 0.5689,
      "step": 952
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001826961713195737,
      "loss": 0.5781,
      "step": 953
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001826551563914622,
      "loss": 0.5575,
      "step": 954
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00018261409752773773,
      "loss": 0.5708,
      "step": 955
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0001825729947502252,
      "loss": 0.5579,
      "step": 956
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018253184808077303,
      "loss": 0.5723,
      "step": 957
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018249065754125278,
      "loss": 0.5764,
      "step": 958
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018244942315355948,
      "loss": 0.5682,
      "step": 959
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018240814493961134,
      "loss": 0.5626,
      "step": 960
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018236682292134995,
      "loss": 0.5666,
      "step": 961
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018232545712074013,
      "loss": 0.5717,
      "step": 962
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0001822840475597701,
      "loss": 0.5627,
      "step": 963
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018224259426045114,
      "loss": 0.5834,
      "step": 964
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018220109724481796,
      "loss": 0.5657,
      "step": 965
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018215955653492837,
      "loss": 0.5695,
      "step": 966
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0001821179721528635,
      "loss": 0.5605,
      "step": 967
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018207634412072764,
      "loss": 0.5632,
      "step": 968
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018203467246064834,
      "loss": 0.5811,
      "step": 969
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00018199295719477628,
      "loss": 0.5699,
      "step": 970
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018195119834528534,
      "loss": 0.5635,
      "step": 971
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018190939593437256,
      "loss": 0.5835,
      "step": 972
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0001818675499842582,
      "loss": 0.5927,
      "step": 973
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018182566051718556,
      "loss": 0.5836,
      "step": 974
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018178372755542113,
      "loss": 0.5748,
      "step": 975
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018174175112125452,
      "loss": 0.5667,
      "step": 976
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0001816997312369984,
      "loss": 0.5737,
      "step": 977
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018165766792498862,
      "loss": 0.5705,
      "step": 978
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018161556120758405,
      "loss": 0.5603,
      "step": 979
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018157341110716665,
      "loss": 0.5646,
      "step": 980
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018153121764614145,
      "loss": 0.5667,
      "step": 981
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0001814889808469365,
      "loss": 0.5733,
      "step": 982
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018144670073200292,
      "loss": 0.5579,
      "step": 983
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00018140437732381485,
      "loss": 0.5724,
      "step": 984
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018136201064486942,
      "loss": 0.5768,
      "step": 985
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018131960071768677,
      "loss": 0.5754,
      "step": 986
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018127714756481001,
      "loss": 0.566,
      "step": 987
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0001812346512088053,
      "loss": 0.5791,
      "step": 988
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0001811921116722617,
      "loss": 0.5682,
      "step": 989
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018114952897779123,
      "loss": 0.5733,
      "step": 990
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018110690314802885,
      "loss": 0.5772,
      "step": 991
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018106423420563247,
      "loss": 0.566,
      "step": 992
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018102152217328284,
      "loss": 0.5743,
      "step": 993
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0001809787670736838,
      "loss": 0.5742,
      "step": 994
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018093596892956187,
      "loss": 0.5678,
      "step": 995
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018089312776366657,
      "loss": 0.5688,
      "step": 996
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018085024359877023,
      "loss": 0.5661,
      "step": 997
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0001808073164576681,
      "loss": 0.5635,
      "step": 998
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00018076434636317822,
      "loss": 0.5902,
      "step": 999
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0001807213333381415,
      "loss": 0.5727,
      "step": 1000
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018067827740542164,
      "loss": 0.5557,
      "step": 1001
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018063517858790516,
      "loss": 0.5617,
      "step": 1002
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018059203690850133,
      "loss": 0.5723,
      "step": 1003
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018054885239014232,
      "loss": 0.5629,
      "step": 1004
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018050562505578298,
      "loss": 0.5619,
      "step": 1005
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0001804623549284009,
      "loss": 0.5694,
      "step": 1006
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018041904203099652,
      "loss": 0.5724,
      "step": 1007
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018037568638659286,
      "loss": 0.5728,
      "step": 1008
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018033228801823584,
      "loss": 0.5646,
      "step": 1009
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018028884694899395,
      "loss": 0.5765,
      "step": 1010
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018024536320195844,
      "loss": 0.5751,
      "step": 1011
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018020183680024325,
      "loss": 0.5611,
      "step": 1012
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018015826776698494,
      "loss": 0.5733,
      "step": 1013
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00018011465612534285,
      "loss": 0.5518,
      "step": 1014
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00018007100189849876,
      "loss": 0.5632,
      "step": 1015
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00018002730510965735,
      "loss": 0.5792,
      "step": 1016
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017998356578204566,
      "loss": 0.5791,
      "step": 1017
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0001799397839389136,
      "loss": 0.5714,
      "step": 1018
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017989595960353345,
      "loss": 0.5776,
      "step": 1019
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017985209279920022,
      "loss": 0.5703,
      "step": 1020
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017980818354923145,
      "loss": 0.5558,
      "step": 1021
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017976423187696718,
      "loss": 0.5601,
      "step": 1022
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017972023780577016,
      "loss": 0.5703,
      "step": 1023
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017967620135902554,
      "loss": 0.5576,
      "step": 1024
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017963212256014098,
      "loss": 0.5615,
      "step": 1025
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017958800143254682,
      "loss": 0.565,
      "step": 1026
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0001795438379996957,
      "loss": 0.5645,
      "step": 1027
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0001794996322850628,
      "loss": 0.5738,
      "step": 1028
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017945538431214593,
      "loss": 0.5812,
      "step": 1029
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017941109410446514,
      "loss": 0.5648,
      "step": 1030
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017936676168556307,
      "loss": 0.5532,
      "step": 1031
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0001793223870790048,
      "loss": 0.5798,
      "step": 1032
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017927797030837768,
      "loss": 0.5853,
      "step": 1033
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0001792335113972917,
      "loss": 0.5634,
      "step": 1034
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017918901036937908,
      "loss": 0.5722,
      "step": 1035
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017914446724829453,
      "loss": 0.5659,
      "step": 1036
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017909988205771504,
      "loss": 0.5826,
      "step": 1037
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017905525482134,
      "loss": 0.5634,
      "step": 1038
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0001790105855628912,
      "loss": 0.5629,
      "step": 1039
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017896587430611273,
      "loss": 0.5544,
      "step": 1040
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000178921121074771,
      "loss": 0.5693,
      "step": 1041
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017887632589265467,
      "loss": 0.5622,
      "step": 1042
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017883148878357482,
      "loss": 0.5702,
      "step": 1043
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017878660977136478,
      "loss": 0.5728,
      "step": 1044
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017874168887988012,
      "loss": 0.5609,
      "step": 1045
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017869672613299864,
      "loss": 0.5608,
      "step": 1046
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017865172155462044,
      "loss": 0.5632,
      "step": 1047
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017860667516866793,
      "loss": 0.5665,
      "step": 1048
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017856158699908553,
      "loss": 0.5553,
      "step": 1049
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0001785164570698401,
      "loss": 0.5773,
      "step": 1050
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017847128540492054,
      "loss": 0.5832,
      "step": 1051
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017842607202833805,
      "loss": 0.5678,
      "step": 1052
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017838081696412592,
      "loss": 0.5749,
      "step": 1053
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0001783355202363396,
      "loss": 0.5704,
      "step": 1054
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017829018186905674,
      "loss": 0.5682,
      "step": 1055
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017824480188637708,
      "loss": 0.5701,
      "step": 1056
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00017819938031242248,
      "loss": 0.5702,
      "step": 1057
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000178153917171337,
      "loss": 0.5444,
      "step": 1058
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017810841248728663,
      "loss": 0.556,
      "step": 1059
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017806286628445955,
      "loss": 0.5605,
      "step": 1060
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000178017278587066,
      "loss": 0.5691,
      "step": 1061
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0001779716494193383,
      "loss": 0.5554,
      "step": 1062
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017792597880553071,
      "loss": 0.5767,
      "step": 1063
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017788026676991963,
      "loss": 0.5604,
      "step": 1064
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017783451333680344,
      "loss": 0.5523,
      "step": 1065
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017778871853050255,
      "loss": 0.5648,
      "step": 1066
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017774288237535924,
      "loss": 0.5658,
      "step": 1067
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000177697004895738,
      "loss": 0.5618,
      "step": 1068
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017765108611602506,
      "loss": 0.5611,
      "step": 1069
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0001776051260606287,
      "loss": 0.5659,
      "step": 1070
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0001775591247539792,
      "loss": 0.5738,
      "step": 1071
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017751308222052864,
      "loss": 0.571,
      "step": 1072
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00017746699848475105,
      "loss": 0.5683,
      "step": 1073
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017742087357114244,
      "loss": 0.5732,
      "step": 1074
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017737470750422068,
      "loss": 0.5746,
      "step": 1075
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017732850030852544,
      "loss": 0.5705,
      "step": 1076
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001772822520086183,
      "loss": 0.5757,
      "step": 1077
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001772359626290827,
      "loss": 0.5655,
      "step": 1078
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017718963219452393,
      "loss": 0.573,
      "step": 1079
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001771432607295691,
      "loss": 0.5771,
      "step": 1080
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017709684825886702,
      "loss": 0.5748,
      "step": 1081
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017705039480708846,
      "loss": 0.57,
      "step": 1082
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017700390039892595,
      "loss": 0.5579,
      "step": 1083
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017695736505909362,
      "loss": 0.5605,
      "step": 1084
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017691078881232753,
      "loss": 0.5664,
      "step": 1085
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017686417168338544,
      "loss": 0.5697,
      "step": 1086
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017681751369704685,
      "loss": 0.5738,
      "step": 1087
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017677081487811296,
      "loss": 0.5817,
      "step": 1088
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017672407525140665,
      "loss": 0.5644,
      "step": 1089
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001766772948417725,
      "loss": 0.5644,
      "step": 1090
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017663047367407684,
      "loss": 0.5559,
      "step": 1091
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001765836117732076,
      "loss": 0.5704,
      "step": 1092
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017653670916407434,
      "loss": 0.5485,
      "step": 1093
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017648976587160832,
      "loss": 0.5545,
      "step": 1094
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001764427819207624,
      "loss": 0.567,
      "step": 1095
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017639575733651103,
      "loss": 0.5528,
      "step": 1096
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001763486921438503,
      "loss": 0.5598,
      "step": 1097
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017630158636779784,
      "loss": 0.5694,
      "step": 1098
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001762544400333929,
      "loss": 0.5763,
      "step": 1099
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017620725316569625,
      "loss": 0.5635,
      "step": 1100
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017616002578979022,
      "loss": 0.5679,
      "step": 1101
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017611275793077868,
      "loss": 0.5685,
      "step": 1102
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00017606544961378703,
      "loss": 0.567,
      "step": 1103
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017601810086396213,
      "loss": 0.5719,
      "step": 1104
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017597071170647236,
      "loss": 0.5634,
      "step": 1105
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017592328216650756,
      "loss": 0.5654,
      "step": 1106
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001758758122692791,
      "loss": 0.5658,
      "step": 1107
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017582830204001973,
      "loss": 0.5709,
      "step": 1108
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017578075150398366,
      "loss": 0.5681,
      "step": 1109
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017573316068644656,
      "loss": 0.5618,
      "step": 1110
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017568552961270544,
      "loss": 0.5679,
      "step": 1111
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001756378583080788,
      "loss": 0.5531,
      "step": 1112
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017559014679790642,
      "loss": 0.5665,
      "step": 1113
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017554239510754954,
      "loss": 0.5696,
      "step": 1114
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001754946032623907,
      "loss": 0.566,
      "step": 1115
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001754467712878338,
      "loss": 0.5737,
      "step": 1116
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017539889920930413,
      "loss": 0.5593,
      "step": 1117
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00017535098705224824,
      "loss": 0.5637,
      "step": 1118
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00017530303484213392,
      "loss": 0.5602,
      "step": 1119
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001752550426044504,
      "loss": 0.5477,
      "step": 1120
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001752070103647081,
      "loss": 0.549,
      "step": 1121
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00017515893814843865,
      "loss": 0.5658,
      "step": 1122
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000175110825981195,
      "loss": 0.5706,
      "step": 1123
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00017506267388855143,
      "loss": 0.5749,
      "step": 1124
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001750144818961032,
      "loss": 0.5565,
      "step": 1125
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000174966250029467,
      "loss": 0.5575,
      "step": 1126
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00017491797831428063,
      "loss": 0.5606,
      "step": 1127
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00017486966677620304,
      "loss": 0.5607,
      "step": 1128
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001748213154409144,
      "loss": 0.5814,
      "step": 1129
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00017477292433411602,
      "loss": 0.5621,
      "step": 1130
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00017472449348153037,
      "loss": 0.5553,
      "step": 1131
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000174676022908901,
      "loss": 0.5677,
      "step": 1132
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00017462751264199258,
      "loss": 0.573,
      "step": 1133
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00017457896270659093,
      "loss": 0.5694,
      "step": 1134
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001745303731285029,
      "loss": 0.5516,
      "step": 1135
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00017448174393355643,
      "loss": 0.5524,
      "step": 1136
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00017443307514760056,
      "loss": 0.5616,
      "step": 1137
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00017438436679650534,
      "loss": 0.5731,
      "step": 1138
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001743356189061618,
      "loss": 0.5556,
      "step": 1139
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00017428683150248208,
      "loss": 0.5764,
      "step": 1140
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00017423800461139924,
      "loss": 0.5481,
      "step": 1141
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00017418913825886746,
      "loss": 0.5506,
      "step": 1142
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001741402324708617,
      "loss": 0.5816,
      "step": 1143
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00017409128727337802,
      "loss": 0.5594,
      "step": 1144
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001740423026924334,
      "loss": 0.5537,
      "step": 1145
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001739932787540658,
      "loss": 0.5716,
      "step": 1146
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.000173944215484334,
      "loss": 0.5548,
      "step": 1147
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00017389511290931775,
      "loss": 0.5653,
      "step": 1148
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001738459710551177,
      "loss": 0.5718,
      "step": 1149
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001737967899478553,
      "loss": 0.5536,
      "step": 1150
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00017374756961367296,
      "loss": 0.5794,
      "step": 1151
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00017369831007873395,
      "loss": 0.561,
      "step": 1152
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00017364901136922227,
      "loss": 0.5698,
      "step": 1153
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001735996735113428,
      "loss": 0.5705,
      "step": 1154
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001735502965313213,
      "loss": 0.5531,
      "step": 1155
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001735008804554042,
      "loss": 0.5498,
      "step": 1156
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00017345142530985887,
      "loss": 0.5741,
      "step": 1157
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00017340193112097321,
      "loss": 0.5697,
      "step": 1158
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00017335239791505615,
      "loss": 0.5832,
      "step": 1159
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00017330282571843714,
      "loss": 0.5551,
      "step": 1160
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00017325321455746644,
      "loss": 0.5651,
      "step": 1161
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001732035644585151,
      "loss": 0.5583,
      "step": 1162
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00017315387544797467,
      "loss": 0.5624,
      "step": 1163
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001731041475522576,
      "loss": 0.5637,
      "step": 1164
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00017305438079779685,
      "loss": 0.5614,
      "step": 1165
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00017300457521104612,
      "loss": 0.5553,
      "step": 1166
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00017295473081847975,
      "loss": 0.5636,
      "step": 1167
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00017290484764659268,
      "loss": 0.5693,
      "step": 1168
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00017285492572190046,
      "loss": 0.5647,
      "step": 1169
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001728049650709392,
      "loss": 0.5688,
      "step": 1170
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00017275496572026578,
      "loss": 0.5635,
      "step": 1171
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00017270492769645736,
      "loss": 0.5611,
      "step": 1172
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001726548510261119,
      "loss": 0.5656,
      "step": 1173
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001726047357358478,
      "loss": 0.5546,
      "step": 1174
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.000172554581852304,
      "loss": 0.5595,
      "step": 1175
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00017250438940213993,
      "loss": 0.5652,
      "step": 1176
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001724541584120356,
      "loss": 0.5589,
      "step": 1177
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001724038889086914,
      "loss": 0.5736,
      "step": 1178
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001723535809188283,
      "loss": 0.5698,
      "step": 1179
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017230323446918762,
      "loss": 0.5632,
      "step": 1180
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017225284958653122,
      "loss": 0.5636,
      "step": 1181
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001722024262976413,
      "loss": 0.5661,
      "step": 1182
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017215196462932054,
      "loss": 0.5532,
      "step": 1183
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017210146460839204,
      "loss": 0.5698,
      "step": 1184
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001720509262616992,
      "loss": 0.5755,
      "step": 1185
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017200034961610587,
      "loss": 0.5632,
      "step": 1186
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001719497346984962,
      "loss": 0.5625,
      "step": 1187
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017189908153577473,
      "loss": 0.5599,
      "step": 1188
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017184839015486629,
      "loss": 0.5675,
      "step": 1189
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017179766058271606,
      "loss": 0.5717,
      "step": 1190
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017174689284628954,
      "loss": 0.5695,
      "step": 1191
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00017169608697257242,
      "loss": 0.5559,
      "step": 1192
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017164524298857078,
      "loss": 0.5601,
      "step": 1193
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017159436092131086,
      "loss": 0.565,
      "step": 1194
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017154344079783922,
      "loss": 0.5566,
      "step": 1195
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0001714924826452226,
      "loss": 0.5724,
      "step": 1196
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017144148649054795,
      "loss": 0.5599,
      "step": 1197
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017139045236092248,
      "loss": 0.5617,
      "step": 1198
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0001713393802834735,
      "loss": 0.5619,
      "step": 1199
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017128827028534856,
      "loss": 0.5581,
      "step": 1200
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017123712239371533,
      "loss": 0.5571,
      "step": 1201
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017118593663576168,
      "loss": 0.5562,
      "step": 1202
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0001711347130386955,
      "loss": 0.5616,
      "step": 1203
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0001710834516297449,
      "loss": 0.5443,
      "step": 1204
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017103215243615804,
      "loss": 0.5527,
      "step": 1205
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017098081548520315,
      "loss": 0.5637,
      "step": 1206
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017092944080416856,
      "loss": 0.5653,
      "step": 1207
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001708780284203627,
      "loss": 0.553,
      "step": 1208
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001708265783611139,
      "loss": 0.5748,
      "step": 1209
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001707750906537706,
      "loss": 0.5584,
      "step": 1210
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017072356532570138,
      "loss": 0.5631,
      "step": 1211
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017067200240429456,
      "loss": 0.5555,
      "step": 1212
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017062040191695866,
      "loss": 0.5718,
      "step": 1213
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017056876389112203,
      "loss": 0.5731,
      "step": 1214
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017051708835423302,
      "loss": 0.5615,
      "step": 1215
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017046537533376,
      "loss": 0.5591,
      "step": 1216
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017041362485719112,
      "loss": 0.5444,
      "step": 1217
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001703618369520345,
      "loss": 0.5534,
      "step": 1218
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017031001164581828,
      "loss": 0.5605,
      "step": 1219
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017025814896609026,
      "loss": 0.5664,
      "step": 1220
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00017020624894041825,
      "loss": 0.5601,
      "step": 1221
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00017015431159638988,
      "loss": 0.5608,
      "step": 1222
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001701023369616126,
      "loss": 0.5686,
      "step": 1223
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00017005032506371367,
      "loss": 0.5716,
      "step": 1224
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016999827593034024,
      "loss": 0.5428,
      "step": 1225
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016994618958915915,
      "loss": 0.5484,
      "step": 1226
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016989406606785707,
      "loss": 0.582,
      "step": 1227
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001698419053941404,
      "loss": 0.5506,
      "step": 1228
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001697897075957354,
      "loss": 0.5712,
      "step": 1229
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016973747270038786,
      "loss": 0.5555,
      "step": 1230
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001696852007358635,
      "loss": 0.5552,
      "step": 1231
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016963289172994758,
      "loss": 0.5702,
      "step": 1232
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016958054571044516,
      "loss": 0.5647,
      "step": 1233
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016952816270518096,
      "loss": 0.5694,
      "step": 1234
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00016947574274199924,
      "loss": 0.5599,
      "step": 1235
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001694232858487641,
      "loss": 0.5817,
      "step": 1236
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016937079205335914,
      "loss": 0.557,
      "step": 1237
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016931826138368757,
      "loss": 0.5666,
      "step": 1238
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016926569386767226,
      "loss": 0.5669,
      "step": 1239
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016921308953325563,
      "loss": 0.5497,
      "step": 1240
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016916044840839973,
      "loss": 0.5628,
      "step": 1241
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016910777052108603,
      "loss": 0.5366,
      "step": 1242
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0001690550558993157,
      "loss": 0.5621,
      "step": 1243
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016900230457110937,
      "loss": 0.5553,
      "step": 1244
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016894951656450715,
      "loss": 0.5617,
      "step": 1245
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016889669190756868,
      "loss": 0.5586,
      "step": 1246
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016884383062837306,
      "loss": 0.5624,
      "step": 1247
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016879093275501888,
      "loss": 0.5567,
      "step": 1248
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016873799831562417,
      "loss": 0.566,
      "step": 1249
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016868502733832644,
      "loss": 0.5352,
      "step": 1250
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00016863201985128256,
      "loss": 0.5678,
      "step": 1251
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001685789758826688,
      "loss": 0.5807,
      "step": 1252
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016852589546068088,
      "loss": 0.5521,
      "step": 1253
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016847277861353389,
      "loss": 0.5743,
      "step": 1254
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001684196253694622,
      "loss": 0.5645,
      "step": 1255
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016836643575671961,
      "loss": 0.5685,
      "step": 1256
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016831320980357925,
      "loss": 0.554,
      "step": 1257
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016825994753833352,
      "loss": 0.5558,
      "step": 1258
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001682066489892942,
      "loss": 0.56,
      "step": 1259
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016815331418479224,
      "loss": 0.5554,
      "step": 1260
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016809994315317796,
      "loss": 0.5621,
      "step": 1261
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001680465359228209,
      "loss": 0.5636,
      "step": 1262
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001679930925221098,
      "loss": 0.5506,
      "step": 1263
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016793961297945277,
      "loss": 0.5656,
      "step": 1264
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016788609732327694,
      "loss": 0.554,
      "step": 1265
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00016783254558202875,
      "loss": 0.547,
      "step": 1266
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00016777895778417382,
      "loss": 0.5553,
      "step": 1267
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00016772533395819692,
      "loss": 0.5527,
      "step": 1268
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00016767167413260197,
      "loss": 0.5606,
      "step": 1269
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00016761797833591196,
      "loss": 0.5729,
      "step": 1270
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001675642465966691,
      "loss": 0.5667,
      "step": 1271
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001675104789434347,
      "loss": 0.5522,
      "step": 1272
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001674566754047891,
      "loss": 0.5515,
      "step": 1273
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00016740283600933176,
      "loss": 0.5547,
      "step": 1274
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00016734896078568116,
      "loss": 0.5538,
      "step": 1275
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00016729504976247482,
      "loss": 0.559,
      "step": 1276
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00016724110296836942,
      "loss": 0.5561,
      "step": 1277
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001671871204320404,
      "loss": 0.5489,
      "step": 1278
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001671331021821825,
      "loss": 0.5719,
      "step": 1279
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00016707904824750922,
      "loss": 0.5793,
      "step": 1280
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001670249586567531,
      "loss": 0.5497,
      "step": 1281
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0001669708334386656,
      "loss": 0.5683,
      "step": 1282
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016691667262201727,
      "loss": 0.5634,
      "step": 1283
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0001668624762355974,
      "loss": 0.5531,
      "step": 1284
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016680824430821418,
      "loss": 0.5633,
      "step": 1285
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016675397686869488,
      "loss": 0.552,
      "step": 1286
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016669967394588545,
      "loss": 0.5485,
      "step": 1287
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016664533556865084,
      "loss": 0.5653,
      "step": 1288
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016659096176587475,
      "loss": 0.5694,
      "step": 1289
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016653655256645977,
      "loss": 0.5449,
      "step": 1290
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016648210799932728,
      "loss": 0.5616,
      "step": 1291
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016642762809341744,
      "loss": 0.5663,
      "step": 1292
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016637311287768925,
      "loss": 0.5585,
      "step": 1293
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016631856238112045,
      "loss": 0.5732,
      "step": 1294
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00016626397663270748,
      "loss": 0.5529,
      "step": 1295
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016620935566146562,
      "loss": 0.5488,
      "step": 1296
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016615469949642883,
      "loss": 0.5627,
      "step": 1297
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016610000816664973,
      "loss": 0.5566,
      "step": 1298
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001660452817011997,
      "loss": 0.5572,
      "step": 1299
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016599052012916877,
      "loss": 0.5579,
      "step": 1300
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016593572347966558,
      "loss": 0.5631,
      "step": 1301
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016588089178181753,
      "loss": 0.5591,
      "step": 1302
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001658260250647705,
      "loss": 0.5529,
      "step": 1303
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001657711233576892,
      "loss": 0.5652,
      "step": 1304
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001657161866897567,
      "loss": 0.5716,
      "step": 1305
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016566121509017478,
      "loss": 0.5619,
      "step": 1306
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016560620858816378,
      "loss": 0.561,
      "step": 1307
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001655511672129626,
      "loss": 0.5648,
      "step": 1308
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016549609099382866,
      "loss": 0.5586,
      "step": 1309
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00016544097996003786,
      "loss": 0.563,
      "step": 1310
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016538583414088472,
      "loss": 0.5658,
      "step": 1311
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016533065356568206,
      "loss": 0.5634,
      "step": 1312
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001652754382637614,
      "loss": 0.5729,
      "step": 1313
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016522018826447255,
      "loss": 0.5673,
      "step": 1314
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016516490359718387,
      "loss": 0.5769,
      "step": 1315
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016510958429128206,
      "loss": 0.5491,
      "step": 1316
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001650542303761723,
      "loss": 0.5627,
      "step": 1317
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016499884188127815,
      "loss": 0.5691,
      "step": 1318
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001649434188360415,
      "loss": 0.5648,
      "step": 1319
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016488796126992267,
      "loss": 0.548,
      "step": 1320
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001648324692124003,
      "loss": 0.5703,
      "step": 1321
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001647769426929714,
      "loss": 0.5589,
      "step": 1322
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016472138174115122,
      "loss": 0.5521,
      "step": 1323
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016466578638647342,
      "loss": 0.5675,
      "step": 1324
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00016461015665848985,
      "loss": 0.5795,
      "step": 1325
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016455449258677066,
      "loss": 0.5559,
      "step": 1326
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016449879420090427,
      "loss": 0.5491,
      "step": 1327
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0001644430615304974,
      "loss": 0.5603,
      "step": 1328
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0001643872946051748,
      "loss": 0.544,
      "step": 1329
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0001643314934545797,
      "loss": 0.5678,
      "step": 1330
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016427565810837332,
      "loss": 0.5489,
      "step": 1331
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0001642197885962351,
      "loss": 0.5584,
      "step": 1332
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016416388494786265,
      "loss": 0.5714,
      "step": 1333
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016410794719297177,
      "loss": 0.549,
      "step": 1334
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0001640519753612963,
      "loss": 0.5635,
      "step": 1335
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016399596948258834,
      "loss": 0.5679,
      "step": 1336
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016393992958661793,
      "loss": 0.549,
      "step": 1337
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016388385570317325,
      "loss": 0.5578,
      "step": 1338
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016382774786206054,
      "loss": 0.5593,
      "step": 1339
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00016377160609310415,
      "loss": 0.5656,
      "step": 1340
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001637154304261464,
      "loss": 0.5567,
      "step": 1341
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016365922089104764,
      "loss": 0.5596,
      "step": 1342
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001636029775176862,
      "loss": 0.5547,
      "step": 1343
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016354670033595848,
      "loss": 0.5663,
      "step": 1344
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001634903893757788,
      "loss": 0.5501,
      "step": 1345
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016343404466707937,
      "loss": 0.5535,
      "step": 1346
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016337766623981042,
      "loss": 0.5429,
      "step": 1347
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016332125412394012,
      "loss": 0.5474,
      "step": 1348
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016326480834945444,
      "loss": 0.5612,
      "step": 1349
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016320832894635743,
      "loss": 0.5499,
      "step": 1350
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016315181594467077,
      "loss": 0.5595,
      "step": 1351
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016309526937443423,
      "loss": 0.5513,
      "step": 1352
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001630386892657052,
      "loss": 0.5489,
      "step": 1353
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016298207564855916,
      "loss": 0.5585,
      "step": 1354
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016292542855308914,
      "loss": 0.5607,
      "step": 1355
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016286874800940616,
      "loss": 0.569,
      "step": 1356
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016281203404763895,
      "loss": 0.5652,
      "step": 1357
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016275528669793392,
      "loss": 0.559,
      "step": 1358
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016269850599045537,
      "loss": 0.5694,
      "step": 1359
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016264169195538526,
      "loss": 0.5618,
      "step": 1360
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016258484462292324,
      "loss": 0.5509,
      "step": 1361
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016252796402328672,
      "loss": 0.5609,
      "step": 1362
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0001624710501867108,
      "loss": 0.5504,
      "step": 1363
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016241410314344812,
      "loss": 0.5788,
      "step": 1364
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016235712292376914,
      "loss": 0.5543,
      "step": 1365
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0001623001095579618,
      "loss": 0.5684,
      "step": 1366
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016224306307633185,
      "loss": 0.558,
      "step": 1367
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00016218598350920243,
      "loss": 0.5574,
      "step": 1368
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0001621288708869144,
      "loss": 0.5666,
      "step": 1369
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016207172523982624,
      "loss": 0.546,
      "step": 1370
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016201454659831374,
      "loss": 0.5772,
      "step": 1371
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0001619573349927705,
      "loss": 0.5709,
      "step": 1372
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016190009045360753,
      "loss": 0.5614,
      "step": 1373
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0001618428130112533,
      "loss": 0.5683,
      "step": 1374
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016178550269615384,
      "loss": 0.5513,
      "step": 1375
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016172815953877266,
      "loss": 0.5526,
      "step": 1376
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016167078356959066,
      "loss": 0.5506,
      "step": 1377
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0001616133748191062,
      "loss": 0.5681,
      "step": 1378
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016155593331783515,
      "loss": 0.5509,
      "step": 1379
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016149845909631068,
      "loss": 0.5626,
      "step": 1380
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016144095218508344,
      "loss": 0.5454,
      "step": 1381
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016138341261472133,
      "loss": 0.5534,
      "step": 1382
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016132584041580978,
      "loss": 0.5726,
      "step": 1383
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00016126823561895143,
      "loss": 0.5639,
      "step": 1384
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001612105982547663,
      "loss": 0.5849,
      "step": 1385
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016115292835389172,
      "loss": 0.5542,
      "step": 1386
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016109522594698233,
      "loss": 0.5558,
      "step": 1387
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016103749106470997,
      "loss": 0.5533,
      "step": 1388
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016097972373776384,
      "loss": 0.5489,
      "step": 1389
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001609219239968504,
      "loss": 0.5696,
      "step": 1390
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001608640918726932,
      "loss": 0.5603,
      "step": 1391
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016080622739603312,
      "loss": 0.5616,
      "step": 1392
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016074833059762824,
      "loss": 0.559,
      "step": 1393
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016069040150825376,
      "loss": 0.5562,
      "step": 1394
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016063244015870204,
      "loss": 0.5544,
      "step": 1395
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001605744465797827,
      "loss": 0.5553,
      "step": 1396
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016051642080232236,
      "loss": 0.5434,
      "step": 1397
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001604583628571648,
      "loss": 0.5617,
      "step": 1398
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016040027277517094,
      "loss": 0.583,
      "step": 1399
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00016034215058721872,
      "loss": 0.5584,
      "step": 1400
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00016028399632420318,
      "loss": 0.5543,
      "step": 1401
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00016022581001703644,
      "loss": 0.5637,
      "step": 1402
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00016016759169664758,
      "loss": 0.5416,
      "step": 1403
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00016010934139398274,
      "loss": 0.5568,
      "step": 1404
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00016005105914000507,
      "loss": 0.5619,
      "step": 1405
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00015999274496569465,
      "loss": 0.5858,
      "step": 1406
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00015993439890204863,
      "loss": 0.5529,
      "step": 1407
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00015987602098008097,
      "loss": 0.5562,
      "step": 1408
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00015981761123082272,
      "loss": 0.5624,
      "step": 1409
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00015975916968532172,
      "loss": 0.5542,
      "step": 1410
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00015970069637464277,
      "loss": 0.5563,
      "step": 1411
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00015964219132986752,
      "loss": 0.5551,
      "step": 1412
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0001595836545820946,
      "loss": 0.554,
      "step": 1413
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00015952508616243934,
      "loss": 0.5536,
      "step": 1414
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015946648610203393,
      "loss": 0.5678,
      "step": 1415
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001594078544320275,
      "loss": 0.5592,
      "step": 1416
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015934919118358587,
      "loss": 0.5659,
      "step": 1417
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015929049638789168,
      "loss": 0.5595,
      "step": 1418
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015923177007614428,
      "loss": 0.5648,
      "step": 1419
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015917301227955994,
      "loss": 0.5654,
      "step": 1420
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001591142230293715,
      "loss": 0.5653,
      "step": 1421
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015905540235682853,
      "loss": 0.5594,
      "step": 1422
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001589965502931974,
      "loss": 0.5589,
      "step": 1423
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015893766686976115,
      "loss": 0.55,
      "step": 1424
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015887875211781938,
      "loss": 0.5548,
      "step": 1425
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001588198060686884,
      "loss": 0.5547,
      "step": 1426
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015876082875370125,
      "loss": 0.5492,
      "step": 1427
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015870182020420747,
      "loss": 0.5632,
      "step": 1428
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00015864278045157323,
      "loss": 0.5517,
      "step": 1429
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0001585837095271813,
      "loss": 0.567,
      "step": 1430
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015852460746243105,
      "loss": 0.5657,
      "step": 1431
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015846547428873838,
      "loss": 0.5619,
      "step": 1432
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015840631003753564,
      "loss": 0.5683,
      "step": 1433
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015834711474027187,
      "loss": 0.5564,
      "step": 1434
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015828788842841242,
      "loss": 0.5583,
      "step": 1435
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015822863113343935,
      "loss": 0.5523,
      "step": 1436
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015816934288685096,
      "loss": 0.5635,
      "step": 1437
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015811002372016216,
      "loss": 0.5508,
      "step": 1438
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015805067366490424,
      "loss": 0.5522,
      "step": 1439
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0001579912927526249,
      "loss": 0.5419,
      "step": 1440
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015793188101488826,
      "loss": 0.558,
      "step": 1441
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0001578724384832748,
      "loss": 0.5539,
      "step": 1442
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00015781296518938137,
      "loss": 0.5662,
      "step": 1443
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015775346116482126,
      "loss": 0.5526,
      "step": 1444
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.000157693926441224,
      "loss": 0.5649,
      "step": 1445
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015763436105023545,
      "loss": 0.5686,
      "step": 1446
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015757476502351777,
      "loss": 0.557,
      "step": 1447
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015751513839274943,
      "loss": 0.5506,
      "step": 1448
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015745548118962512,
      "loss": 0.559,
      "step": 1449
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0001573957934458559,
      "loss": 0.5546,
      "step": 1450
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015733607519316896,
      "loss": 0.5559,
      "step": 1451
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015727632646330767,
      "loss": 0.5709,
      "step": 1452
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015721654728803171,
      "loss": 0.5591,
      "step": 1453
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015715673769911686,
      "loss": 0.5527,
      "step": 1454
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015709689772835513,
      "loss": 0.5508,
      "step": 1455
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015703702740755464,
      "loss": 0.555,
      "step": 1456
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00015697712676853963,
      "loss": 0.5681,
      "step": 1457
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0001569171958431505,
      "loss": 0.5469,
      "step": 1458
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015685723466324372,
      "loss": 0.5531,
      "step": 1459
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015679724326069186,
      "loss": 0.5481,
      "step": 1460
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015673722166738348,
      "loss": 0.5584,
      "step": 1461
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001566771699152233,
      "loss": 0.5534,
      "step": 1462
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015661708803613205,
      "loss": 0.5422,
      "step": 1463
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001565569760620464,
      "loss": 0.5623,
      "step": 1464
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015649683402491906,
      "loss": 0.5696,
      "step": 1465
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015643666195671875,
      "loss": 0.5541,
      "step": 1466
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001563764598894301,
      "loss": 0.5457,
      "step": 1467
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001563162278550537,
      "loss": 0.5669,
      "step": 1468
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015625596588560617,
      "loss": 0.5564,
      "step": 1469
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015619567401311982,
      "loss": 0.558,
      "step": 1470
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015613535226964307,
      "loss": 0.5646,
      "step": 1471
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00015607500068724015,
      "loss": 0.541,
      "step": 1472
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001560146192979911,
      "loss": 0.5781,
      "step": 1473
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00015595420813399184,
      "loss": 0.5424,
      "step": 1474
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00015589376722735412,
      "loss": 0.5568,
      "step": 1475
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001558332966102055,
      "loss": 0.5702,
      "step": 1476
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00015577279631468933,
      "loss": 0.5569,
      "step": 1477
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00015571226637296474,
      "loss": 0.5624,
      "step": 1478
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00015565170681720663,
      "loss": 0.5537,
      "step": 1479
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001555911176796056,
      "loss": 0.556,
      "step": 1480
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.000155530498992368,
      "loss": 0.5709,
      "step": 1481
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001554698507877159,
      "loss": 0.5081,
      "step": 1482
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00015540917309788706,
      "loss": 0.5194,
      "step": 1483
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00015534846595513486,
      "loss": 0.5148,
      "step": 1484
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001552877293917284,
      "loss": 0.5077,
      "step": 1485
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001552269634399524,
      "loss": 0.5018,
      "step": 1486
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001551661681321072,
      "loss": 0.5152,
      "step": 1487
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00015510534350050874,
      "loss": 0.5222,
      "step": 1488
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.0001550444895774885,
      "loss": 0.5115,
      "step": 1489
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015498360639539365,
      "loss": 0.5177,
      "step": 1490
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.0001549226939865868,
      "loss": 0.4945,
      "step": 1491
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015486175238344613,
      "loss": 0.5146,
      "step": 1492
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015480078161836538,
      "loss": 0.5204,
      "step": 1493
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015473978172375372,
      "loss": 0.5139,
      "step": 1494
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015467875273203583,
      "loss": 0.5094,
      "step": 1495
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015461769467565192,
      "loss": 0.5103,
      "step": 1496
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015455660758705757,
      "loss": 0.5219,
      "step": 1497
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015449549149872376,
      "loss": 0.5131,
      "step": 1498
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015443434644313704,
      "loss": 0.5182,
      "step": 1499
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.0001543731724527992,
      "loss": 0.4948,
      "step": 1500
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015431196956022745,
      "loss": 0.5066,
      "step": 1501
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015425073779795447,
      "loss": 0.5155,
      "step": 1502
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00015418947719852809,
      "loss": 0.5115,
      "step": 1503
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015412818779451167,
      "loss": 0.5083,
      "step": 1504
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015406686961848376,
      "loss": 0.5099,
      "step": 1505
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015400552270303822,
      "loss": 0.5247,
      "step": 1506
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015394414708078417,
      "loss": 0.5108,
      "step": 1507
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015388274278434606,
      "loss": 0.508,
      "step": 1508
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015382130984636358,
      "loss": 0.5159,
      "step": 1509
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015375984829949153,
      "loss": 0.5084,
      "step": 1510
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015369835817640002,
      "loss": 0.5167,
      "step": 1511
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0001536368395097743,
      "loss": 0.5167,
      "step": 1512
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015357529233231486,
      "loss": 0.4991,
      "step": 1513
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015351371667673728,
      "loss": 0.5049,
      "step": 1514
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015345211257577227,
      "loss": 0.5184,
      "step": 1515
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015339048006216577,
      "loss": 0.5021,
      "step": 1516
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00015332881916867857,
      "loss": 0.5044,
      "step": 1517
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0001532671299280869,
      "loss": 0.5111,
      "step": 1518
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0001532054123731818,
      "loss": 0.4991,
      "step": 1519
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015314366653676942,
      "loss": 0.5045,
      "step": 1520
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015308189245167095,
      "loss": 0.5171,
      "step": 1521
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015302009015072266,
      "loss": 0.5044,
      "step": 1522
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015295825966677568,
      "loss": 0.5074,
      "step": 1523
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015289640103269625,
      "loss": 0.5073,
      "step": 1524
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015283451428136556,
      "loss": 0.5096,
      "step": 1525
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015277259944567964,
      "loss": 0.5076,
      "step": 1526
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015271065655854952,
      "loss": 0.514,
      "step": 1527
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0001526486856529012,
      "loss": 0.516,
      "step": 1528
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015258668676167546,
      "loss": 0.4939,
      "step": 1529
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015252465991782806,
      "loss": 0.5108,
      "step": 1530
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0001524626051543295,
      "loss": 0.5084,
      "step": 1531
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00015240052250416528,
      "loss": 0.5051,
      "step": 1532
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015233841200033553,
      "loss": 0.5072,
      "step": 1533
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.0001522762736758554,
      "loss": 0.4969,
      "step": 1534
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015221410756375458,
      "loss": 0.5081,
      "step": 1535
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015215191369707774,
      "loss": 0.5063,
      "step": 1536
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015208969210888426,
      "loss": 0.516,
      "step": 1537
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015202744283224822,
      "loss": 0.5128,
      "step": 1538
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015196516590025832,
      "loss": 0.507,
      "step": 1539
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015190286134601818,
      "loss": 0.5123,
      "step": 1540
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.0001518405292026459,
      "loss": 0.5054,
      "step": 1541
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.0001517781695032744,
      "loss": 0.5174,
      "step": 1542
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015171578228105116,
      "loss": 0.5246,
      "step": 1543
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015165336756913826,
      "loss": 0.5098,
      "step": 1544
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.0001515909254007125,
      "loss": 0.5134,
      "step": 1545
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015152845580896517,
      "loss": 0.5151,
      "step": 1546
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00015146595882710223,
      "loss": 0.5041,
      "step": 1547
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0001514034344883441,
      "loss": 0.5014,
      "step": 1548
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015134088282592583,
      "loss": 0.4972,
      "step": 1549
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015127830387309693,
      "loss": 0.5113,
      "step": 1550
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015121569766312146,
      "loss": 0.5091,
      "step": 1551
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015115306422927794,
      "loss": 0.5223,
      "step": 1552
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015109040360485935,
      "loss": 0.4976,
      "step": 1553
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0001510277158231732,
      "loss": 0.5095,
      "step": 1554
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015096500091754138,
      "loss": 0.5212,
      "step": 1555
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015090225892130013,
      "loss": 0.5199,
      "step": 1556
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015083948986780024,
      "loss": 0.5039,
      "step": 1557
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015077669379040672,
      "loss": 0.4991,
      "step": 1558
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015071387072249909,
      "loss": 0.5169,
      "step": 1559
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.00015065102069747118,
      "loss": 0.5074,
      "step": 1560
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.000150588143748731,
      "loss": 0.511,
      "step": 1561
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0001505252399097011,
      "loss": 0.4983,
      "step": 1562
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00015046230921381818,
      "loss": 0.5113,
      "step": 1563
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00015039935169453329,
      "loss": 0.5223,
      "step": 1564
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00015033636738531162,
      "loss": 0.511,
      "step": 1565
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00015027335631963276,
      "loss": 0.5138,
      "step": 1566
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00015021031853099033,
      "loss": 0.5122,
      "step": 1567
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00015014725405289246,
      "loss": 0.5132,
      "step": 1568
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.0001500841629188611,
      "loss": 0.5219,
      "step": 1569
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00015002104516243265,
      "loss": 0.5181,
      "step": 1570
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00014995790081715754,
      "loss": 0.5167,
      "step": 1571
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00014989472991660025,
      "loss": 0.5176,
      "step": 1572
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.0001498315324943396,
      "loss": 0.493,
      "step": 1573
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.0001497683085839683,
      "loss": 0.5265,
      "step": 1574
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00014970505821909326,
      "loss": 0.5138,
      "step": 1575
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00014964178143333544,
      "loss": 0.5068,
      "step": 1576
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.00014957847826032973,
      "loss": 0.5084,
      "step": 1577
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00014951514873372523,
      "loss": 0.5193,
      "step": 1578
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0001494517928871849,
      "loss": 0.5203,
      "step": 1579
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0001493884107543858,
      "loss": 0.5174,
      "step": 1580
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00014932500236901878,
      "loss": 0.5051,
      "step": 1581
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0001492615677647889,
      "loss": 0.5118,
      "step": 1582
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00014919810697541498,
      "loss": 0.514,
      "step": 1583
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0001491346200346298,
      "loss": 0.5114,
      "step": 1584
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00014907110697618003,
      "loss": 0.5025,
      "step": 1585
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00014900756783382628,
      "loss": 0.5247,
      "step": 1586
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00014894400264134298,
      "loss": 0.5053,
      "step": 1587
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00014888041143251837,
      "loss": 0.5185,
      "step": 1588
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0001488167942411546,
      "loss": 0.5183,
      "step": 1589
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00014875315110106756,
      "loss": 0.5306,
      "step": 1590
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.000148689482046087,
      "loss": 0.5013,
      "step": 1591
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.00014862578711005638,
      "loss": 0.4972,
      "step": 1592
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014856206632683297,
      "loss": 0.5111,
      "step": 1593
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014849831973028774,
      "loss": 0.5094,
      "step": 1594
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014843454735430543,
      "loss": 0.513,
      "step": 1595
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0001483707492327844,
      "loss": 0.5152,
      "step": 1596
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0001483069253996368,
      "loss": 0.5176,
      "step": 1597
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0001482430758887883,
      "loss": 0.5185,
      "step": 1598
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0001481792007341784,
      "loss": 0.5176,
      "step": 1599
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014811529996976015,
      "loss": 0.5227,
      "step": 1600
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014805137362950015,
      "loss": 0.5208,
      "step": 1601
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014798742174737867,
      "loss": 0.5116,
      "step": 1602
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014792344435738953,
      "loss": 0.5156,
      "step": 1603
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014785944149354018,
      "loss": 0.5165,
      "step": 1604
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014779541318985146,
      "loss": 0.5074,
      "step": 1605
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.00014773135948035787,
      "loss": 0.5023,
      "step": 1606
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014766728039910734,
      "loss": 0.5074,
      "step": 1607
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014760317598016133,
      "loss": 0.5128,
      "step": 1608
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014753904625759473,
      "loss": 0.4985,
      "step": 1609
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014747489126549593,
      "loss": 0.5149,
      "step": 1610
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014741071103796668,
      "loss": 0.5251,
      "step": 1611
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014734650560912222,
      "loss": 0.512,
      "step": 1612
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014728227501309112,
      "loss": 0.5167,
      "step": 1613
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001472180192840154,
      "loss": 0.5121,
      "step": 1614
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014715373845605037,
      "loss": 0.5024,
      "step": 1615
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001470894325633647,
      "loss": 0.5089,
      "step": 1616
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001470251016401404,
      "loss": 0.5076,
      "step": 1617
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001469607457205728,
      "loss": 0.5189,
      "step": 1618
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014689636483887047,
      "loss": 0.5111,
      "step": 1619
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0001468319590292553,
      "loss": 0.5221,
      "step": 1620
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00014676752832596242,
      "loss": 0.5168,
      "step": 1621
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0001467030727632401,
      "loss": 0.514,
      "step": 1622
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014663859237534993,
      "loss": 0.5122,
      "step": 1623
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0001465740871965667,
      "loss": 0.5005,
      "step": 1624
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014650955726117835,
      "loss": 0.4966,
      "step": 1625
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014644500260348596,
      "loss": 0.5019,
      "step": 1626
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0001463804232578037,
      "loss": 0.5068,
      "step": 1627
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014631581925845903,
      "loss": 0.5088,
      "step": 1628
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014625119063979236,
      "loss": 0.5111,
      "step": 1629
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014618653743615724,
      "loss": 0.516,
      "step": 1630
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014612185968192028,
      "loss": 0.5045,
      "step": 1631
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014605715741146117,
      "loss": 0.5092,
      "step": 1632
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014599243065917264,
      "loss": 0.5201,
      "step": 1633
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0001459276794594603,
      "loss": 0.515,
      "step": 1634
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014586290384674294,
      "loss": 0.518,
      "step": 1635
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00014579810385545222,
      "loss": 0.5225,
      "step": 1636
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014573327952003282,
      "loss": 0.5022,
      "step": 1637
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014566843087494225,
      "loss": 0.5193,
      "step": 1638
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014560355795465108,
      "loss": 0.5077,
      "step": 1639
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014553866079364267,
      "loss": 0.5069,
      "step": 1640
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014547373942641335,
      "loss": 0.5237,
      "step": 1641
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014540879388747228,
      "loss": 0.5029,
      "step": 1642
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014534382421134144,
      "loss": 0.5133,
      "step": 1643
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0001452788304325557,
      "loss": 0.5098,
      "step": 1644
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014521381258566268,
      "loss": 0.509,
      "step": 1645
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014514877070522284,
      "loss": 0.5305,
      "step": 1646
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0001450837048258094,
      "loss": 0.4991,
      "step": 1647
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014501861498200834,
      "loss": 0.5185,
      "step": 1648
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014495350120841838,
      "loss": 0.4893,
      "step": 1649
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014488836353965094,
      "loss": 0.5193,
      "step": 1650
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.00014482320201033015,
      "loss": 0.505,
      "step": 1651
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014475801665509284,
      "loss": 0.52,
      "step": 1652
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014469280750858854,
      "loss": 0.5086,
      "step": 1653
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.0001446275746054793,
      "loss": 0.5078,
      "step": 1654
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014456231798043993,
      "loss": 0.5034,
      "step": 1655
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014449703766815781,
      "loss": 0.5289,
      "step": 1656
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014443173370333286,
      "loss": 0.5081,
      "step": 1657
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014436640612067758,
      "loss": 0.5181,
      "step": 1658
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014430105495491714,
      "loss": 0.5251,
      "step": 1659
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014423568024078912,
      "loss": 0.5005,
      "step": 1660
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014417028201304367,
      "loss": 0.5138,
      "step": 1661
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.0001441048603064434,
      "loss": 0.5059,
      "step": 1662
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014403941515576344,
      "loss": 0.5189,
      "step": 1663
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.0001439739465957914,
      "loss": 0.5114,
      "step": 1664
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00014390845466132724,
      "loss": 0.509,
      "step": 1665
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.0001438429393871835,
      "loss": 0.5034,
      "step": 1666
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00014377740080818492,
      "loss": 0.5222,
      "step": 1667
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00014371183895916883,
      "loss": 0.5123,
      "step": 1668
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0001436462538749848,
      "loss": 0.5134,
      "step": 1669
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0001435806455904948,
      "loss": 0.5034,
      "step": 1670
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00014351501414057316,
      "loss": 0.5147,
      "step": 1671
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00014344935956010638,
      "loss": 0.5189,
      "step": 1672
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0001433836818839935,
      "loss": 0.5136,
      "step": 1673
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00014331798114714564,
      "loss": 0.5181,
      "step": 1674
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00014325225738448616,
      "loss": 0.5177,
      "step": 1675
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0001431865106309509,
      "loss": 0.5251,
      "step": 1676
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00014312074092148757,
      "loss": 0.5189,
      "step": 1677
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00014305494829105643,
      "loss": 0.5077,
      "step": 1678
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0001429891327746297,
      "loss": 0.5224,
      "step": 1679
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00014292329440719178,
      "loss": 0.5121,
      "step": 1680
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014285743322373936,
      "loss": 0.5088,
      "step": 1681
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0001427915492592811,
      "loss": 0.5158,
      "step": 1682
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014272564254883783,
      "loss": 0.5071,
      "step": 1683
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014265971312744252,
      "loss": 0.5164,
      "step": 1684
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014259376103014008,
      "loss": 0.5145,
      "step": 1685
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014252778629198767,
      "loss": 0.5148,
      "step": 1686
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0001424617889480543,
      "loss": 0.5096,
      "step": 1687
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014239576903342109,
      "loss": 0.5102,
      "step": 1688
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0001423297265831811,
      "loss": 0.5101,
      "step": 1689
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014226366163243946,
      "loss": 0.515,
      "step": 1690
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0001421975742163132,
      "loss": 0.518,
      "step": 1691
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014213146436993127,
      "loss": 0.5224,
      "step": 1692
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0001420653321284346,
      "loss": 0.5136,
      "step": 1693
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014199917752697597,
      "loss": 0.5178,
      "step": 1694
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.00014193300060072003,
      "loss": 0.5151,
      "step": 1695
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014186680138484343,
      "loss": 0.5172,
      "step": 1696
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0001418005799145345,
      "loss": 0.5276,
      "step": 1697
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014173433622499353,
      "loss": 0.5145,
      "step": 1698
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014166807035143254,
      "loss": 0.5066,
      "step": 1699
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014160178232907534,
      "loss": 0.5185,
      "step": 1700
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0001415354721931576,
      "loss": 0.5114,
      "step": 1701
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014146913997892664,
      "loss": 0.5147,
      "step": 1702
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014140278572164156,
      "loss": 0.5108,
      "step": 1703
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0001413364094565732,
      "loss": 0.5133,
      "step": 1704
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014127001121900406,
      "loss": 0.516,
      "step": 1705
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014120359104422838,
      "loss": 0.5197,
      "step": 1706
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014113714896755194,
      "loss": 0.527,
      "step": 1707
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014107068502429228,
      "loss": 0.5147,
      "step": 1708
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00014100419924977854,
      "loss": 0.504,
      "step": 1709
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0001409376916793514,
      "loss": 0.5087,
      "step": 1710
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00014087116234836323,
      "loss": 0.5308,
      "step": 1711
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00014080461129217786,
      "loss": 0.5133,
      "step": 1712
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0001407380385461707,
      "loss": 0.5163,
      "step": 1713
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0001406714441457288,
      "loss": 0.5067,
      "step": 1714
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00014060482812625055,
      "loss": 0.5037,
      "step": 1715
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00014053819052314593,
      "loss": 0.5054,
      "step": 1716
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00014047153137183632,
      "loss": 0.5224,
      "step": 1717
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00014040485070775468,
      "loss": 0.5172,
      "step": 1718
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0001403381485663453,
      "loss": 0.5159,
      "step": 1719
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00014027142498306393,
      "loss": 0.508,
      "step": 1720
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0001402046799933777,
      "loss": 0.5062,
      "step": 1721
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0001401379136327651,
      "loss": 0.5044,
      "step": 1722
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00014007112593671602,
      "loss": 0.517,
      "step": 1723
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.00014000431694073169,
      "loss": 0.4972,
      "step": 1724
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0001399374866803246,
      "loss": 0.5096,
      "step": 1725
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013987063519101863,
      "loss": 0.4942,
      "step": 1726
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013980376250834886,
      "loss": 0.5139,
      "step": 1727
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013973686866786174,
      "loss": 0.4993,
      "step": 1728
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013966995370511486,
      "loss": 0.513,
      "step": 1729
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0001396030176556771,
      "loss": 0.5151,
      "step": 1730
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013953606055512854,
      "loss": 0.5224,
      "step": 1731
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013946908243906044,
      "loss": 0.5176,
      "step": 1732
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013940208334307518,
      "loss": 0.5122,
      "step": 1733
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013933506330278644,
      "loss": 0.5203,
      "step": 1734
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013926802235381884,
      "loss": 0.5233,
      "step": 1735
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013920096053180825,
      "loss": 0.5081,
      "step": 1736
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013913387787240161,
      "loss": 0.5138,
      "step": 1737
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013906677441125694,
      "loss": 0.532,
      "step": 1738
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00013899965018404324,
      "loss": 0.5202,
      "step": 1739
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0001389325052264406,
      "loss": 0.5173,
      "step": 1740
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013886533957414024,
      "loss": 0.5169,
      "step": 1741
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0001387981532628442,
      "loss": 0.5107,
      "step": 1742
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0001387309463282656,
      "loss": 0.5237,
      "step": 1743
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013866371880612847,
      "loss": 0.5228,
      "step": 1744
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013859647073216784,
      "loss": 0.5067,
      "step": 1745
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013852920214212964,
      "loss": 0.5035,
      "step": 1746
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0001384619130717707,
      "loss": 0.5086,
      "step": 1747
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013839460355685875,
      "loss": 0.5056,
      "step": 1748
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013832727363317237,
      "loss": 0.5054,
      "step": 1749
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013825992333650099,
      "loss": 0.5198,
      "step": 1750
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013819255270264486,
      "loss": 0.5039,
      "step": 1751
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013812516176741512,
      "loss": 0.5262,
      "step": 1752
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013805775056663358,
      "loss": 0.5131,
      "step": 1753
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00013799031913613288,
      "loss": 0.5297,
      "step": 1754
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013792286751175645,
      "loss": 0.5174,
      "step": 1755
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0001378553957293584,
      "loss": 0.5137,
      "step": 1756
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013778790382480355,
      "loss": 0.5146,
      "step": 1757
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013772039183396746,
      "loss": 0.5243,
      "step": 1758
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013765285979273635,
      "loss": 0.5096,
      "step": 1759
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013758530773700707,
      "loss": 0.5041,
      "step": 1760
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013751773570268714,
      "loss": 0.5172,
      "step": 1761
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0001374501437256947,
      "loss": 0.5158,
      "step": 1762
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0001373825318419585,
      "loss": 0.5185,
      "step": 1763
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0001373149000874178,
      "loss": 0.5029,
      "step": 1764
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013724724849802254,
      "loss": 0.5242,
      "step": 1765
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013717957710973307,
      "loss": 0.5062,
      "step": 1766
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013711188595852037,
      "loss": 0.5032,
      "step": 1767
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013704417508036588,
      "loss": 0.5302,
      "step": 1768
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.00013697644451126156,
      "loss": 0.5213,
      "step": 1769
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0001369086942872098,
      "loss": 0.5129,
      "step": 1770
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013684092444422343,
      "loss": 0.5195,
      "step": 1771
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0001367731350183257,
      "loss": 0.5132,
      "step": 1772
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013670532604555038,
      "loss": 0.5072,
      "step": 1773
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013663749756194152,
      "loss": 0.5198,
      "step": 1774
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013656964960355347,
      "loss": 0.5084,
      "step": 1775
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013650178220645117,
      "loss": 0.5226,
      "step": 1776
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013643389540670962,
      "loss": 0.5094,
      "step": 1777
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013636598924041438,
      "loss": 0.5159,
      "step": 1778
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013629806374366118,
      "loss": 0.5204,
      "step": 1779
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013623011895255594,
      "loss": 0.513,
      "step": 1780
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.000136162154903215,
      "loss": 0.5051,
      "step": 1781
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0001360941716317649,
      "loss": 0.5076,
      "step": 1782
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0001360261691743423,
      "loss": 0.5027,
      "step": 1783
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.00013595814756709414,
      "loss": 0.5235,
      "step": 1784
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013589010684617752,
      "loss": 0.5114,
      "step": 1785
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013582204704775972,
      "loss": 0.5176,
      "step": 1786
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013575396820801814,
      "loss": 0.494,
      "step": 1787
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013568587036314034,
      "loss": 0.5147,
      "step": 1788
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013561775354932386,
      "loss": 0.5029,
      "step": 1789
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013554961780277647,
      "loss": 0.5102,
      "step": 1790
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013548146315971594,
      "loss": 0.5135,
      "step": 1791
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.0001354132896563701,
      "loss": 0.5193,
      "step": 1792
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.0001353450973289768,
      "loss": 0.5222,
      "step": 1793
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013527688621378382,
      "loss": 0.5294,
      "step": 1794
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013520865634704905,
      "loss": 0.5008,
      "step": 1795
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.0001351404077650403,
      "loss": 0.5229,
      "step": 1796
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013507214050403532,
      "loss": 0.5075,
      "step": 1797
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.0001350038546003218,
      "loss": 0.5161,
      "step": 1798
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00013493555009019725,
      "loss": 0.5096,
      "step": 1799
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013486722700996925,
      "loss": 0.5024,
      "step": 1800
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.0001347988853959551,
      "loss": 0.5145,
      "step": 1801
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013473052528448201,
      "loss": 0.5123,
      "step": 1802
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013466214671188707,
      "loss": 0.5031,
      "step": 1803
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013459374971451705,
      "loss": 0.5114,
      "step": 1804
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013452533432872866,
      "loss": 0.508,
      "step": 1805
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013445690059088827,
      "loss": 0.5173,
      "step": 1806
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013438844853737205,
      "loss": 0.5107,
      "step": 1807
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013431997820456592,
      "loss": 0.5071,
      "step": 1808
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013425148962886553,
      "loss": 0.5223,
      "step": 1809
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013418298284667622,
      "loss": 0.4992,
      "step": 1810
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.0001341144578944129,
      "loss": 0.5078,
      "step": 1811
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013404591480850028,
      "loss": 0.5125,
      "step": 1812
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013397735362537266,
      "loss": 0.5194,
      "step": 1813
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.00013390877438147395,
      "loss": 0.5252,
      "step": 1814
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.0001338401771132577,
      "loss": 0.5016,
      "step": 1815
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013377156185718695,
      "loss": 0.512,
      "step": 1816
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.0001337029286497344,
      "loss": 0.5109,
      "step": 1817
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.0001336342775273822,
      "loss": 0.5099,
      "step": 1818
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013356560852662214,
      "loss": 0.5204,
      "step": 1819
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.0001334969216839554,
      "loss": 0.5079,
      "step": 1820
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013342821703589275,
      "loss": 0.5073,
      "step": 1821
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013335949461895424,
      "loss": 0.51,
      "step": 1822
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013329075446966965,
      "loss": 0.511,
      "step": 1823
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013322199662457794,
      "loss": 0.5234,
      "step": 1824
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013315322112022755,
      "loss": 0.514,
      "step": 1825
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013308442799317636,
      "loss": 0.5023,
      "step": 1826
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013301561727999153,
      "loss": 0.5152,
      "step": 1827
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013294678901724967,
      "loss": 0.513,
      "step": 1828
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00013287794324153663,
      "loss": 0.5082,
      "step": 1829
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013280907998944764,
      "loss": 0.5077,
      "step": 1830
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013274019929758712,
      "loss": 0.5201,
      "step": 1831
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013267130120256889,
      "loss": 0.5235,
      "step": 1832
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013260238574101587,
      "loss": 0.5214,
      "step": 1833
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0001325334529495604,
      "loss": 0.5097,
      "step": 1834
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0001324645028648438,
      "loss": 0.5131,
      "step": 1835
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013239553552351683,
      "loss": 0.5058,
      "step": 1836
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013232655096223923,
      "loss": 0.5162,
      "step": 1837
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013225754921767998,
      "loss": 0.5163,
      "step": 1838
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0001321885303265172,
      "loss": 0.4994,
      "step": 1839
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013211949432543805,
      "loss": 0.5052,
      "step": 1840
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013205044125113888,
      "loss": 0.5057,
      "step": 1841
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013198137114032508,
      "loss": 0.5274,
      "step": 1842
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.00013191228402971105,
      "loss": 0.5066,
      "step": 1843
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0001318431799560203,
      "loss": 0.5213,
      "step": 1844
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013177405895598528,
      "loss": 0.5092,
      "step": 1845
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013170492106634755,
      "loss": 0.5186,
      "step": 1846
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013163576632385752,
      "loss": 0.5166,
      "step": 1847
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0001315665947652746,
      "loss": 0.5179,
      "step": 1848
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013149740642736722,
      "loss": 0.5064,
      "step": 1849
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013142820134691262,
      "loss": 0.5132,
      "step": 1850
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013135897956069695,
      "loss": 0.5144,
      "step": 1851
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013128974110551532,
      "loss": 0.5151,
      "step": 1852
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013122048601817165,
      "loss": 0.5144,
      "step": 1853
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0001311512143354787,
      "loss": 0.5251,
      "step": 1854
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013108192609425804,
      "loss": 0.4999,
      "step": 1855
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013101262133134007,
      "loss": 0.5115,
      "step": 1856
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013094330008356394,
      "loss": 0.5088,
      "step": 1857
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00013087396238777762,
      "loss": 0.4978,
      "step": 1858
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00013080460828083774,
      "loss": 0.5092,
      "step": 1859
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00013073523779960972,
      "loss": 0.5243,
      "step": 1860
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00013066585098096766,
      "loss": 0.5119,
      "step": 1861
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00013059644786179436,
      "loss": 0.5197,
      "step": 1862
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0001305270284789812,
      "loss": 0.5038,
      "step": 1863
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00013045759286942838,
      "loss": 0.5161,
      "step": 1864
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0001303881410700446,
      "loss": 0.5054,
      "step": 1865
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00013031867311774707,
      "loss": 0.5221,
      "step": 1866
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00013024918904946183,
      "loss": 0.5055,
      "step": 1867
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0001301796889021233,
      "loss": 0.5095,
      "step": 1868
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00013011017271267455,
      "loss": 0.5141,
      "step": 1869
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0001300406405180671,
      "loss": 0.5128,
      "step": 1870
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00012997109235526102,
      "loss": 0.5129,
      "step": 1871
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0001299015282612248,
      "loss": 0.4936,
      "step": 1872
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00012983194827293558,
      "loss": 0.5076,
      "step": 1873
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012976235242737874,
      "loss": 0.5216,
      "step": 1874
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012969274076154815,
      "loss": 0.5131,
      "step": 1875
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012962311331244622,
      "loss": 0.4956,
      "step": 1876
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012955347011708352,
      "loss": 0.5092,
      "step": 1877
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012948381121247922,
      "loss": 0.5171,
      "step": 1878
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012941413663566066,
      "loss": 0.5063,
      "step": 1879
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012934444642366365,
      "loss": 0.5219,
      "step": 1880
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012927474061353214,
      "loss": 0.5143,
      "step": 1881
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.0001292050192423186,
      "loss": 0.5126,
      "step": 1882
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012913528234708354,
      "loss": 0.5104,
      "step": 1883
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012906552996489593,
      "loss": 0.508,
      "step": 1884
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.0001289957621328328,
      "loss": 0.4864,
      "step": 1885
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012892597888797947,
      "loss": 0.5143,
      "step": 1886
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.00012885618026742948,
      "loss": 0.5119,
      "step": 1887
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.0001287863663082845,
      "loss": 0.5189,
      "step": 1888
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012871653704765436,
      "loss": 0.5122,
      "step": 1889
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.000128646692522657,
      "loss": 0.5106,
      "step": 1890
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012857683277041855,
      "loss": 0.5224,
      "step": 1891
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012850695782807317,
      "loss": 0.5087,
      "step": 1892
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012843706773276307,
      "loss": 0.5098,
      "step": 1893
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012836716252163862,
      "loss": 0.5126,
      "step": 1894
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012829724223185806,
      "loss": 0.5126,
      "step": 1895
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012822730690058786,
      "loss": 0.5076,
      "step": 1896
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012815735656500232,
      "loss": 0.5078,
      "step": 1897
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012808739126228373,
      "loss": 0.5178,
      "step": 1898
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012801741102962238,
      "loss": 0.5098,
      "step": 1899
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012794741590421658,
      "loss": 0.519,
      "step": 1900
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.0001278774059232723,
      "loss": 0.5066,
      "step": 1901
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00012780738112400374,
      "loss": 0.5045,
      "step": 1902
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.0001277373415436327,
      "loss": 0.5155,
      "step": 1903
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00012766728721938893,
      "loss": 0.5127,
      "step": 1904
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0001275972181885101,
      "loss": 0.5233,
      "step": 1905
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00012752713448824162,
      "loss": 0.504,
      "step": 1906
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0001274570361558367,
      "loss": 0.5121,
      "step": 1907
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00012738692322855627,
      "loss": 0.5126,
      "step": 1908
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00012731679574366913,
      "loss": 0.5158,
      "step": 1909
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00012724665373845177,
      "loss": 0.5224,
      "step": 1910
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0001271764972501884,
      "loss": 0.5102,
      "step": 1911
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00012710632631617093,
      "loss": 0.5116,
      "step": 1912
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0001270361409736989,
      "loss": 0.5147,
      "step": 1913
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0001269659412600796,
      "loss": 0.511,
      "step": 1914
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00012689572721262792,
      "loss": 0.5142,
      "step": 1915
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00012682549886866635,
      "loss": 0.515,
      "step": 1916
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00012675525626552496,
      "loss": 0.5078,
      "step": 1917
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012668499944054143,
      "loss": 0.5066,
      "step": 1918
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012661472843106107,
      "loss": 0.5142,
      "step": 1919
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012654444327443658,
      "loss": 0.5119,
      "step": 1920
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012647414400802833,
      "loss": 0.5106,
      "step": 1921
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012640383066920402,
      "loss": 0.5181,
      "step": 1922
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012633350329533904,
      "loss": 0.507,
      "step": 1923
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.0001262631619238161,
      "loss": 0.5132,
      "step": 1924
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.0001261928065920253,
      "loss": 0.5022,
      "step": 1925
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012612243733736434,
      "loss": 0.5073,
      "step": 1926
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012605205419723823,
      "loss": 0.5074,
      "step": 1927
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012598165720905925,
      "loss": 0.5127,
      "step": 1928
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012591124641024725,
      "loss": 0.5016,
      "step": 1929
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012584082183822928,
      "loss": 0.52,
      "step": 1930
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.0001257703835304397,
      "loss": 0.5023,
      "step": 1931
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.00012569993152432028,
      "loss": 0.5164,
      "step": 1932
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012562946585732002,
      "loss": 0.5155,
      "step": 1933
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012555898656689508,
      "loss": 0.5097,
      "step": 1934
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012548849369050906,
      "loss": 0.5058,
      "step": 1935
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012541798726563256,
      "loss": 0.512,
      "step": 1936
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012534746732974363,
      "loss": 0.5094,
      "step": 1937
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012527693392032726,
      "loss": 0.4901,
      "step": 1938
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012520638707487578,
      "loss": 0.4954,
      "step": 1939
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012513582683088855,
      "loss": 0.5056,
      "step": 1940
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012506525322587207,
      "loss": 0.5042,
      "step": 1941
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012499466629734002,
      "loss": 0.5166,
      "step": 1942
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012492406608281308,
      "loss": 0.5078,
      "step": 1943
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012485345261981906,
      "loss": 0.5152,
      "step": 1944
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012478282594589267,
      "loss": 0.5029,
      "step": 1945
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012471218609857586,
      "loss": 0.5057,
      "step": 1946
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00012464153311541744,
      "loss": 0.5087,
      "step": 1947
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0001245708670339732,
      "loss": 0.5186,
      "step": 1948
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0001245001878918059,
      "loss": 0.5119,
      "step": 1949
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0001244294957264853,
      "loss": 0.5151,
      "step": 1950
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012435879057558804,
      "loss": 0.5107,
      "step": 1951
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0001242880724766977,
      "loss": 0.5186,
      "step": 1952
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012421734146740468,
      "loss": 0.511,
      "step": 1953
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012414659758530624,
      "loss": 0.5034,
      "step": 1954
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012407584086800662,
      "loss": 0.5163,
      "step": 1955
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012400507135311667,
      "loss": 0.5011,
      "step": 1956
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012393428907825422,
      "loss": 0.5131,
      "step": 1957
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012386349408104378,
      "loss": 0.5108,
      "step": 1958
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012379268639911665,
      "loss": 0.5109,
      "step": 1959
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012372186607011092,
      "loss": 0.5051,
      "step": 1960
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012365103313167136,
      "loss": 0.5034,
      "step": 1961
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.00012358018762144942,
      "loss": 0.5081,
      "step": 1962
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0001235093295771032,
      "loss": 0.5157,
      "step": 1963
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012343845903629765,
      "loss": 0.521,
      "step": 1964
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012336757603670416,
      "loss": 0.507,
      "step": 1965
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012329668061600085,
      "loss": 0.5033,
      "step": 1966
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012322577281187233,
      "loss": 0.507,
      "step": 1967
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012315485266200992,
      "loss": 0.5172,
      "step": 1968
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0001230839202041115,
      "loss": 0.5006,
      "step": 1969
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012301297547588137,
      "loss": 0.5056,
      "step": 1970
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012294201851503045,
      "loss": 0.4981,
      "step": 1971
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012287104935927617,
      "loss": 0.51,
      "step": 1972
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012280006804634235,
      "loss": 0.5389,
      "step": 1973
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0001227290746139594,
      "loss": 0.4971,
      "step": 1974
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012265806909986406,
      "loss": 0.5139,
      "step": 1975
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0001225870515417996,
      "loss": 0.5007,
      "step": 1976
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00012251602197751554,
      "loss": 0.5184,
      "step": 1977
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012244498044476794,
      "loss": 0.5106,
      "step": 1978
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012237392698131913,
      "loss": 0.5316,
      "step": 1979
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012230286162493782,
      "loss": 0.5178,
      "step": 1980
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012223178441339902,
      "loss": 0.5146,
      "step": 1981
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.000122160695384484,
      "loss": 0.5176,
      "step": 1982
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012208959457598047,
      "loss": 0.5134,
      "step": 1983
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012201848202568218,
      "loss": 0.5053,
      "step": 1984
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012194735777138933,
      "loss": 0.5082,
      "step": 1985
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012187622185090815,
      "loss": 0.5128,
      "step": 1986
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012180507430205124,
      "loss": 0.5133,
      "step": 1987
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012173391516263728,
      "loss": 0.4967,
      "step": 1988
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012166274447049116,
      "loss": 0.5112,
      "step": 1989
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012159156226344386,
      "loss": 0.5024,
      "step": 1990
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00012152036857933249,
      "loss": 0.499,
      "step": 1991
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012144916345600033,
      "loss": 0.5172,
      "step": 1992
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0001213779469312967,
      "loss": 0.5118,
      "step": 1993
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0001213067190430769,
      "loss": 0.5218,
      "step": 1994
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0001212354798292024,
      "loss": 0.5062,
      "step": 1995
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012116422932754061,
      "loss": 0.516,
      "step": 1996
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.000121092967575965,
      "loss": 0.5072,
      "step": 1997
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012102169461235493,
      "loss": 0.5122,
      "step": 1998
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012095041047459576,
      "loss": 0.5063,
      "step": 1999
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012087911520057889,
      "loss": 0.5186,
      "step": 2000
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012080780882820149,
      "loss": 0.5097,
      "step": 2001
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012073649139536666,
      "loss": 0.5068,
      "step": 2002
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012066516293998348,
      "loss": 0.508,
      "step": 2003
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012059382349996677,
      "loss": 0.5117,
      "step": 2004
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012052247311323725,
      "loss": 0.5126,
      "step": 2005
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.00012045111181772148,
      "loss": 0.5132,
      "step": 2006
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00012037973965135174,
      "loss": 0.5101,
      "step": 2007
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00012030835665206614,
      "loss": 0.5178,
      "step": 2008
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00012023696285780854,
      "loss": 0.5224,
      "step": 2009
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00012016555830652858,
      "loss": 0.4982,
      "step": 2010
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00012009414303618154,
      "loss": 0.5122,
      "step": 2011
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00012002271708472845,
      "loss": 0.514,
      "step": 2012
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00011995128049013595,
      "loss": 0.5047,
      "step": 2013
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00011987983329037649,
      "loss": 0.5086,
      "step": 2014
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00011980837552342799,
      "loss": 0.5325,
      "step": 2015
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00011973690722727407,
      "loss": 0.5165,
      "step": 2016
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00011966542843990394,
      "loss": 0.4949,
      "step": 2017
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00011959393919931235,
      "loss": 0.5093,
      "step": 2018
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00011952243954349965,
      "loss": 0.5065,
      "step": 2019
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00011945092951047174,
      "loss": 0.5082,
      "step": 2020
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00011937940913823997,
      "loss": 0.5102,
      "step": 2021
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011930787846482123,
      "loss": 0.5117,
      "step": 2022
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011923633752823788,
      "loss": 0.5094,
      "step": 2023
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011916478636651776,
      "loss": 0.5157,
      "step": 2024
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011909322501769406,
      "loss": 0.5117,
      "step": 2025
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.0001190216535198055,
      "loss": 0.5161,
      "step": 2026
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011895007191089613,
      "loss": 0.5128,
      "step": 2027
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011887848022901535,
      "loss": 0.5032,
      "step": 2028
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011880687851221798,
      "loss": 0.5073,
      "step": 2029
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011873526679856415,
      "loss": 0.5061,
      "step": 2030
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011866364512611927,
      "loss": 0.509,
      "step": 2031
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011859201353295406,
      "loss": 0.5063,
      "step": 2032
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011852037205714454,
      "loss": 0.5242,
      "step": 2033
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.000118448720736772,
      "loss": 0.4927,
      "step": 2034
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011837705960992287,
      "loss": 0.5134,
      "step": 2035
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00011830538871468885,
      "loss": 0.5079,
      "step": 2036
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.0001182337080891669,
      "loss": 0.5005,
      "step": 2037
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011816201777145904,
      "loss": 0.5189,
      "step": 2038
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011809031779967247,
      "loss": 0.5197,
      "step": 2039
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011801860821191958,
      "loss": 0.5046,
      "step": 2040
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.0001179468890463178,
      "loss": 0.5162,
      "step": 2041
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011787516034098973,
      "loss": 0.5079,
      "step": 2042
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011780342213406292,
      "loss": 0.5155,
      "step": 2043
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011773167446367011,
      "loss": 0.5215,
      "step": 2044
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011765991736794895,
      "loss": 0.5068,
      "step": 2045
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.0001175881508850422,
      "loss": 0.5119,
      "step": 2046
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011751637505309754,
      "loss": 0.5077,
      "step": 2047
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011744458991026766,
      "loss": 0.5048,
      "step": 2048
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011737279549471012,
      "loss": 0.5187,
      "step": 2049
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.00011730099184458756,
      "loss": 0.5118,
      "step": 2050
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.0001172291789980674,
      "loss": 0.5061,
      "step": 2051
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011715735699332198,
      "loss": 0.5113,
      "step": 2052
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011708552586852853,
      "loss": 0.5063,
      "step": 2053
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.0001170136856618691,
      "loss": 0.5104,
      "step": 2054
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.0001169418364115306,
      "loss": 0.5075,
      "step": 2055
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011686997815570473,
      "loss": 0.5047,
      "step": 2056
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011679811093258797,
      "loss": 0.5082,
      "step": 2057
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011672623478038156,
      "loss": 0.5067,
      "step": 2058
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011665434973729151,
      "loss": 0.5091,
      "step": 2059
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011658245584152853,
      "loss": 0.502,
      "step": 2060
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011651055313130807,
      "loss": 0.5169,
      "step": 2061
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011643864164485027,
      "loss": 0.5085,
      "step": 2062
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011636672142037986,
      "loss": 0.5035,
      "step": 2063
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011629479249612627,
      "loss": 0.5046,
      "step": 2064
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00011622285491032356,
      "loss": 0.5103,
      "step": 2065
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011615090870121042,
      "loss": 0.5008,
      "step": 2066
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011607895390703005,
      "loss": 0.514,
      "step": 2067
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011600699056603024,
      "loss": 0.5163,
      "step": 2068
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011593501871646338,
      "loss": 0.5137,
      "step": 2069
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011586303839658627,
      "loss": 0.5116,
      "step": 2070
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011579104964466036,
      "loss": 0.5158,
      "step": 2071
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011571905249895146,
      "loss": 0.4999,
      "step": 2072
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011564704699772986,
      "loss": 0.5001,
      "step": 2073
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011557503317927034,
      "loss": 0.4945,
      "step": 2074
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011550301108185207,
      "loss": 0.5189,
      "step": 2075
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011543098074375866,
      "loss": 0.509,
      "step": 2076
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.000115358942203278,
      "loss": 0.5135,
      "step": 2077
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011528689549870246,
      "loss": 0.4894,
      "step": 2078
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.0001152148406683287,
      "loss": 0.514,
      "step": 2079
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.00011514277775045768,
      "loss": 0.5037,
      "step": 2080
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011507070678339466,
      "loss": 0.5034,
      "step": 2081
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011499862780544919,
      "loss": 0.5087,
      "step": 2082
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0001149265408549351,
      "loss": 0.5207,
      "step": 2083
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011485444597017046,
      "loss": 0.5021,
      "step": 2084
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011478234318947749,
      "loss": 0.5152,
      "step": 2085
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011471023255118267,
      "loss": 0.5198,
      "step": 2086
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011463811409361667,
      "loss": 0.5062,
      "step": 2087
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011456598785511425,
      "loss": 0.4994,
      "step": 2088
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011449385387401437,
      "loss": 0.5183,
      "step": 2089
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011442171218866006,
      "loss": 0.5149,
      "step": 2090
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011434956283739843,
      "loss": 0.5113,
      "step": 2091
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0001142774058585808,
      "loss": 0.4975,
      "step": 2092
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011420524129056235,
      "loss": 0.5075,
      "step": 2093
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011413306917170244,
      "loss": 0.5253,
      "step": 2094
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.00011406088954036432,
      "loss": 0.5076,
      "step": 2095
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011398870243491539,
      "loss": 0.5171,
      "step": 2096
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.0001139165078937269,
      "loss": 0.503,
      "step": 2097
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011384430595517408,
      "loss": 0.5018,
      "step": 2098
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011377209665763611,
      "loss": 0.5231,
      "step": 2099
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011369988003949608,
      "loss": 0.5082,
      "step": 2100
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011362765613914097,
      "loss": 0.5168,
      "step": 2101
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011355542499496157,
      "loss": 0.514,
      "step": 2102
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011348318664535267,
      "loss": 0.5115,
      "step": 2103
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011341094112871274,
      "loss": 0.5069,
      "step": 2104
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011333868848344406,
      "loss": 0.5126,
      "step": 2105
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011326642874795288,
      "loss": 0.5022,
      "step": 2106
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011319416196064902,
      "loss": 0.5103,
      "step": 2107
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011312188815994616,
      "loss": 0.506,
      "step": 2108
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011304960738426157,
      "loss": 0.5212,
      "step": 2109
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.00011297731967201646,
      "loss": 0.5072,
      "step": 2110
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011290502506163551,
      "loss": 0.499,
      "step": 2111
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.0001128327235915472,
      "loss": 0.5114,
      "step": 2112
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011276041530018358,
      "loss": 0.5012,
      "step": 2113
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011268810022598034,
      "loss": 0.5076,
      "step": 2114
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011261577840737681,
      "loss": 0.5213,
      "step": 2115
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011254344988281589,
      "loss": 0.5037,
      "step": 2116
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011247111469074403,
      "loss": 0.505,
      "step": 2117
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011239877286961122,
      "loss": 0.5009,
      "step": 2118
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.000112326424457871,
      "loss": 0.5069,
      "step": 2119
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011225406949398043,
      "loss": 0.5019,
      "step": 2120
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011218170801639997,
      "loss": 0.5055,
      "step": 2121
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011210934006359364,
      "loss": 0.5085,
      "step": 2122
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011203696567402879,
      "loss": 0.5099,
      "step": 2123
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011196458488617634,
      "loss": 0.5164,
      "step": 2124
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.00011189219773851047,
      "loss": 0.5046,
      "step": 2125
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011181980426950884,
      "loss": 0.5041,
      "step": 2126
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011174740451765239,
      "loss": 0.5044,
      "step": 2127
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011167499852142548,
      "loss": 0.504,
      "step": 2128
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011160258631931572,
      "loss": 0.5163,
      "step": 2129
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0001115301679498141,
      "loss": 0.516,
      "step": 2130
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011145774345141477,
      "loss": 0.5059,
      "step": 2131
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0001113853128626152,
      "loss": 0.5103,
      "step": 2132
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011131287622191617,
      "loss": 0.5038,
      "step": 2133
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011124043356782156,
      "loss": 0.5166,
      "step": 2134
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011116798493883851,
      "loss": 0.5097,
      "step": 2135
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011109553037347728,
      "loss": 0.5173,
      "step": 2136
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0001110230699102514,
      "loss": 0.5118,
      "step": 2137
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011095060358767741,
      "loss": 0.5102,
      "step": 2138
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011087813144427504,
      "loss": 0.5149,
      "step": 2139
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.00011080565351856707,
      "loss": 0.5132,
      "step": 2140
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011073316984907935,
      "loss": 0.509,
      "step": 2141
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011066068047434088,
      "loss": 0.5136,
      "step": 2142
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011058818543288356,
      "loss": 0.5,
      "step": 2143
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011051568476324239,
      "loss": 0.4949,
      "step": 2144
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011044317850395531,
      "loss": 0.5026,
      "step": 2145
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011037066669356326,
      "loss": 0.5024,
      "step": 2146
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011029814937061014,
      "loss": 0.4959,
      "step": 2147
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011022562657364275,
      "loss": 0.5048,
      "step": 2148
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011015309834121081,
      "loss": 0.5162,
      "step": 2149
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011008056471186697,
      "loss": 0.5154,
      "step": 2150
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00011000802572416666,
      "loss": 0.5111,
      "step": 2151
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00010993548141666825,
      "loss": 0.51,
      "step": 2152
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.0001098629318279329,
      "loss": 0.5071,
      "step": 2153
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00010979037699652455,
      "loss": 0.5042,
      "step": 2154
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010971781696100996,
      "loss": 0.5135,
      "step": 2155
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010964525175995865,
      "loss": 0.504,
      "step": 2156
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010957268143194292,
      "loss": 0.5048,
      "step": 2157
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010950010601553773,
      "loss": 0.5145,
      "step": 2158
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010942752554932075,
      "loss": 0.4921,
      "step": 2159
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.0001093549400718724,
      "loss": 0.5065,
      "step": 2160
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.0001092823496217757,
      "loss": 0.5001,
      "step": 2161
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010920975423761636,
      "loss": 0.5019,
      "step": 2162
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010913715395798267,
      "loss": 0.5095,
      "step": 2163
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010906454882146553,
      "loss": 0.5128,
      "step": 2164
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010899193886665844,
      "loss": 0.5014,
      "step": 2165
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010891932413215748,
      "loss": 0.5231,
      "step": 2166
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010884670465656122,
      "loss": 0.501,
      "step": 2167
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010877408047847076,
      "loss": 0.501,
      "step": 2168
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.00010870145163648975,
      "loss": 0.5113,
      "step": 2169
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010862881816922428,
      "loss": 0.5316,
      "step": 2170
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010855618011528289,
      "loss": 0.516,
      "step": 2171
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010848353751327656,
      "loss": 0.4997,
      "step": 2172
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010841089040181868,
      "loss": 0.5099,
      "step": 2173
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010833823881952511,
      "loss": 0.5112,
      "step": 2174
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010826558280501401,
      "loss": 0.5096,
      "step": 2175
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010819292239690588,
      "loss": 0.5021,
      "step": 2176
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010812025763382365,
      "loss": 0.5092,
      "step": 2177
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010804758855439246,
      "loss": 0.4991,
      "step": 2178
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010797491519723983,
      "loss": 0.5142,
      "step": 2179
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010790223760099549,
      "loss": 0.4975,
      "step": 2180
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010782955580429145,
      "loss": 0.5087,
      "step": 2181
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010775686984576196,
      "loss": 0.5134,
      "step": 2182
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.0001076841797640435,
      "loss": 0.5233,
      "step": 2183
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00010761148559777467,
      "loss": 0.518,
      "step": 2184
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010753878738559631,
      "loss": 0.517,
      "step": 2185
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010746608516615137,
      "loss": 0.5131,
      "step": 2186
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010739337897808495,
      "loss": 0.5081,
      "step": 2187
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010732066886004428,
      "loss": 0.5144,
      "step": 2188
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010724795485067864,
      "loss": 0.5028,
      "step": 2189
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010717523698863941,
      "loss": 0.5073,
      "step": 2190
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010710251531257995,
      "loss": 0.5135,
      "step": 2191
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010702978986115578,
      "loss": 0.5063,
      "step": 2192
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010695706067302429,
      "loss": 0.5034,
      "step": 2193
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010688432778684494,
      "loss": 0.5132,
      "step": 2194
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010681159124127911,
      "loss": 0.5085,
      "step": 2195
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010673885107499013,
      "loss": 0.5055,
      "step": 2196
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010666610732664333,
      "loss": 0.5228,
      "step": 2197
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0001065933600349058,
      "loss": 0.5134,
      "step": 2198
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.00010652060923844664,
      "loss": 0.504,
      "step": 2199
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010644785497593679,
      "loss": 0.5136,
      "step": 2200
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010637509728604898,
      "loss": 0.5075,
      "step": 2201
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010630233620745777,
      "loss": 0.512,
      "step": 2202
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010622957177883961,
      "loss": 0.5139,
      "step": 2203
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010615680403887262,
      "loss": 0.4922,
      "step": 2204
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010608403302623673,
      "loss": 0.5113,
      "step": 2205
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010601125877961362,
      "loss": 0.504,
      "step": 2206
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010593848133768664,
      "loss": 0.5193,
      "step": 2207
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010586570073914092,
      "loss": 0.5114,
      "step": 2208
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010579291702266316,
      "loss": 0.5009,
      "step": 2209
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010572013022694184,
      "loss": 0.5131,
      "step": 2210
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010564734039066699,
      "loss": 0.5043,
      "step": 2211
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010557454755253028,
      "loss": 0.5011,
      "step": 2212
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010550175175122497,
      "loss": 0.5122,
      "step": 2213
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00010542895302544587,
      "loss": 0.5033,
      "step": 2214
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010535615141388944,
      "loss": 0.5076,
      "step": 2215
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010528334695525359,
      "loss": 0.5123,
      "step": 2216
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010521053968823773,
      "loss": 0.5041,
      "step": 2217
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010513772965154283,
      "loss": 0.5037,
      "step": 2218
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010506491688387127,
      "loss": 0.5109,
      "step": 2219
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010499210142392694,
      "loss": 0.5239,
      "step": 2220
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010491928331041514,
      "loss": 0.5161,
      "step": 2221
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010484646258204254,
      "loss": 0.5102,
      "step": 2222
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.0001047736392775172,
      "loss": 0.497,
      "step": 2223
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010470081343554865,
      "loss": 0.5032,
      "step": 2224
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010462798509484768,
      "loss": 0.5137,
      "step": 2225
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010455515429412639,
      "loss": 0.5112,
      "step": 2226
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010448232107209828,
      "loss": 0.5025,
      "step": 2227
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00010440948546747802,
      "loss": 0.5161,
      "step": 2228
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010433664751898167,
      "loss": 0.5102,
      "step": 2229
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010426380726532645,
      "loss": 0.5126,
      "step": 2230
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010419096474523082,
      "loss": 0.5207,
      "step": 2231
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010411811999741445,
      "loss": 0.5067,
      "step": 2232
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010404527306059826,
      "loss": 0.4968,
      "step": 2233
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.0001039724239735042,
      "loss": 0.508,
      "step": 2234
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010389957277485545,
      "loss": 0.4958,
      "step": 2235
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010382671950337633,
      "loss": 0.5015,
      "step": 2236
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010375386419779217,
      "loss": 0.5104,
      "step": 2237
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010368100689682951,
      "loss": 0.5098,
      "step": 2238
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010360814763921585,
      "loss": 0.5123,
      "step": 2239
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010353528646367974,
      "loss": 0.5061,
      "step": 2240
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.0001034624234089508,
      "loss": 0.5121,
      "step": 2241
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010338955851375962,
      "loss": 0.5007,
      "step": 2242
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.00010331669181683777,
      "loss": 0.5107,
      "step": 2243
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010324382335691776,
      "loss": 0.5116,
      "step": 2244
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010317095317273306,
      "loss": 0.5106,
      "step": 2245
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010309808130301804,
      "loss": 0.5018,
      "step": 2246
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.000103025207786508,
      "loss": 0.5077,
      "step": 2247
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010295233266193906,
      "loss": 0.5155,
      "step": 2248
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010287945596804826,
      "loss": 0.5078,
      "step": 2249
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0001028065777435734,
      "loss": 0.5077,
      "step": 2250
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0001027336980272532,
      "loss": 0.4946,
      "step": 2251
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010266081685782701,
      "loss": 0.4994,
      "step": 2252
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010258793427403513,
      "loss": 0.5152,
      "step": 2253
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0001025150503146185,
      "loss": 0.5052,
      "step": 2254
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0001024421650183188,
      "loss": 0.5053,
      "step": 2255
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010236927842387848,
      "loss": 0.5183,
      "step": 2256
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010229639057004066,
      "loss": 0.5083,
      "step": 2257
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00010222350149554907,
      "loss": 0.5207,
      "step": 2258
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010215061123914811,
      "loss": 0.5109,
      "step": 2259
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.0001020777198395829,
      "loss": 0.505,
      "step": 2260
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010200482733559904,
      "loss": 0.5061,
      "step": 2261
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010193193376594279,
      "loss": 0.5164,
      "step": 2262
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010185903916936097,
      "loss": 0.516,
      "step": 2263
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.0001017861435846009,
      "loss": 0.4893,
      "step": 2264
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010171324705041052,
      "loss": 0.5185,
      "step": 2265
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010164034960553819,
      "loss": 0.499,
      "step": 2266
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010156745128873277,
      "loss": 0.5118,
      "step": 2267
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010149455213874362,
      "loss": 0.4929,
      "step": 2268
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010142165219432044,
      "loss": 0.5086,
      "step": 2269
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010134875149421352,
      "loss": 0.4967,
      "step": 2270
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010127585007717342,
      "loss": 0.5196,
      "step": 2271
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010120294798195114,
      "loss": 0.5052,
      "step": 2272
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.00010113004524729799,
      "loss": 0.512,
      "step": 2273
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010105714191196566,
      "loss": 0.5109,
      "step": 2274
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.0001009842380147062,
      "loss": 0.5161,
      "step": 2275
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010091133359427185,
      "loss": 0.5168,
      "step": 2276
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010083842868941522,
      "loss": 0.5036,
      "step": 2277
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010076552333888911,
      "loss": 0.5007,
      "step": 2278
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010069261758144664,
      "loss": 0.5191,
      "step": 2279
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010061971145584108,
      "loss": 0.5177,
      "step": 2280
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010054680500082592,
      "loss": 0.5211,
      "step": 2281
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010047389825515479,
      "loss": 0.5019,
      "step": 2282
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010040099125758154,
      "loss": 0.5223,
      "step": 2283
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.0001003280840468601,
      "loss": 0.5036,
      "step": 2284
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010025517666174453,
      "loss": 0.5129,
      "step": 2285
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.000100182269140989,
      "loss": 0.5229,
      "step": 2286
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010010936152334772,
      "loss": 0.4989,
      "step": 2287
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.00010003645384757496,
      "loss": 0.5189,
      "step": 2288
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.996354615242503e-05,
      "loss": 0.4954,
      "step": 2289
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.989063847665229e-05,
      "loss": 0.5068,
      "step": 2290
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.981773085901102e-05,
      "loss": 0.5011,
      "step": 2291
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.974482333825548e-05,
      "loss": 0.5054,
      "step": 2292
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.967191595313992e-05,
      "loss": 0.4901,
      "step": 2293
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.959900874241848e-05,
      "loss": 0.5071,
      "step": 2294
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.952610174484524e-05,
      "loss": 0.5017,
      "step": 2295
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.945319499917413e-05,
      "loss": 0.5052,
      "step": 2296
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.938028854415894e-05,
      "loss": 0.5105,
      "step": 2297
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.930738241855336e-05,
      "loss": 0.5089,
      "step": 2298
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.923447666111088e-05,
      "loss": 0.51,
      "step": 2299
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.916157131058481e-05,
      "loss": 0.5094,
      "step": 2300
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.908866640572818e-05,
      "loss": 0.5071,
      "step": 2301
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.901576198529382e-05,
      "loss": 0.5136,
      "step": 2302
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.894285808803435e-05,
      "loss": 0.5059,
      "step": 2303
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.886995475270205e-05,
      "loss": 0.4945,
      "step": 2304
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.87970520180489e-05,
      "loss": 0.5172,
      "step": 2305
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.872414992282662e-05,
      "loss": 0.501,
      "step": 2306
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.865124850578648e-05,
      "loss": 0.515,
      "step": 2307
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.857834780567957e-05,
      "loss": 0.5065,
      "step": 2308
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.850544786125642e-05,
      "loss": 0.4992,
      "step": 2309
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.843254871126724e-05,
      "loss": 0.5172,
      "step": 2310
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.835965039446182e-05,
      "loss": 0.508,
      "step": 2311
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.828675294958949e-05,
      "loss": 0.5086,
      "step": 2312
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.82138564153991e-05,
      "loss": 0.5018,
      "step": 2313
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.814096083063907e-05,
      "loss": 0.505,
      "step": 2314
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.806806623405724e-05,
      "loss": 0.5037,
      "step": 2315
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.799517266440096e-05,
      "loss": 0.4988,
      "step": 2316
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.792228016041711e-05,
      "loss": 0.4896,
      "step": 2317
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.784938876085188e-05,
      "loss": 0.5108,
      "step": 2318
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.777649850445096e-05,
      "loss": 0.5079,
      "step": 2319
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.770360942995935e-05,
      "loss": 0.5093,
      "step": 2320
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.763072157612153e-05,
      "loss": 0.5099,
      "step": 2321
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.755783498168122e-05,
      "loss": 0.51,
      "step": 2322
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.748494968538153e-05,
      "loss": 0.5176,
      "step": 2323
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.74120657259649e-05,
      "loss": 0.5033,
      "step": 2324
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.7339183142173e-05,
      "loss": 0.4907,
      "step": 2325
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.726630197274683e-05,
      "loss": 0.5019,
      "step": 2326
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.719342225642662e-05,
      "loss": 0.5066,
      "step": 2327
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.712054403195175e-05,
      "loss": 0.5211,
      "step": 2328
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.704766733806094e-05,
      "loss": 0.5227,
      "step": 2329
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.697479221349203e-05,
      "loss": 0.5089,
      "step": 2330
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.690191869698199e-05,
      "loss": 0.4964,
      "step": 2331
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.682904682726698e-05,
      "loss": 0.5155,
      "step": 2332
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.675617664308229e-05,
      "loss": 0.503,
      "step": 2333
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.668330818316228e-05,
      "loss": 0.5064,
      "step": 2334
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.661044148624037e-05,
      "loss": 0.4937,
      "step": 2335
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.653757659104921e-05,
      "loss": 0.511,
      "step": 2336
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.646471353632027e-05,
      "loss": 0.5092,
      "step": 2337
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.639185236078419e-05,
      "loss": 0.5062,
      "step": 2338
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.631899310317052e-05,
      "loss": 0.5172,
      "step": 2339
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.624613580220786e-05,
      "loss": 0.5162,
      "step": 2340
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.617328049662372e-05,
      "loss": 0.5125,
      "step": 2341
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.610042722514459e-05,
      "loss": 0.5106,
      "step": 2342
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.602757602649586e-05,
      "loss": 0.5028,
      "step": 2343
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.595472693940175e-05,
      "loss": 0.5084,
      "step": 2344
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.588188000258553e-05,
      "loss": 0.5142,
      "step": 2345
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.580903525476919e-05,
      "loss": 0.5128,
      "step": 2346
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.573619273467356e-05,
      "loss": 0.5131,
      "step": 2347
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.566335248101835e-05,
      "loss": 0.5079,
      "step": 2348
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.5590514532522e-05,
      "loss": 0.5084,
      "step": 2349
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.551767892790176e-05,
      "loss": 0.5187,
      "step": 2350
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.544484570587363e-05,
      "loss": 0.5066,
      "step": 2351
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.537201490515236e-05,
      "loss": 0.513,
      "step": 2352
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.529918656445137e-05,
      "loss": 0.5039,
      "step": 2353
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.522636072248282e-05,
      "loss": 0.5013,
      "step": 2354
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.51535374179575e-05,
      "loss": 0.513,
      "step": 2355
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.508071668958488e-05,
      "loss": 0.509,
      "step": 2356
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.500789857607307e-05,
      "loss": 0.5,
      "step": 2357
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.493508311612874e-05,
      "loss": 0.5063,
      "step": 2358
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.48622703484572e-05,
      "loss": 0.5016,
      "step": 2359
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.47894603117623e-05,
      "loss": 0.499,
      "step": 2360
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.471665304474646e-05,
      "loss": 0.511,
      "step": 2361
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.464384858611055e-05,
      "loss": 0.5129,
      "step": 2362
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.457104697455413e-05,
      "loss": 0.5039,
      "step": 2363
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.449824824877505e-05,
      "loss": 0.5065,
      "step": 2364
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.442545244746974e-05,
      "loss": 0.5089,
      "step": 2365
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.435265960933302e-05,
      "loss": 0.497,
      "step": 2366
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.427986977305817e-05,
      "loss": 0.4936,
      "step": 2367
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.420708297733686e-05,
      "loss": 0.5133,
      "step": 2368
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.413429926085911e-05,
      "loss": 0.5045,
      "step": 2369
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.40615186623134e-05,
      "loss": 0.5148,
      "step": 2370
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.398874122038639e-05,
      "loss": 0.5101,
      "step": 2371
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.391596697376327e-05,
      "loss": 0.4962,
      "step": 2372
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.38431959611274e-05,
      "loss": 0.4933,
      "step": 2373
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.37704282211604e-05,
      "loss": 0.5023,
      "step": 2374
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.369766379254223e-05,
      "loss": 0.5077,
      "step": 2375
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.362490271395103e-05,
      "loss": 0.5133,
      "step": 2376
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.355214502406324e-05,
      "loss": 0.5082,
      "step": 2377
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.347939076155338e-05,
      "loss": 0.4959,
      "step": 2378
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.340663996509423e-05,
      "loss": 0.5113,
      "step": 2379
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.33338926733567e-05,
      "loss": 0.4931,
      "step": 2380
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.326114892500986e-05,
      "loss": 0.501,
      "step": 2381
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.31884087587209e-05,
      "loss": 0.5061,
      "step": 2382
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.311567221315509e-05,
      "loss": 0.5131,
      "step": 2383
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.304293932697572e-05,
      "loss": 0.5052,
      "step": 2384
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.297021013884423e-05,
      "loss": 0.5008,
      "step": 2385
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.289748468742006e-05,
      "loss": 0.5087,
      "step": 2386
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.282476301136064e-05,
      "loss": 0.5089,
      "step": 2387
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.27520451493214e-05,
      "loss": 0.5032,
      "step": 2388
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.267933113995572e-05,
      "loss": 0.516,
      "step": 2389
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.260662102191505e-05,
      "loss": 0.4952,
      "step": 2390
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.253391483384865e-05,
      "loss": 0.5182,
      "step": 2391
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.246121261440372e-05,
      "loss": 0.5011,
      "step": 2392
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.238851440222536e-05,
      "loss": 0.5145,
      "step": 2393
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.231582023595653e-05,
      "loss": 0.4838,
      "step": 2394
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.224313015423806e-05,
      "loss": 0.5017,
      "step": 2395
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.217044419570857e-05,
      "loss": 0.493,
      "step": 2396
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.209776239900453e-05,
      "loss": 0.5056,
      "step": 2397
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.202508480276018e-05,
      "loss": 0.513,
      "step": 2398
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.195241144560754e-05,
      "loss": 0.5021,
      "step": 2399
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.187974236617636e-05,
      "loss": 0.5101,
      "step": 2400
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.180707760309414e-05,
      "loss": 0.5062,
      "step": 2401
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.173441719498601e-05,
      "loss": 0.5079,
      "step": 2402
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.166176118047491e-05,
      "loss": 0.5006,
      "step": 2403
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.158910959818133e-05,
      "loss": 0.4974,
      "step": 2404
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.151646248672349e-05,
      "loss": 0.515,
      "step": 2405
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.144381988471715e-05,
      "loss": 0.5086,
      "step": 2406
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.137118183077573e-05,
      "loss": 0.5012,
      "step": 2407
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.129854836351025e-05,
      "loss": 0.5025,
      "step": 2408
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.122591952152924e-05,
      "loss": 0.5069,
      "step": 2409
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.115329534343879e-05,
      "loss": 0.5147,
      "step": 2410
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.108067586784253e-05,
      "loss": 0.5084,
      "step": 2411
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.100806113334157e-05,
      "loss": 0.5176,
      "step": 2412
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.09354511785345e-05,
      "loss": 0.4864,
      "step": 2413
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.086284604201738e-05,
      "loss": 0.516,
      "step": 2414
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.079024576238369e-05,
      "loss": 0.5034,
      "step": 2415
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.071765037822435e-05,
      "loss": 0.5029,
      "step": 2416
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.06450599281276e-05,
      "loss": 0.5082,
      "step": 2417
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.057247445067925e-05,
      "loss": 0.5064,
      "step": 2418
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.049989398446229e-05,
      "loss": 0.5067,
      "step": 2419
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.04273185680571e-05,
      "loss": 0.4952,
      "step": 2420
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.035474824004136e-05,
      "loss": 0.4991,
      "step": 2421
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.028218303899007e-05,
      "loss": 0.5179,
      "step": 2422
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.02096230034755e-05,
      "loss": 0.5045,
      "step": 2423
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.013706817206713e-05,
      "loss": 0.5061,
      "step": 2424
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.006451858333176e-05,
      "loss": 0.5027,
      "step": 2425
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.999197427583335e-05,
      "loss": 0.5108,
      "step": 2426
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.991943528813306e-05,
      "loss": 0.5045,
      "step": 2427
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.984690165878921e-05,
      "loss": 0.5168,
      "step": 2428
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.977437342635726e-05,
      "loss": 0.4944,
      "step": 2429
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.970185062938988e-05,
      "loss": 0.4996,
      "step": 2430
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.962933330643676e-05,
      "loss": 0.5083,
      "step": 2431
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.955682149604471e-05,
      "loss": 0.5185,
      "step": 2432
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.948431523675765e-05,
      "loss": 0.5036,
      "step": 2433
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.941181456711646e-05,
      "loss": 0.5107,
      "step": 2434
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.933931952565911e-05,
      "loss": 0.504,
      "step": 2435
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.926683015092063e-05,
      "loss": 0.5127,
      "step": 2436
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.919434648143294e-05,
      "loss": 0.511,
      "step": 2437
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.912186855572497e-05,
      "loss": 0.5089,
      "step": 2438
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.90493964123226e-05,
      "loss": 0.5153,
      "step": 2439
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.897693008974861e-05,
      "loss": 0.5168,
      "step": 2440
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.890446962652273e-05,
      "loss": 0.5129,
      "step": 2441
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.883201506116154e-05,
      "loss": 0.5213,
      "step": 2442
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.875956643217847e-05,
      "loss": 0.5033,
      "step": 2443
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.868712377808382e-05,
      "loss": 0.5016,
      "step": 2444
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.86146871373848e-05,
      "loss": 0.4913,
      "step": 2445
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.854225654858525e-05,
      "loss": 0.5053,
      "step": 2446
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.846983205018594e-05,
      "loss": 0.4961,
      "step": 2447
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.839741368068429e-05,
      "loss": 0.5075,
      "step": 2448
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.832500147857454e-05,
      "loss": 0.5175,
      "step": 2449
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.825259548234762e-05,
      "loss": 0.5092,
      "step": 2450
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.81801957304912e-05,
      "loss": 0.5057,
      "step": 2451
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.810780226148956e-05,
      "loss": 0.5061,
      "step": 2452
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.803541511382368e-05,
      "loss": 0.5111,
      "step": 2453
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.796303432597122e-05,
      "loss": 0.5201,
      "step": 2454
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.789065993640639e-05,
      "loss": 0.5149,
      "step": 2455
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.781829198360004e-05,
      "loss": 0.5105,
      "step": 2456
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.774593050601959e-05,
      "loss": 0.5112,
      "step": 2457
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.767357554212901e-05,
      "loss": 0.5137,
      "step": 2458
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.760122713038881e-05,
      "loss": 0.5117,
      "step": 2459
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.752888530925602e-05,
      "loss": 0.5192,
      "step": 2460
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.745655011718416e-05,
      "loss": 0.5207,
      "step": 2461
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.73842215926232e-05,
      "loss": 0.4901,
      "step": 2462
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.731189977401967e-05,
      "loss": 0.5176,
      "step": 2463
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.723958469981645e-05,
      "loss": 0.5076,
      "step": 2464
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.716727640845282e-05,
      "loss": 0.499,
      "step": 2465
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.70949749383645e-05,
      "loss": 0.499,
      "step": 2466
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.702268032798356e-05,
      "loss": 0.5105,
      "step": 2467
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.695039261573844e-05,
      "loss": 0.5119,
      "step": 2468
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.68781118400539e-05,
      "loss": 0.505,
      "step": 2469
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.6805838039351e-05,
      "loss": 0.4981,
      "step": 2470
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.67335712520471e-05,
      "loss": 0.5055,
      "step": 2471
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.666131151655592e-05,
      "loss": 0.503,
      "step": 2472
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.65890588712873e-05,
      "loss": 0.5124,
      "step": 2473
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.651681335464737e-05,
      "loss": 0.522,
      "step": 2474
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.644457500503844e-05,
      "loss": 0.4984,
      "step": 2475
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.637234386085906e-05,
      "loss": 0.5107,
      "step": 2476
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.630011996050394e-05,
      "loss": 0.5016,
      "step": 2477
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.622790334236391e-05,
      "loss": 0.504,
      "step": 2478
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.615569404482596e-05,
      "loss": 0.5007,
      "step": 2479
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.608349210627313e-05,
      "loss": 0.4979,
      "step": 2480
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.601129756508462e-05,
      "loss": 0.5079,
      "step": 2481
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.593911045963568e-05,
      "loss": 0.4984,
      "step": 2482
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.58669308282976e-05,
      "loss": 0.5012,
      "step": 2483
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.579475870943767e-05,
      "loss": 0.5,
      "step": 2484
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.572259414141923e-05,
      "loss": 0.5071,
      "step": 2485
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.565043716260158e-05,
      "loss": 0.5195,
      "step": 2486
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.557828781133999e-05,
      "loss": 0.5041,
      "step": 2487
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.550614612598568e-05,
      "loss": 0.5018,
      "step": 2488
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.543401214488575e-05,
      "loss": 0.5126,
      "step": 2489
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.536188590638334e-05,
      "loss": 0.4992,
      "step": 2490
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.528976744881732e-05,
      "loss": 0.5059,
      "step": 2491
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.521765681052252e-05,
      "loss": 0.5157,
      "step": 2492
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.514555402982957e-05,
      "loss": 0.5011,
      "step": 2493
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.507345914506492e-05,
      "loss": 0.4963,
      "step": 2494
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.500137219455085e-05,
      "loss": 0.5062,
      "step": 2495
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.492929321660539e-05,
      "loss": 0.5188,
      "step": 2496
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.485722224954237e-05,
      "loss": 0.4985,
      "step": 2497
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.478515933167132e-05,
      "loss": 0.5081,
      "step": 2498
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.471310450129752e-05,
      "loss": 0.5024,
      "step": 2499
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.4641057796722e-05,
      "loss": 0.4946,
      "step": 2500
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.456901925624137e-05,
      "loss": 0.493,
      "step": 2501
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.449698891814794e-05,
      "loss": 0.5007,
      "step": 2502
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.442496682072968e-05,
      "loss": 0.5074,
      "step": 2503
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.435295300227018e-05,
      "loss": 0.5147,
      "step": 2504
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.428094750104858e-05,
      "loss": 0.487,
      "step": 2505
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.420895035533968e-05,
      "loss": 0.5193,
      "step": 2506
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.413696160341375e-05,
      "loss": 0.5033,
      "step": 2507
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.406498128353663e-05,
      "loss": 0.5083,
      "step": 2508
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.399300943396977e-05,
      "loss": 0.5037,
      "step": 2509
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.392104609296997e-05,
      "loss": 0.4949,
      "step": 2510
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.384909129878959e-05,
      "loss": 0.4948,
      "step": 2511
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.377714508967645e-05,
      "loss": 0.501,
      "step": 2512
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.370520750387376e-05,
      "loss": 0.4919,
      "step": 2513
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.363327857962019e-05,
      "loss": 0.5016,
      "step": 2514
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.356135835514978e-05,
      "loss": 0.5108,
      "step": 2515
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.348944686869195e-05,
      "loss": 0.5014,
      "step": 2516
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.341754415847147e-05,
      "loss": 0.5075,
      "step": 2517
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.33456502627085e-05,
      "loss": 0.5073,
      "step": 2518
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.327376521961847e-05,
      "loss": 0.496,
      "step": 2519
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.320188906741205e-05,
      "loss": 0.5188,
      "step": 2520
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.313002184429529e-05,
      "loss": 0.4938,
      "step": 2521
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.305816358846942e-05,
      "loss": 0.502,
      "step": 2522
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.298631433813093e-05,
      "loss": 0.498,
      "step": 2523
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.291447413147148e-05,
      "loss": 0.513,
      "step": 2524
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.284264300667804e-05,
      "loss": 0.4919,
      "step": 2525
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.277082100193261e-05,
      "loss": 0.5034,
      "step": 2526
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.269900815541245e-05,
      "loss": 0.4982,
      "step": 2527
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.262720450528989e-05,
      "loss": 0.5052,
      "step": 2528
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.255541008973238e-05,
      "loss": 0.5066,
      "step": 2529
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.248362494690249e-05,
      "loss": 0.5131,
      "step": 2530
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.241184911495782e-05,
      "loss": 0.5016,
      "step": 2531
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.234008263205108e-05,
      "loss": 0.4908,
      "step": 2532
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.226832553632993e-05,
      "loss": 0.4996,
      "step": 2533
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.21965778659371e-05,
      "loss": 0.5078,
      "step": 2534
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.212483965901028e-05,
      "loss": 0.4927,
      "step": 2535
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.205311095368219e-05,
      "loss": 0.5069,
      "step": 2536
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.198139178808043e-05,
      "loss": 0.5133,
      "step": 2537
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.190968220032754e-05,
      "loss": 0.4999,
      "step": 2538
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.183798222854098e-05,
      "loss": 0.5142,
      "step": 2539
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.176629191083313e-05,
      "loss": 0.5128,
      "step": 2540
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.169461128531117e-05,
      "loss": 0.5007,
      "step": 2541
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.162294039007717e-05,
      "loss": 0.5195,
      "step": 2542
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.155127926322805e-05,
      "loss": 0.509,
      "step": 2543
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.147962794285545e-05,
      "loss": 0.5032,
      "step": 2544
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.140798646704595e-05,
      "loss": 0.5124,
      "step": 2545
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.133635487388075e-05,
      "loss": 0.5031,
      "step": 2546
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.126473320143587e-05,
      "loss": 0.5124,
      "step": 2547
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.119312148778203e-05,
      "loss": 0.5026,
      "step": 2548
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.112151977098467e-05,
      "loss": 0.5029,
      "step": 2549
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.104992808910389e-05,
      "loss": 0.5139,
      "step": 2550
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.097834648019451e-05,
      "loss": 0.502,
      "step": 2551
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.090677498230596e-05,
      "loss": 0.5078,
      "step": 2552
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.083521363348228e-05,
      "loss": 0.5128,
      "step": 2553
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.076366247176215e-05,
      "loss": 0.5039,
      "step": 2554
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.069212153517878e-05,
      "loss": 0.4989,
      "step": 2555
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.062059086176004e-05,
      "loss": 0.5045,
      "step": 2556
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.05490704895283e-05,
      "loss": 0.5035,
      "step": 2557
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.047756045650036e-05,
      "loss": 0.5127,
      "step": 2558
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.040606080068769e-05,
      "loss": 0.5015,
      "step": 2559
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.033457156009611e-05,
      "loss": 0.5037,
      "step": 2560
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.026309277272597e-05,
      "loss": 0.4912,
      "step": 2561
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.019162447657201e-05,
      "loss": 0.5012,
      "step": 2562
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.012016670962351e-05,
      "loss": 0.5103,
      "step": 2563
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.004871950986403e-05,
      "loss": 0.5143,
      "step": 2564
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.997728291527158e-05,
      "loss": 0.5057,
      "step": 2565
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.990585696381849e-05,
      "loss": 0.5061,
      "step": 2566
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.983444169347144e-05,
      "loss": 0.5139,
      "step": 2567
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.976303714219146e-05,
      "loss": 0.4939,
      "step": 2568
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.969164334793389e-05,
      "loss": 0.5091,
      "step": 2569
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.96202603486483e-05,
      "loss": 0.5061,
      "step": 2570
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.954888818227853e-05,
      "loss": 0.4998,
      "step": 2571
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.947752688676273e-05,
      "loss": 0.5124,
      "step": 2572
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.940617650003324e-05,
      "loss": 0.5043,
      "step": 2573
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.933483706001654e-05,
      "loss": 0.5055,
      "step": 2574
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.926350860463335e-05,
      "loss": 0.523,
      "step": 2575
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.919219117179854e-05,
      "loss": 0.5091,
      "step": 2576
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.912088479942112e-05,
      "loss": 0.507,
      "step": 2577
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.904958952540425e-05,
      "loss": 0.4998,
      "step": 2578
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.897830538764512e-05,
      "loss": 0.5028,
      "step": 2579
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.890703242403505e-05,
      "loss": 0.513,
      "step": 2580
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.883577067245939e-05,
      "loss": 0.5055,
      "step": 2581
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.876452017079761e-05,
      "loss": 0.5041,
      "step": 2582
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.869328095692312e-05,
      "loss": 0.5038,
      "step": 2583
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.862205306870334e-05,
      "loss": 0.508,
      "step": 2584
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.855083654399968e-05,
      "loss": 0.494,
      "step": 2585
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.847963142066753e-05,
      "loss": 0.5001,
      "step": 2586
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.840843773655619e-05,
      "loss": 0.5033,
      "step": 2587
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.83372555295089e-05,
      "loss": 0.4934,
      "step": 2588
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.826608483736275e-05,
      "loss": 0.503,
      "step": 2589
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.819492569794875e-05,
      "loss": 0.4989,
      "step": 2590
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.812377814909186e-05,
      "loss": 0.4951,
      "step": 2591
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.805264222861071e-05,
      "loss": 0.494,
      "step": 2592
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.798151797431783e-05,
      "loss": 0.4971,
      "step": 2593
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.791040542401957e-05,
      "loss": 0.4958,
      "step": 2594
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.783930461551602e-05,
      "loss": 0.5115,
      "step": 2595
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.776821558660104e-05,
      "loss": 0.495,
      "step": 2596
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.769713837506223e-05,
      "loss": 0.5006,
      "step": 2597
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.76260730186809e-05,
      "loss": 0.5109,
      "step": 2598
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.755501955523207e-05,
      "loss": 0.4891,
      "step": 2599
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.748397802248447e-05,
      "loss": 0.4998,
      "step": 2600
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.741294845820044e-05,
      "loss": 0.5045,
      "step": 2601
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.734193090013595e-05,
      "loss": 0.5066,
      "step": 2602
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.727092538604062e-05,
      "loss": 0.507,
      "step": 2603
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.719993195365767e-05,
      "loss": 0.504,
      "step": 2604
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.712895064072388e-05,
      "loss": 0.5036,
      "step": 2605
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.705798148496958e-05,
      "loss": 0.52,
      "step": 2606
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.698702452411868e-05,
      "loss": 0.5085,
      "step": 2607
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.69160797958885e-05,
      "loss": 0.4965,
      "step": 2608
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.684514733799008e-05,
      "loss": 0.4867,
      "step": 2609
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.677422718812769e-05,
      "loss": 0.5004,
      "step": 2610
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.670331938399919e-05,
      "loss": 0.5036,
      "step": 2611
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.663242396329585e-05,
      "loss": 0.5055,
      "step": 2612
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.656154096370236e-05,
      "loss": 0.5037,
      "step": 2613
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.64906704228968e-05,
      "loss": 0.4932,
      "step": 2614
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.641981237855065e-05,
      "loss": 0.5015,
      "step": 2615
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.63489668683287e-05,
      "loss": 0.4986,
      "step": 2616
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.627813392988907e-05,
      "loss": 0.4994,
      "step": 2617
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.620731360088335e-05,
      "loss": 0.5089,
      "step": 2618
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.613650591895624e-05,
      "loss": 0.5118,
      "step": 2619
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.606571092174581e-05,
      "loss": 0.5033,
      "step": 2620
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.599492864688335e-05,
      "loss": 0.5053,
      "step": 2621
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.592415913199342e-05,
      "loss": 0.5019,
      "step": 2622
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.585340241469376e-05,
      "loss": 0.504,
      "step": 2623
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.578265853259534e-05,
      "loss": 0.4848,
      "step": 2624
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.571192752330231e-05,
      "loss": 0.5027,
      "step": 2625
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.564120942441195e-05,
      "loss": 0.5068,
      "step": 2626
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.55705042735147e-05,
      "loss": 0.5069,
      "step": 2627
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.549981210819413e-05,
      "loss": 0.5044,
      "step": 2628
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.542913296602683e-05,
      "loss": 0.5036,
      "step": 2629
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.53584668845826e-05,
      "loss": 0.5008,
      "step": 2630
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.528781390142415e-05,
      "loss": 0.5088,
      "step": 2631
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.521717405410734e-05,
      "loss": 0.4924,
      "step": 2632
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.514654738018099e-05,
      "loss": 0.506,
      "step": 2633
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.507593391718696e-05,
      "loss": 0.4946,
      "step": 2634
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.500533370265999e-05,
      "loss": 0.5148,
      "step": 2635
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.493474677412794e-05,
      "loss": 0.5014,
      "step": 2636
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.486417316911149e-05,
      "loss": 0.5162,
      "step": 2637
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.479361292512426e-05,
      "loss": 0.5009,
      "step": 2638
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.472306607967275e-05,
      "loss": 0.4972,
      "step": 2639
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.46525326702564e-05,
      "loss": 0.4967,
      "step": 2640
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.458201273436745e-05,
      "loss": 0.5033,
      "step": 2641
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.4511506309491e-05,
      "loss": 0.4985,
      "step": 2642
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.444101343310496e-05,
      "loss": 0.5015,
      "step": 2643
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.437053414268e-05,
      "loss": 0.5005,
      "step": 2644
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.430006847567972e-05,
      "loss": 0.5162,
      "step": 2645
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.42296164695603e-05,
      "loss": 0.5034,
      "step": 2646
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.415917816177076e-05,
      "loss": 0.5011,
      "step": 2647
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.408875358975277e-05,
      "loss": 0.5018,
      "step": 2648
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.401834279094077e-05,
      "loss": 0.5023,
      "step": 2649
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.39479458027618e-05,
      "loss": 0.5045,
      "step": 2650
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.387756266263565e-05,
      "loss": 0.5025,
      "step": 2651
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.380719340797471e-05,
      "loss": 0.5051,
      "step": 2652
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.373683807618394e-05,
      "loss": 0.492,
      "step": 2653
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.366649670466099e-05,
      "loss": 0.5126,
      "step": 2654
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.359616933079597e-05,
      "loss": 0.4904,
      "step": 2655
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.35258559919717e-05,
      "loss": 0.5178,
      "step": 2656
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.345555672556343e-05,
      "loss": 0.5096,
      "step": 2657
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.338527156893896e-05,
      "loss": 0.5059,
      "step": 2658
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.331500055945858e-05,
      "loss": 0.4908,
      "step": 2659
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.324474373447508e-05,
      "loss": 0.4996,
      "step": 2660
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.31745011313337e-05,
      "loss": 0.4955,
      "step": 2661
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.31042727873721e-05,
      "loss": 0.4974,
      "step": 2662
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.303405873992038e-05,
      "loss": 0.4888,
      "step": 2663
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.29638590263011e-05,
      "loss": 0.4987,
      "step": 2664
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.28936736838291e-05,
      "loss": 0.503,
      "step": 2665
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.282350274981161e-05,
      "loss": 0.4999,
      "step": 2666
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.275334626154824e-05,
      "loss": 0.5016,
      "step": 2667
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.268320425633089e-05,
      "loss": 0.4949,
      "step": 2668
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.261307677144377e-05,
      "loss": 0.5078,
      "step": 2669
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.254296384416335e-05,
      "loss": 0.4975,
      "step": 2670
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.247286551175841e-05,
      "loss": 0.4901,
      "step": 2671
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.240278181148987e-05,
      "loss": 0.4924,
      "step": 2672
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.233271278061106e-05,
      "loss": 0.4997,
      "step": 2673
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.226265845636733e-05,
      "loss": 0.4983,
      "step": 2674
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.219261887599628e-05,
      "loss": 0.5145,
      "step": 2675
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.21225940767277e-05,
      "loss": 0.5102,
      "step": 2676
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.205258409578347e-05,
      "loss": 0.4904,
      "step": 2677
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.198258897037761e-05,
      "loss": 0.5022,
      "step": 2678
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.191260873771632e-05,
      "loss": 0.5089,
      "step": 2679
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.184264343499774e-05,
      "loss": 0.4929,
      "step": 2680
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.177269309941216e-05,
      "loss": 0.4976,
      "step": 2681
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.170275776814192e-05,
      "loss": 0.492,
      "step": 2682
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.163283747836142e-05,
      "loss": 0.5094,
      "step": 2683
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.156293226723695e-05,
      "loss": 0.4995,
      "step": 2684
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.149304217192687e-05,
      "loss": 0.5056,
      "step": 2685
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.142316722958147e-05,
      "loss": 0.5101,
      "step": 2686
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.135330747734302e-05,
      "loss": 0.4913,
      "step": 2687
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.128346295234569e-05,
      "loss": 0.498,
      "step": 2688
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.121363369171554e-05,
      "loss": 0.5077,
      "step": 2689
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.114381973257051e-05,
      "loss": 0.4993,
      "step": 2690
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.107402111202053e-05,
      "loss": 0.5056,
      "step": 2691
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.100423786716723e-05,
      "loss": 0.5045,
      "step": 2692
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.09344700351041e-05,
      "loss": 0.5102,
      "step": 2693
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.086471765291647e-05,
      "loss": 0.5048,
      "step": 2694
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.079498075768145e-05,
      "loss": 0.5139,
      "step": 2695
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.072525938646788e-05,
      "loss": 0.488,
      "step": 2696
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.065555357633642e-05,
      "loss": 0.5088,
      "step": 2697
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.058586336433936e-05,
      "loss": 0.502,
      "step": 2698
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.051618878752079e-05,
      "loss": 0.5018,
      "step": 2699
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.044652988291647e-05,
      "loss": 0.4888,
      "step": 2700
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.037688668755382e-05,
      "loss": 0.503,
      "step": 2701
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.030725923845185e-05,
      "loss": 0.4876,
      "step": 2702
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.023764757262128e-05,
      "loss": 0.5109,
      "step": 2703
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.016805172706444e-05,
      "loss": 0.495,
      "step": 2704
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.009847173877519e-05,
      "loss": 0.5025,
      "step": 2705
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.002890764473903e-05,
      "loss": 0.4996,
      "step": 2706
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.995935948193294e-05,
      "loss": 0.4973,
      "step": 2707
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.988982728732544e-05,
      "loss": 0.4842,
      "step": 2708
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.982031109787669e-05,
      "loss": 0.5044,
      "step": 2709
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.975081095053817e-05,
      "loss": 0.5158,
      "step": 2710
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.968132688225294e-05,
      "loss": 0.5084,
      "step": 2711
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.961185892995545e-05,
      "loss": 0.5098,
      "step": 2712
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.954240713057163e-05,
      "loss": 0.4974,
      "step": 2713
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.94729715210188e-05,
      "loss": 0.513,
      "step": 2714
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.940355213820568e-05,
      "loss": 0.5078,
      "step": 2715
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.933414901903237e-05,
      "loss": 0.4983,
      "step": 2716
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.926476220039027e-05,
      "loss": 0.5072,
      "step": 2717
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.919539171916227e-05,
      "loss": 0.4874,
      "step": 2718
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.912603761222238e-05,
      "loss": 0.5007,
      "step": 2719
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.905669991643606e-05,
      "loss": 0.51,
      "step": 2720
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.898737866865994e-05,
      "loss": 0.4929,
      "step": 2721
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.891807390574199e-05,
      "loss": 0.5142,
      "step": 2722
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.884878566452133e-05,
      "loss": 0.4999,
      "step": 2723
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.877951398182836e-05,
      "loss": 0.5055,
      "step": 2724
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.871025889448469e-05,
      "loss": 0.4948,
      "step": 2725
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.864102043930305e-05,
      "loss": 0.4934,
      "step": 2726
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.857179865308742e-05,
      "loss": 0.5012,
      "step": 2727
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.850259357263281e-05,
      "loss": 0.4902,
      "step": 2728
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.843340523472539e-05,
      "loss": 0.4994,
      "step": 2729
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.83642336761425e-05,
      "loss": 0.4926,
      "step": 2730
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.829507893365247e-05,
      "loss": 0.5114,
      "step": 2731
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.822594104401472e-05,
      "loss": 0.5048,
      "step": 2732
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.815682004397973e-05,
      "loss": 0.4974,
      "step": 2733
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.808771597028898e-05,
      "loss": 0.5097,
      "step": 2734
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.801862885967493e-05,
      "loss": 0.494,
      "step": 2735
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.794955874886111e-05,
      "loss": 0.5099,
      "step": 2736
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.788050567456196e-05,
      "loss": 0.4871,
      "step": 2737
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.781146967348284e-05,
      "loss": 0.5075,
      "step": 2738
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.774245078232004e-05,
      "loss": 0.5103,
      "step": 2739
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.767344903776079e-05,
      "loss": 0.5068,
      "step": 2740
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.760446447648321e-05,
      "loss": 0.503,
      "step": 2741
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.753549713515621e-05,
      "loss": 0.5091,
      "step": 2742
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.746654705043966e-05,
      "loss": 0.513,
      "step": 2743
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.739761425898412e-05,
      "loss": 0.5063,
      "step": 2744
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.732869879743114e-05,
      "loss": 0.4925,
      "step": 2745
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.725980070241289e-05,
      "loss": 0.4998,
      "step": 2746
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.719092001055239e-05,
      "loss": 0.5024,
      "step": 2747
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.712205675846337e-05,
      "loss": 0.5121,
      "step": 2748
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.705321098275035e-05,
      "loss": 0.4961,
      "step": 2749
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.698438272000848e-05,
      "loss": 0.4979,
      "step": 2750
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.691557200682368e-05,
      "loss": 0.4931,
      "step": 2751
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.684677887977249e-05,
      "loss": 0.4996,
      "step": 2752
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.677800337542211e-05,
      "loss": 0.5036,
      "step": 2753
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.670924553033038e-05,
      "loss": 0.5009,
      "step": 2754
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.664050538104575e-05,
      "loss": 0.4928,
      "step": 2755
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.657178296410728e-05,
      "loss": 0.5005,
      "step": 2756
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.65030783160446e-05,
      "loss": 0.5064,
      "step": 2757
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.643439147337789e-05,
      "loss": 0.5024,
      "step": 2758
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.636572247261783e-05,
      "loss": 0.5081,
      "step": 2759
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.629707135026565e-05,
      "loss": 0.4881,
      "step": 2760
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.62284381428131e-05,
      "loss": 0.5019,
      "step": 2761
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.615982288674235e-05,
      "loss": 0.4938,
      "step": 2762
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.609122561852603e-05,
      "loss": 0.5036,
      "step": 2763
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.602264637462734e-05,
      "loss": 0.4913,
      "step": 2764
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.595408519149973e-05,
      "loss": 0.4963,
      "step": 2765
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.588554210558713e-05,
      "loss": 0.491,
      "step": 2766
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.581701715332383e-05,
      "loss": 0.5058,
      "step": 2767
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.574851037113447e-05,
      "loss": 0.5029,
      "step": 2768
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.568002179543409e-05,
      "loss": 0.5001,
      "step": 2769
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.561155146262798e-05,
      "loss": 0.4961,
      "step": 2770
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.554309940911178e-05,
      "loss": 0.5103,
      "step": 2771
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.547466567127135e-05,
      "loss": 0.4982,
      "step": 2772
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.540625028548295e-05,
      "loss": 0.5018,
      "step": 2773
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.533785328811295e-05,
      "loss": 0.4932,
      "step": 2774
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.526947471551798e-05,
      "loss": 0.5056,
      "step": 2775
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.520111460404492e-05,
      "loss": 0.501,
      "step": 2776
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.513277299003077e-05,
      "loss": 0.5052,
      "step": 2777
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.506444990980279e-05,
      "loss": 0.5044,
      "step": 2778
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.499614539967827e-05,
      "loss": 0.4913,
      "step": 2779
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.492785949596471e-05,
      "loss": 0.4946,
      "step": 2780
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.485959223495972e-05,
      "loss": 0.51,
      "step": 2781
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.479134365295094e-05,
      "loss": 0.511,
      "step": 2782
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.472311378621619e-05,
      "loss": 0.4909,
      "step": 2783
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.465490267102324e-05,
      "loss": 0.5079,
      "step": 2784
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.458671034362991e-05,
      "loss": 0.5,
      "step": 2785
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.451853684028407e-05,
      "loss": 0.4985,
      "step": 2786
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.445038219722354e-05,
      "loss": 0.5044,
      "step": 2787
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.438224645067617e-05,
      "loss": 0.5032,
      "step": 2788
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.431412963685973e-05,
      "loss": 0.5038,
      "step": 2789
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.424603179198186e-05,
      "loss": 0.5084,
      "step": 2790
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.417795295224028e-05,
      "loss": 0.5065,
      "step": 2791
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.41098931538225e-05,
      "loss": 0.5017,
      "step": 2792
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.40418524329059e-05,
      "loss": 0.5002,
      "step": 2793
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.397383082565775e-05,
      "loss": 0.5016,
      "step": 2794
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.390582836823514e-05,
      "loss": 0.5147,
      "step": 2795
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.383784509678502e-05,
      "loss": 0.5034,
      "step": 2796
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.37698810474441e-05,
      "loss": 0.4867,
      "step": 2797
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.370193625633887e-05,
      "loss": 0.4967,
      "step": 2798
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.363401075958561e-05,
      "loss": 0.4956,
      "step": 2799
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.356610459329038e-05,
      "loss": 0.5065,
      "step": 2800
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.349821779354887e-05,
      "loss": 0.4898,
      "step": 2801
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.343035039644655e-05,
      "loss": 0.4855,
      "step": 2802
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.336250243805852e-05,
      "loss": 0.4962,
      "step": 2803
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.329467395444963e-05,
      "loss": 0.4948,
      "step": 2804
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.32268649816743e-05,
      "loss": 0.4824,
      "step": 2805
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.31590755557766e-05,
      "loss": 0.4986,
      "step": 2806
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.309130571279024e-05,
      "loss": 0.508,
      "step": 2807
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.302355548873843e-05,
      "loss": 0.5034,
      "step": 2808
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.29558249196341e-05,
      "loss": 0.4874,
      "step": 2809
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.288811404147963e-05,
      "loss": 0.4932,
      "step": 2810
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.282042289026695e-05,
      "loss": 0.5014,
      "step": 2811
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.275275150197751e-05,
      "loss": 0.4924,
      "step": 2812
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.268509991258222e-05,
      "loss": 0.5017,
      "step": 2813
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.261746815804152e-05,
      "loss": 0.5093,
      "step": 2814
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.254985627430531e-05,
      "loss": 0.5027,
      "step": 2815
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.248226429731289e-05,
      "loss": 0.5006,
      "step": 2816
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.241469226299293e-05,
      "loss": 0.5071,
      "step": 2817
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.234714020726365e-05,
      "loss": 0.4977,
      "step": 2818
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.227960816603254e-05,
      "loss": 0.514,
      "step": 2819
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.221209617519645e-05,
      "loss": 0.5204,
      "step": 2820
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.214460427064161e-05,
      "loss": 0.5054,
      "step": 2821
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.207713248824356e-05,
      "loss": 0.5026,
      "step": 2822
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.200968086386714e-05,
      "loss": 0.4982,
      "step": 2823
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.194224943336646e-05,
      "loss": 0.4939,
      "step": 2824
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.18748382325849e-05,
      "loss": 0.5025,
      "step": 2825
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.180744729735512e-05,
      "loss": 0.5107,
      "step": 2826
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.174007666349904e-05,
      "loss": 0.5061,
      "step": 2827
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.167272636682766e-05,
      "loss": 0.4988,
      "step": 2828
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.160539644314124e-05,
      "loss": 0.5061,
      "step": 2829
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.153808692822931e-05,
      "loss": 0.5216,
      "step": 2830
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.147079785787038e-05,
      "loss": 0.5082,
      "step": 2831
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.140352926783218e-05,
      "loss": 0.5045,
      "step": 2832
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.133628119387156e-05,
      "loss": 0.4966,
      "step": 2833
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.126905367173445e-05,
      "loss": 0.4967,
      "step": 2834
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.120184673715583e-05,
      "loss": 0.5044,
      "step": 2835
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.113466042585975e-05,
      "loss": 0.5017,
      "step": 2836
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.106749477355938e-05,
      "loss": 0.4934,
      "step": 2837
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.100034981595679e-05,
      "loss": 0.5075,
      "step": 2838
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.0933225588743094e-05,
      "loss": 0.4932,
      "step": 2839
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.08661221275984e-05,
      "loss": 0.4878,
      "step": 2840
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.079903946819177e-05,
      "loss": 0.486,
      "step": 2841
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.073197764618119e-05,
      "loss": 0.5071,
      "step": 2842
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.0664936697213616e-05,
      "loss": 0.5084,
      "step": 2843
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.059791665692485e-05,
      "loss": 0.494,
      "step": 2844
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.053091756093957e-05,
      "loss": 0.4942,
      "step": 2845
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.046393944487145e-05,
      "loss": 0.501,
      "step": 2846
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.039698234432289e-05,
      "loss": 0.4967,
      "step": 2847
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.033004629488515e-05,
      "loss": 0.5044,
      "step": 2848
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.026313133213829e-05,
      "loss": 0.4901,
      "step": 2849
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.019623749165117e-05,
      "loss": 0.4951,
      "step": 2850
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.012936480898142e-05,
      "loss": 0.5005,
      "step": 2851
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.006251331967545e-05,
      "loss": 0.5118,
      "step": 2852
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.999568305926837e-05,
      "loss": 0.4942,
      "step": 2853
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9928874063284004e-05,
      "loss": 0.4975,
      "step": 2854
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9862086367234935e-05,
      "loss": 0.4933,
      "step": 2855
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9795320006622315e-05,
      "loss": 0.4937,
      "step": 2856
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.972857501693607e-05,
      "loss": 0.4984,
      "step": 2857
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9661851433654704e-05,
      "loss": 0.4913,
      "step": 2858
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.959514929224534e-05,
      "loss": 0.503,
      "step": 2859
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.95284686281637e-05,
      "loss": 0.5043,
      "step": 2860
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.946180947685414e-05,
      "loss": 0.5123,
      "step": 2861
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.93951718737495e-05,
      "loss": 0.5081,
      "step": 2862
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.932855585427119e-05,
      "loss": 0.4897,
      "step": 2863
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.926196145382927e-05,
      "loss": 0.5063,
      "step": 2864
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9195388707822154e-05,
      "loss": 0.4968,
      "step": 2865
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.9128837651636795e-05,
      "loss": 0.5019,
      "step": 2866
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.9062308320648605e-05,
      "loss": 0.4976,
      "step": 2867
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.899580075022149e-05,
      "loss": 0.4955,
      "step": 2868
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.892931497570774e-05,
      "loss": 0.5066,
      "step": 2869
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.88628510324481e-05,
      "loss": 0.4982,
      "step": 2870
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.8796408955771676e-05,
      "loss": 0.5234,
      "step": 2871
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.872998878099592e-05,
      "loss": 0.4977,
      "step": 2872
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.8663590543426806e-05,
      "loss": 0.4957,
      "step": 2873
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.8597214278358446e-05,
      "loss": 0.5049,
      "step": 2874
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.853086002107338e-05,
      "loss": 0.4898,
      "step": 2875
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.846452780684243e-05,
      "loss": 0.5128,
      "step": 2876
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.839821767092466e-05,
      "loss": 0.5063,
      "step": 2877
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.8331929648567485e-05,
      "loss": 0.4934,
      "step": 2878
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.8265663775006496e-05,
      "loss": 0.5016,
      "step": 2879
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.819942008546553e-05,
      "loss": 0.5064,
      "step": 2880
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.813319861515658e-05,
      "loss": 0.5033,
      "step": 2881
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.8066999399279976e-05,
      "loss": 0.5025,
      "step": 2882
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.8000822473024074e-05,
      "loss": 0.4798,
      "step": 2883
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.793466787156544e-05,
      "loss": 0.4809,
      "step": 2884
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.786853563006876e-05,
      "loss": 0.4979,
      "step": 2885
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.780242578368682e-05,
      "loss": 0.5019,
      "step": 2886
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.773633836756056e-05,
      "loss": 0.4963,
      "step": 2887
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.767027341681892e-05,
      "loss": 0.504,
      "step": 2888
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.7604230966578965e-05,
      "loss": 0.5054,
      "step": 2889
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.75382110519457e-05,
      "loss": 0.5032,
      "step": 2890
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.7472213708012334e-05,
      "loss": 0.4918,
      "step": 2891
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.7406238969859904e-05,
      "loss": 0.4848,
      "step": 2892
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.734028687255751e-05,
      "loss": 0.4857,
      "step": 2893
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.727435745116218e-05,
      "loss": 0.4945,
      "step": 2894
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.720845074071893e-05,
      "loss": 0.4896,
      "step": 2895
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.714256677626067e-05,
      "loss": 0.4985,
      "step": 2896
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.707670559280824e-05,
      "loss": 0.5084,
      "step": 2897
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.7010867225370344e-05,
      "loss": 0.497,
      "step": 2898
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.6945051708943575e-05,
      "loss": 0.4925,
      "step": 2899
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.6879259078512416e-05,
      "loss": 0.5075,
      "step": 2900
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.681348936904914e-05,
      "loss": 0.5068,
      "step": 2901
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.6747742615513846e-05,
      "loss": 0.4997,
      "step": 2902
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.668201885285442e-05,
      "loss": 0.5114,
      "step": 2903
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.661631811600654e-05,
      "loss": 0.5,
      "step": 2904
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.655064043989364e-05,
      "loss": 0.5131,
      "step": 2905
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.648498585942692e-05,
      "loss": 0.4961,
      "step": 2906
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.641935440950525e-05,
      "loss": 0.4992,
      "step": 2907
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.635374612501523e-05,
      "loss": 0.5057,
      "step": 2908
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.62881610408312e-05,
      "loss": 0.4931,
      "step": 2909
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.622259919181512e-05,
      "loss": 0.5116,
      "step": 2910
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.615706061281656e-05,
      "loss": 0.5012,
      "step": 2911
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.609154533867279e-05,
      "loss": 0.496,
      "step": 2912
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.6026053404208656e-05,
      "loss": 0.5092,
      "step": 2913
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.596058484423656e-05,
      "loss": 0.5032,
      "step": 2914
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.589513969355662e-05,
      "loss": 0.4934,
      "step": 2915
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.582971798695635e-05,
      "loss": 0.4985,
      "step": 2916
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.57643197592109e-05,
      "loss": 0.5021,
      "step": 2917
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.569894504508289e-05,
      "loss": 0.5095,
      "step": 2918
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.56335938793224e-05,
      "loss": 0.4928,
      "step": 2919
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.556826629666716e-05,
      "loss": 0.5009,
      "step": 2920
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.55029623318422e-05,
      "loss": 0.4943,
      "step": 2921
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.5437682019560054e-05,
      "loss": 0.485,
      "step": 2922
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.53724253945207e-05,
      "loss": 0.4979,
      "step": 2923
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.530719249141147e-05,
      "loss": 0.5025,
      "step": 2924
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.524198334490716e-05,
      "loss": 0.4988,
      "step": 2925
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.517679798966987e-05,
      "loss": 0.5033,
      "step": 2926
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.511163646034907e-05,
      "loss": 0.5149,
      "step": 2927
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.5046498791581634e-05,
      "loss": 0.504,
      "step": 2928
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.498138501799167e-05,
      "loss": 0.4991,
      "step": 2929
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.491629517419061e-05,
      "loss": 0.4949,
      "step": 2930
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.48512292947772e-05,
      "loss": 0.4959,
      "step": 2931
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.4786187414337366e-05,
      "loss": 0.5141,
      "step": 2932
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.4721169567444355e-05,
      "loss": 0.4963,
      "step": 2933
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.4656175788658606e-05,
      "loss": 0.5035,
      "step": 2934
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.459120611252777e-05,
      "loss": 0.5041,
      "step": 2935
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.452626057358665e-05,
      "loss": 0.493,
      "step": 2936
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.446133920635733e-05,
      "loss": 0.5014,
      "step": 2937
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.4396442045348926e-05,
      "loss": 0.497,
      "step": 2938
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.4331569125057756e-05,
      "loss": 0.4975,
      "step": 2939
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.4266720479967216e-05,
      "loss": 0.4971,
      "step": 2940
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.420189614454778e-05,
      "loss": 0.4919,
      "step": 2941
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.413709615325708e-05,
      "loss": 0.4921,
      "step": 2942
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.4072320540539736e-05,
      "loss": 0.4865,
      "step": 2943
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.4007569340827426e-05,
      "loss": 0.508,
      "step": 2944
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.3942842588538823e-05,
      "loss": 0.4984,
      "step": 2945
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.3878140318079714e-05,
      "loss": 0.5028,
      "step": 2946
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.381346256384279e-05,
      "loss": 0.5067,
      "step": 2947
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.3748809360207667e-05,
      "loss": 0.4928,
      "step": 2948
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.3684180741541e-05,
      "loss": 0.4968,
      "step": 2949
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.361957674219632e-05,
      "loss": 0.5019,
      "step": 2950
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.35549973965141e-05,
      "loss": 0.5059,
      "step": 2951
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.349044273882169e-05,
      "loss": 0.5049,
      "step": 2952
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.342591280343334e-05,
      "loss": 0.5042,
      "step": 2953
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.336140762465007e-05,
      "loss": 0.492,
      "step": 2954
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.329692723675994e-05,
      "loss": 0.4993,
      "step": 2955
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.3232471674037645e-05,
      "loss": 0.4845,
      "step": 2956
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.3168040970744724e-05,
      "loss": 0.5018,
      "step": 2957
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.310363516112955e-05,
      "loss": 0.503,
      "step": 2958
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.3039254279427245e-05,
      "loss": 0.5008,
      "step": 2959
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.2974898359859645e-05,
      "loss": 0.4907,
      "step": 2960
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.291056743663536e-05,
      "loss": 0.5101,
      "step": 2961
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.2846261543949674e-05,
      "loss": 0.4523,
      "step": 2962
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.2781980715984634e-05,
      "loss": 0.4419,
      "step": 2963
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.27177249869089e-05,
      "loss": 0.4444,
      "step": 2964
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.265349439087781e-05,
      "loss": 0.4401,
      "step": 2965
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.258928896203336e-05,
      "loss": 0.4492,
      "step": 2966
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.252510873450407e-05,
      "loss": 0.4341,
      "step": 2967
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.246095374240527e-05,
      "loss": 0.4353,
      "step": 2968
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.2396824019838676e-05,
      "loss": 0.4333,
      "step": 2969
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.233271960089267e-05,
      "loss": 0.4423,
      "step": 2970
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.226864051964214e-05,
      "loss": 0.4424,
      "step": 2971
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.220458681014852e-05,
      "loss": 0.4292,
      "step": 2972
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.21405585064598e-05,
      "loss": 0.4314,
      "step": 2973
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.207655564261043e-05,
      "loss": 0.4327,
      "step": 2974
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.201257825262131e-05,
      "loss": 0.4305,
      "step": 2975
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.194862637049985e-05,
      "loss": 0.4284,
      "step": 2976
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1884700030239864e-05,
      "loss": 0.4413,
      "step": 2977
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.182079926582159e-05,
      "loss": 0.4366,
      "step": 2978
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.175692411121171e-05,
      "loss": 0.4292,
      "step": 2979
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1693074600363255e-05,
      "loss": 0.432,
      "step": 2980
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.162925076721561e-05,
      "loss": 0.4315,
      "step": 2981
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.156545264569459e-05,
      "loss": 0.4345,
      "step": 2982
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.150168026971226e-05,
      "loss": 0.4237,
      "step": 2983
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1437933673167035e-05,
      "loss": 0.4359,
      "step": 2984
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.137421288994363e-05,
      "loss": 0.426,
      "step": 2985
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.131051795391302e-05,
      "loss": 0.4282,
      "step": 2986
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.124684889893246e-05,
      "loss": 0.4327,
      "step": 2987
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.118320575884544e-05,
      "loss": 0.4378,
      "step": 2988
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.111958856748167e-05,
      "loss": 0.4323,
      "step": 2989
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.1055997358657025e-05,
      "loss": 0.4257,
      "step": 2990
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0992432166173706e-05,
      "loss": 0.4349,
      "step": 2991
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.092889302381997e-05,
      "loss": 0.4433,
      "step": 2992
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.086537996537022e-05,
      "loss": 0.4421,
      "step": 2993
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0801893024585036e-05,
      "loss": 0.439,
      "step": 2994
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.073843223521111e-05,
      "loss": 0.4258,
      "step": 2995
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0674997630981226e-05,
      "loss": 0.4166,
      "step": 2996
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0611589245614264e-05,
      "loss": 0.4246,
      "step": 2997
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.0548207112815116e-05,
      "loss": 0.4286,
      "step": 2998
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.04848512662748e-05,
      "loss": 0.4371,
      "step": 2999
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.042152173967026e-05,
      "loss": 0.4332,
      "step": 3000
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.035821856666459e-05,
      "loss": 0.4303,
      "step": 3001
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.0294941780906756e-05,
      "loss": 0.426,
      "step": 3002
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.023169141603172e-05,
      "loss": 0.4308,
      "step": 3003
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.016846750566046e-05,
      "loss": 0.4283,
      "step": 3004
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.010527008339979e-05,
      "loss": 0.4287,
      "step": 3005
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.004209918284255e-05,
      "loss": 0.4225,
      "step": 3006
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.9978954837567406e-05,
      "loss": 0.4375,
      "step": 3007
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.991583708113894e-05,
      "loss": 0.4269,
      "step": 3008
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.9852745947107574e-05,
      "loss": 0.4251,
      "step": 3009
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.978968146900966e-05,
      "loss": 0.4213,
      "step": 3010
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.97266436803673e-05,
      "loss": 0.425,
      "step": 3011
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.9663632614688426e-05,
      "loss": 0.4424,
      "step": 3012
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.960064830546678e-05,
      "loss": 0.4235,
      "step": 3013
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.953769078618182e-05,
      "loss": 0.4265,
      "step": 3014
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.947476009029891e-05,
      "loss": 0.4492,
      "step": 3015
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.9411856251269003e-05,
      "loss": 0.4362,
      "step": 3016
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.934897930252886e-05,
      "loss": 0.4292,
      "step": 3017
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.928612927750092e-05,
      "loss": 0.4294,
      "step": 3018
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.9223306209593244e-05,
      "loss": 0.4216,
      "step": 3019
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.916051013219976e-05,
      "loss": 0.4361,
      "step": 3020
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.909774107869985e-05,
      "loss": 0.4416,
      "step": 3021
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.903499908245862e-05,
      "loss": 0.4317,
      "step": 3022
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.897228417682679e-05,
      "loss": 0.4155,
      "step": 3023
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.8909596395140645e-05,
      "loss": 0.441,
      "step": 3024
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.884693577072208e-05,
      "loss": 0.4276,
      "step": 3025
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.8784302336878575e-05,
      "loss": 0.435,
      "step": 3026
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.872169612690306e-05,
      "loss": 0.4281,
      "step": 3027
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.865911717407418e-05,
      "loss": 0.4284,
      "step": 3028
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.85965655116559e-05,
      "loss": 0.4468,
      "step": 3029
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.853404117289778e-05,
      "loss": 0.4306,
      "step": 3030
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.8471544191034836e-05,
      "loss": 0.4203,
      "step": 3031
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.840907459928752e-05,
      "loss": 0.427,
      "step": 3032
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.8346632430861764e-05,
      "loss": 0.4241,
      "step": 3033
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.8284217718948874e-05,
      "loss": 0.4375,
      "step": 3034
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.822183049672563e-05,
      "loss": 0.4417,
      "step": 3035
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.815947079735409e-05,
      "loss": 0.438,
      "step": 3036
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.8097138653981834e-05,
      "loss": 0.4308,
      "step": 3037
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.803483409974169e-05,
      "loss": 0.4383,
      "step": 3038
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.797255716775182e-05,
      "loss": 0.4392,
      "step": 3039
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.791030789111575e-05,
      "loss": 0.4268,
      "step": 3040
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.784808630292226e-05,
      "loss": 0.4364,
      "step": 3041
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.778589243624544e-05,
      "loss": 0.4361,
      "step": 3042
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.772372632414466e-05,
      "loss": 0.4328,
      "step": 3043
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.766158799966448e-05,
      "loss": 0.4188,
      "step": 3044
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.759947749583472e-05,
      "loss": 0.4207,
      "step": 3045
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.753739484567048e-05,
      "loss": 0.4279,
      "step": 3046
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.7475340082171946e-05,
      "loss": 0.42,
      "step": 3047
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.7413313238324556e-05,
      "loss": 0.4298,
      "step": 3048
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.735131434709883e-05,
      "loss": 0.4262,
      "step": 3049
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.7289343441450505e-05,
      "loss": 0.4368,
      "step": 3050
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.7227400554320425e-05,
      "loss": 0.4253,
      "step": 3051
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.71654857186345e-05,
      "loss": 0.4314,
      "step": 3052
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.710359896730379e-05,
      "loss": 0.4318,
      "step": 3053
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.704174033322434e-05,
      "loss": 0.4294,
      "step": 3054
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.697990984927738e-05,
      "loss": 0.4272,
      "step": 3055
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.691810754832907e-05,
      "loss": 0.4287,
      "step": 3056
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.685633346323062e-05,
      "loss": 0.4344,
      "step": 3057
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.679458762681823e-05,
      "loss": 0.435,
      "step": 3058
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.673287007191312e-05,
      "loss": 0.4227,
      "step": 3059
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.667118083132145e-05,
      "loss": 0.433,
      "step": 3060
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.660951993783431e-05,
      "loss": 0.4338,
      "step": 3061
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.654788742422773e-05,
      "loss": 0.4473,
      "step": 3062
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.6486283323262734e-05,
      "loss": 0.4283,
      "step": 3063
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.6424707667685154e-05,
      "loss": 0.4098,
      "step": 3064
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.636316049022571e-05,
      "loss": 0.4345,
      "step": 3065
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.6301641823600014e-05,
      "loss": 0.422,
      "step": 3066
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.624015170050847e-05,
      "loss": 0.4292,
      "step": 3067
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.6178690153636427e-05,
      "loss": 0.4395,
      "step": 3068
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.611725721565392e-05,
      "loss": 0.4321,
      "step": 3069
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.605585291921584e-05,
      "loss": 0.4305,
      "step": 3070
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.5994477296961815e-05,
      "loss": 0.4279,
      "step": 3071
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.5933130381516234e-05,
      "loss": 0.4384,
      "step": 3072
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.5871812205488305e-05,
      "loss": 0.4358,
      "step": 3073
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.581052280147189e-05,
      "loss": 0.4394,
      "step": 3074
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.574926220204554e-05,
      "loss": 0.426,
      "step": 3075
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.568803043977254e-05,
      "loss": 0.4271,
      "step": 3076
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.562682754720082e-05,
      "loss": 0.4401,
      "step": 3077
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.556565355686299e-05,
      "loss": 0.4295,
      "step": 3078
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.550450850127625e-05,
      "loss": 0.4378,
      "step": 3079
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.544339241294248e-05,
      "loss": 0.4188,
      "step": 3080
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.538230532434811e-05,
      "loss": 0.4333,
      "step": 3081
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.532124726796415e-05,
      "loss": 0.4233,
      "step": 3082
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.526021827624629e-05,
      "loss": 0.4199,
      "step": 3083
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.519921838163463e-05,
      "loss": 0.4233,
      "step": 3084
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.5138247616553876e-05,
      "loss": 0.4206,
      "step": 3085
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.507730601341321e-05,
      "loss": 0.4344,
      "step": 3086
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.5016393604606364e-05,
      "loss": 0.4226,
      "step": 3087
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4955510422511514e-05,
      "loss": 0.4319,
      "step": 3088
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4894656499491306e-05,
      "loss": 0.4359,
      "step": 3089
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.483383186789283e-05,
      "loss": 0.4184,
      "step": 3090
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.47730365600476e-05,
      "loss": 0.4248,
      "step": 3091
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.47122706082716e-05,
      "loss": 0.4322,
      "step": 3092
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.465153404486515e-05,
      "loss": 0.4386,
      "step": 3093
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.459082690211297e-05,
      "loss": 0.4343,
      "step": 3094
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4530149212284125e-05,
      "loss": 0.4282,
      "step": 3095
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.446950100763203e-05,
      "loss": 0.4326,
      "step": 3096
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.440888232039444e-05,
      "loss": 0.4257,
      "step": 3097
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.434829318279341e-05,
      "loss": 0.4242,
      "step": 3098
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.42877336270353e-05,
      "loss": 0.4342,
      "step": 3099
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.422720368531068e-05,
      "loss": 0.4281,
      "step": 3100
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.416670338979453e-05,
      "loss": 0.4331,
      "step": 3101
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.410623277264592e-05,
      "loss": 0.4364,
      "step": 3102
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.40457918660082e-05,
      "loss": 0.4315,
      "step": 3103
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.398538070200894e-05,
      "loss": 0.4279,
      "step": 3104
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.392499931275988e-05,
      "loss": 0.4348,
      "step": 3105
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.386464773035695e-05,
      "loss": 0.4318,
      "step": 3106
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.3804325986880227e-05,
      "loss": 0.4276,
      "step": 3107
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.3744034114393907e-05,
      "loss": 0.4274,
      "step": 3108
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.36837721449463e-05,
      "loss": 0.4427,
      "step": 3109
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.3623540110569935e-05,
      "loss": 0.4291,
      "step": 3110
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.356333804328129e-05,
      "loss": 0.4445,
      "step": 3111
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.350316597508098e-05,
      "loss": 0.4289,
      "step": 3112
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.3443023937953654e-05,
      "loss": 0.4259,
      "step": 3113
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.338291196386797e-05,
      "loss": 0.4347,
      "step": 3114
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.33228300847767e-05,
      "loss": 0.4257,
      "step": 3115
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.326277833261654e-05,
      "loss": 0.4288,
      "step": 3116
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.3202756739308183e-05,
      "loss": 0.4223,
      "step": 3117
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.3142765336756306e-05,
      "loss": 0.434,
      "step": 3118
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.3082804156849486e-05,
      "loss": 0.4379,
      "step": 3119
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.3022873231460356e-05,
      "loss": 0.4333,
      "step": 3120
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.296297259244536e-05,
      "loss": 0.4247,
      "step": 3121
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.290310227164487e-05,
      "loss": 0.435,
      "step": 3122
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.284326230088314e-05,
      "loss": 0.4301,
      "step": 3123
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.278345271196831e-05,
      "loss": 0.4325,
      "step": 3124
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2723673536692346e-05,
      "loss": 0.432,
      "step": 3125
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2663924806831065e-05,
      "loss": 0.4324,
      "step": 3126
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2604206554144076e-05,
      "loss": 0.4263,
      "step": 3127
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2544518810374854e-05,
      "loss": 0.4374,
      "step": 3128
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2484861607250595e-05,
      "loss": 0.4265,
      "step": 3129
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2425234976482256e-05,
      "loss": 0.4398,
      "step": 3130
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2365638949764576e-05,
      "loss": 0.4303,
      "step": 3131
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.230607355877602e-05,
      "loss": 0.4273,
      "step": 3132
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.224653883517874e-05,
      "loss": 0.432,
      "step": 3133
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.218703481061863e-05,
      "loss": 0.4246,
      "step": 3134
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.2127561516725246e-05,
      "loss": 0.4378,
      "step": 3135
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.2068118985111746e-05,
      "loss": 0.427,
      "step": 3136
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.20087072473751e-05,
      "loss": 0.4257,
      "step": 3137
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.194932633509576e-05,
      "loss": 0.4379,
      "step": 3138
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.188997627983784e-05,
      "loss": 0.4409,
      "step": 3139
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.183065711314904e-05,
      "loss": 0.4281,
      "step": 3140
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.1771368866560665e-05,
      "loss": 0.439,
      "step": 3141
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.171211157158758e-05,
      "loss": 0.4472,
      "step": 3142
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.165288525972818e-05,
      "loss": 0.4311,
      "step": 3143
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.15936899624644e-05,
      "loss": 0.4161,
      "step": 3144
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.153452571126165e-05,
      "loss": 0.4328,
      "step": 3145
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.1475392537568956e-05,
      "loss": 0.4215,
      "step": 3146
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.141629047281871e-05,
      "loss": 0.4307,
      "step": 3147
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.135721954842679e-05,
      "loss": 0.4305,
      "step": 3148
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.129817979579257e-05,
      "loss": 0.4283,
      "step": 3149
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.123917124629878e-05,
      "loss": 0.4183,
      "step": 3150
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.118019393131162e-05,
      "loss": 0.4143,
      "step": 3151
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.112124788218067e-05,
      "loss": 0.4243,
      "step": 3152
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.10623331302389e-05,
      "loss": 0.4183,
      "step": 3153
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.1003449706802586e-05,
      "loss": 0.4249,
      "step": 3154
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.094459764317147e-05,
      "loss": 0.4315,
      "step": 3155
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.0885776970628534e-05,
      "loss": 0.4229,
      "step": 3156
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.082698772044008e-05,
      "loss": 0.433,
      "step": 3157
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.076822992385574e-05,
      "loss": 0.4366,
      "step": 3158
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.070950361210838e-05,
      "loss": 0.4306,
      "step": 3159
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.0650808816414186e-05,
      "loss": 0.4358,
      "step": 3160
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.059214556797255e-05,
      "loss": 0.4304,
      "step": 3161
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.053351389796608e-05,
      "loss": 0.4374,
      "step": 3162
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.04749138375607e-05,
      "loss": 0.4313,
      "step": 3163
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.041634541790542e-05,
      "loss": 0.4373,
      "step": 3164
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.035780867013248e-05,
      "loss": 0.4266,
      "step": 3165
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.0299303625357266e-05,
      "loss": 0.4458,
      "step": 3166
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.024083031467828e-05,
      "loss": 0.4378,
      "step": 3167
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.018238876917728e-05,
      "loss": 0.4423,
      "step": 3168
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.012397901991902e-05,
      "loss": 0.4293,
      "step": 3169
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.0065601097951386e-05,
      "loss": 0.4352,
      "step": 3170
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.000725503430536e-05,
      "loss": 0.4348,
      "step": 3171
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.9948940859994966e-05,
      "loss": 0.4309,
      "step": 3172
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.989065860601725e-05,
      "loss": 0.4342,
      "step": 3173
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.983240830335242e-05,
      "loss": 0.4334,
      "step": 3174
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.977418998296356e-05,
      "loss": 0.4299,
      "step": 3175
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.9716003675796806e-05,
      "loss": 0.4241,
      "step": 3176
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.965784941278128e-05,
      "loss": 0.4289,
      "step": 3177
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.9599727224829073e-05,
      "loss": 0.4361,
      "step": 3178
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.954163714283521e-05,
      "loss": 0.4251,
      "step": 3179
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.948357919767767e-05,
      "loss": 0.4312,
      "step": 3180
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.942555342021732e-05,
      "loss": 0.4397,
      "step": 3181
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.936755984129793e-05,
      "loss": 0.4272,
      "step": 3182
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.9309598491746246e-05,
      "loss": 0.4247,
      "step": 3183
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.925166940237176e-05,
      "loss": 0.4302,
      "step": 3184
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.919377260396687e-05,
      "loss": 0.4278,
      "step": 3185
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.913590812730681e-05,
      "loss": 0.4295,
      "step": 3186
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.907807600314962e-05,
      "loss": 0.4182,
      "step": 3187
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.9020276262236167e-05,
      "loss": 0.4265,
      "step": 3188
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.8962508935290054e-05,
      "loss": 0.438,
      "step": 3189
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.8904774053017726e-05,
      "loss": 0.4288,
      "step": 3190
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.884707164610829e-05,
      "loss": 0.4353,
      "step": 3191
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.878940174523371e-05,
      "loss": 0.4321,
      "step": 3192
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8731764381048595e-05,
      "loss": 0.4217,
      "step": 3193
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.867415958419024e-05,
      "loss": 0.4394,
      "step": 3194
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.861658738527869e-05,
      "loss": 0.4183,
      "step": 3195
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8559047814916605e-05,
      "loss": 0.4361,
      "step": 3196
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.850154090368935e-05,
      "loss": 0.437,
      "step": 3197
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.844406668216488e-05,
      "loss": 0.4271,
      "step": 3198
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.838662518089383e-05,
      "loss": 0.4376,
      "step": 3199
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8329216430409366e-05,
      "loss": 0.4318,
      "step": 3200
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.827184046122737e-05,
      "loss": 0.4348,
      "step": 3201
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.821449730384618e-05,
      "loss": 0.4261,
      "step": 3202
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.815718698874672e-05,
      "loss": 0.4262,
      "step": 3203
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8099909546392496e-05,
      "loss": 0.4299,
      "step": 3204
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.804266500722952e-05,
      "loss": 0.4351,
      "step": 3205
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.798545340168629e-05,
      "loss": 0.4362,
      "step": 3206
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.792827476017382e-05,
      "loss": 0.4268,
      "step": 3207
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.787112911308561e-05,
      "loss": 0.4384,
      "step": 3208
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.781401649079758e-05,
      "loss": 0.4296,
      "step": 3209
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7756936923668175e-05,
      "loss": 0.4315,
      "step": 3210
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.769989044203821e-05,
      "loss": 0.4318,
      "step": 3211
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.764287707623091e-05,
      "loss": 0.4259,
      "step": 3212
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.758589685655193e-05,
      "loss": 0.4329,
      "step": 3213
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.752894981328927e-05,
      "loss": 0.4314,
      "step": 3214
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7472035976713284e-05,
      "loss": 0.4308,
      "step": 3215
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.741515537707677e-05,
      "loss": 0.4319,
      "step": 3216
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.735830804461476e-05,
      "loss": 0.4253,
      "step": 3217
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7301494009544644e-05,
      "loss": 0.4351,
      "step": 3218
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7244713302066106e-05,
      "loss": 0.4213,
      "step": 3219
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.718796595236106e-05,
      "loss": 0.4217,
      "step": 3220
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.713125199059383e-05,
      "loss": 0.4287,
      "step": 3221
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.707457144691085e-05,
      "loss": 0.4247,
      "step": 3222
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.701792435144086e-05,
      "loss": 0.4329,
      "step": 3223
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6961310734294795e-05,
      "loss": 0.4272,
      "step": 3224
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.690473062556582e-05,
      "loss": 0.4251,
      "step": 3225
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.684818405532925e-05,
      "loss": 0.4279,
      "step": 3226
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.679167105364257e-05,
      "loss": 0.426,
      "step": 3227
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.673519165054553e-05,
      "loss": 0.4151,
      "step": 3228
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6678745876059894e-05,
      "loss": 0.4263,
      "step": 3229
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.662233376018959e-05,
      "loss": 0.4204,
      "step": 3230
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6565955332920654e-05,
      "loss": 0.424,
      "step": 3231
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.650961062422122e-05,
      "loss": 0.4288,
      "step": 3232
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.645329966404152e-05,
      "loss": 0.4375,
      "step": 3233
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6397022482313805e-05,
      "loss": 0.4267,
      "step": 3234
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6340779108952397e-05,
      "loss": 0.43,
      "step": 3235
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.62845695738536e-05,
      "loss": 0.4258,
      "step": 3236
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.622839390689585e-05,
      "loss": 0.4306,
      "step": 3237
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6172252137939466e-05,
      "loss": 0.4322,
      "step": 3238
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6116144296826784e-05,
      "loss": 0.4318,
      "step": 3239
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6060070413382106e-05,
      "loss": 0.44,
      "step": 3240
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.600403051741168e-05,
      "loss": 0.4372,
      "step": 3241
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5948024638703704e-05,
      "loss": 0.428,
      "step": 3242
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5892052807028276e-05,
      "loss": 0.4235,
      "step": 3243
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.583611505213739e-05,
      "loss": 0.4295,
      "step": 3244
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.578021140376496e-05,
      "loss": 0.4331,
      "step": 3245
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.57243418916267e-05,
      "loss": 0.4357,
      "step": 3246
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.56685065454203e-05,
      "loss": 0.4321,
      "step": 3247
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.56127053948252e-05,
      "loss": 0.436,
      "step": 3248
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.555693846950264e-05,
      "loss": 0.4272,
      "step": 3249
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5501205799095746e-05,
      "loss": 0.4358,
      "step": 3250
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5445507413229384e-05,
      "loss": 0.4389,
      "step": 3251
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.538984334151021e-05,
      "loss": 0.444,
      "step": 3252
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.533421361352663e-05,
      "loss": 0.431,
      "step": 3253
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.527861825884882e-05,
      "loss": 0.4288,
      "step": 3254
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.522305730702862e-05,
      "loss": 0.4261,
      "step": 3255
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.516753078759971e-05,
      "loss": 0.4201,
      "step": 3256
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.511203873007736e-05,
      "loss": 0.4222,
      "step": 3257
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5056581163958535e-05,
      "loss": 0.4292,
      "step": 3258
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.500115811872189e-05,
      "loss": 0.437,
      "step": 3259
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.494576962382772e-05,
      "loss": 0.4324,
      "step": 3260
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.489041570871796e-05,
      "loss": 0.4339,
      "step": 3261
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.4835096402816124e-05,
      "loss": 0.4293,
      "step": 3262
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.477981173552745e-05,
      "loss": 0.4344,
      "step": 3263
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.4724561736238615e-05,
      "loss": 0.4278,
      "step": 3264
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.466934643431795e-05,
      "loss": 0.4353,
      "step": 3265
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.461416585911532e-05,
      "loss": 0.4262,
      "step": 3266
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.455902003996212e-05,
      "loss": 0.435,
      "step": 3267
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4503909006171345e-05,
      "loss": 0.4239,
      "step": 3268
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.444883278703739e-05,
      "loss": 0.4165,
      "step": 3269
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.439379141183622e-05,
      "loss": 0.4289,
      "step": 3270
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.433878490982524e-05,
      "loss": 0.4373,
      "step": 3271
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.428381331024333e-05,
      "loss": 0.4304,
      "step": 3272
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4228876642310804e-05,
      "loss": 0.4287,
      "step": 3273
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.417397493522947e-05,
      "loss": 0.4327,
      "step": 3274
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.411910821818249e-05,
      "loss": 0.4299,
      "step": 3275
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4064276520334426e-05,
      "loss": 0.4335,
      "step": 3276
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.400947987083126e-05,
      "loss": 0.4286,
      "step": 3277
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.3954718298800316e-05,
      "loss": 0.4178,
      "step": 3278
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.389999183335029e-05,
      "loss": 0.4201,
      "step": 3279
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.38453005035712e-05,
      "loss": 0.4302,
      "step": 3280
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.379064433853439e-05,
      "loss": 0.4249,
      "step": 3281
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3736023367292513e-05,
      "loss": 0.4309,
      "step": 3282
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3681437618879575e-05,
      "loss": 0.4311,
      "step": 3283
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.362688712231076e-05,
      "loss": 0.4241,
      "step": 3284
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.357237190658258e-05,
      "loss": 0.4212,
      "step": 3285
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3517892000672745e-05,
      "loss": 0.4216,
      "step": 3286
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.346344743354025e-05,
      "loss": 0.4408,
      "step": 3287
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3409038234125276e-05,
      "loss": 0.4262,
      "step": 3288
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.335466443134919e-05,
      "loss": 0.4339,
      "step": 3289
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.330032605411457e-05,
      "loss": 0.4325,
      "step": 3290
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.324602313130514e-05,
      "loss": 0.4254,
      "step": 3291
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.319175569178582e-05,
      "loss": 0.4276,
      "step": 3292
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3137523764402636e-05,
      "loss": 0.4314,
      "step": 3293
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.308332737798274e-05,
      "loss": 0.4407,
      "step": 3294
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.302916656133439e-05,
      "loss": 0.4322,
      "step": 3295
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2975041343246936e-05,
      "loss": 0.4181,
      "step": 3296
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2920951752490814e-05,
      "loss": 0.4295,
      "step": 3297
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2866897817817535e-05,
      "loss": 0.4369,
      "step": 3298
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.281287956795962e-05,
      "loss": 0.4295,
      "step": 3299
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.275889703163062e-05,
      "loss": 0.4304,
      "step": 3300
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.270495023752518e-05,
      "loss": 0.43,
      "step": 3301
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.265103921431888e-05,
      "loss": 0.4297,
      "step": 3302
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.259716399066827e-05,
      "loss": 0.4167,
      "step": 3303
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.254332459521092e-05,
      "loss": 0.434,
      "step": 3304
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.248952105656532e-05,
      "loss": 0.4312,
      "step": 3305
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.243575340333092e-05,
      "loss": 0.4341,
      "step": 3306
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.238202166408809e-05,
      "loss": 0.4319,
      "step": 3307
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.232832586739809e-05,
      "loss": 0.4242,
      "step": 3308
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2274666041803093e-05,
      "loss": 0.4341,
      "step": 3309
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.222104221582618e-05,
      "loss": 0.4265,
      "step": 3310
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.2167454417971255e-05,
      "loss": 0.4287,
      "step": 3311
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.2113902676723085e-05,
      "loss": 0.4244,
      "step": 3312
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.2060387020547264e-05,
      "loss": 0.4251,
      "step": 3313
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.2006907477890215e-05,
      "loss": 0.4306,
      "step": 3314
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.195346407717912e-05,
      "loss": 0.4303,
      "step": 3315
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1900056846822046e-05,
      "loss": 0.4217,
      "step": 3316
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.184668581520778e-05,
      "loss": 0.4316,
      "step": 3317
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.179335101070582e-05,
      "loss": 0.434,
      "step": 3318
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.174005246166648e-05,
      "loss": 0.4322,
      "step": 3319
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.168679019642075e-05,
      "loss": 0.4207,
      "step": 3320
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.163356424328039e-05,
      "loss": 0.4312,
      "step": 3321
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.158037463053783e-05,
      "loss": 0.4338,
      "step": 3322
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1527221386466144e-05,
      "loss": 0.4317,
      "step": 3323
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.147410453931913e-05,
      "loss": 0.4325,
      "step": 3324
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.142102411733121e-05,
      "loss": 0.4339,
      "step": 3325
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.136798014871747e-05,
      "loss": 0.4347,
      "step": 3326
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.131497266167357e-05,
      "loss": 0.4258,
      "step": 3327
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.126200168437581e-05,
      "loss": 0.4271,
      "step": 3328
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.1209067244981114e-05,
      "loss": 0.4314,
      "step": 3329
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.115616937162695e-05,
      "loss": 0.4276,
      "step": 3330
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.110330809243134e-05,
      "loss": 0.4312,
      "step": 3331
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.105048343549287e-05,
      "loss": 0.4253,
      "step": 3332
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0997695428890636e-05,
      "loss": 0.4233,
      "step": 3333
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.09449441006843e-05,
      "loss": 0.4409,
      "step": 3334
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0892229478913995e-05,
      "loss": 0.4193,
      "step": 3335
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0839551591600324e-05,
      "loss": 0.437,
      "step": 3336
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.078691046674437e-05,
      "loss": 0.4197,
      "step": 3337
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.073430613232776e-05,
      "loss": 0.431,
      "step": 3338
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.068173861631245e-05,
      "loss": 0.4276,
      "step": 3339
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.062920794664089e-05,
      "loss": 0.4266,
      "step": 3340
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0576714151235916e-05,
      "loss": 0.4309,
      "step": 3341
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.052425725800078e-05,
      "loss": 0.4342,
      "step": 3342
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0471837294819084e-05,
      "loss": 0.4315,
      "step": 3343
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0419454289554862e-05,
      "loss": 0.4391,
      "step": 3344
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.036710827005246e-05,
      "loss": 0.4357,
      "step": 3345
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.031479926413652e-05,
      "loss": 0.4242,
      "step": 3346
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0262527299612143e-05,
      "loss": 0.4304,
      "step": 3347
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0210292404264627e-05,
      "loss": 0.427,
      "step": 3348
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.01580946058596e-05,
      "loss": 0.4257,
      "step": 3349
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.010593393214296e-05,
      "loss": 0.4229,
      "step": 3350
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0053810410840877e-05,
      "loss": 0.4278,
      "step": 3351
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0001724069659786e-05,
      "loss": 0.428,
      "step": 3352
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.9949674936286355e-05,
      "loss": 0.434,
      "step": 3353
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.9897663038387447e-05,
      "loss": 0.4329,
      "step": 3354
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9845688403610127e-05,
      "loss": 0.4199,
      "step": 3355
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9793751059581753e-05,
      "loss": 0.4169,
      "step": 3356
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.974185103390975e-05,
      "loss": 0.4319,
      "step": 3357
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.968998835418174e-05,
      "loss": 0.4415,
      "step": 3358
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9638163047965496e-05,
      "loss": 0.4357,
      "step": 3359
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9586375142808932e-05,
      "loss": 0.4336,
      "step": 3360
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9534624666240062e-05,
      "loss": 0.431,
      "step": 3361
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9482911645767018e-05,
      "loss": 0.4347,
      "step": 3362
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9431236108878014e-05,
      "loss": 0.4227,
      "step": 3363
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9379598083041382e-05,
      "loss": 0.4315,
      "step": 3364
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9327997595705457e-05,
      "loss": 0.4387,
      "step": 3365
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.927643467429866e-05,
      "loss": 0.4156,
      "step": 3366
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.922490934622937e-05,
      "loss": 0.4179,
      "step": 3367
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.917342163888611e-05,
      "loss": 0.4253,
      "step": 3368
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9121971579637318e-05,
      "loss": 0.429,
      "step": 3369
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.9070559195831426e-05,
      "loss": 0.4311,
      "step": 3370
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.9019184514796848e-05,
      "loss": 0.4308,
      "step": 3371
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8967847563841975e-05,
      "loss": 0.4381,
      "step": 3372
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.891654837025508e-05,
      "loss": 0.4408,
      "step": 3373
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8865286961304483e-05,
      "loss": 0.4236,
      "step": 3374
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.881406336423832e-05,
      "loss": 0.4235,
      "step": 3375
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8762877606284655e-05,
      "loss": 0.4289,
      "step": 3376
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8711729714651446e-05,
      "loss": 0.4216,
      "step": 3377
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8660619716526515e-05,
      "loss": 0.4297,
      "step": 3378
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.860954763907755e-05,
      "loss": 0.4337,
      "step": 3379
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8558513509452066e-05,
      "loss": 0.43,
      "step": 3380
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8507517354777423e-05,
      "loss": 0.4227,
      "step": 3381
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.845655920216077e-05,
      "loss": 0.4337,
      "step": 3382
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8405639078689128e-05,
      "loss": 0.4187,
      "step": 3383
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8354757011429223e-05,
      "loss": 0.4235,
      "step": 3384
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.8303913027427575e-05,
      "loss": 0.4412,
      "step": 3385
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.8253107153710467e-05,
      "loss": 0.4292,
      "step": 3386
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.820233941728393e-05,
      "loss": 0.426,
      "step": 3387
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.815160984513372e-05,
      "loss": 0.427,
      "step": 3388
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.81009184642253e-05,
      "loss": 0.4284,
      "step": 3389
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.805026530150383e-05,
      "loss": 0.4278,
      "step": 3390
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7999650383894138e-05,
      "loss": 0.4336,
      "step": 3391
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.79490737383008e-05,
      "loss": 0.4231,
      "step": 3392
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7898535391607973e-05,
      "loss": 0.4277,
      "step": 3393
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.784803537067946e-05,
      "loss": 0.4221,
      "step": 3394
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7797573702358726e-05,
      "loss": 0.4297,
      "step": 3395
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7747150413468814e-05,
      "loss": 0.4382,
      "step": 3396
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7696765530812407e-05,
      "loss": 0.4267,
      "step": 3397
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.764641908117174e-05,
      "loss": 0.4362,
      "step": 3398
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.759611109130863e-05,
      "loss": 0.4158,
      "step": 3399
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7545841587964416e-05,
      "loss": 0.4246,
      "step": 3400
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7495610597860078e-05,
      "loss": 0.418,
      "step": 3401
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7445418147696023e-05,
      "loss": 0.4322,
      "step": 3402
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7395264264152222e-05,
      "loss": 0.4246,
      "step": 3403
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7345148973888113e-05,
      "loss": 0.4255,
      "step": 3404
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7295072303542658e-05,
      "loss": 0.4259,
      "step": 3405
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.724503427973426e-05,
      "loss": 0.4242,
      "step": 3406
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7195034929060803e-05,
      "loss": 0.423,
      "step": 3407
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7145074278099593e-05,
      "loss": 0.4232,
      "step": 3408
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.709515235340736e-05,
      "loss": 0.4383,
      "step": 3409
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.704526918152026e-05,
      "loss": 0.4371,
      "step": 3410
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.6995424788953895e-05,
      "loss": 0.4326,
      "step": 3411
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.6945619202203187e-05,
      "loss": 0.4273,
      "step": 3412
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.6895852447742453e-05,
      "loss": 0.4175,
      "step": 3413
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6846124552025374e-05,
      "loss": 0.4276,
      "step": 3414
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6796435541484933e-05,
      "loss": 0.426,
      "step": 3415
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.674678544253356e-05,
      "loss": 0.4263,
      "step": 3416
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6697174281562877e-05,
      "loss": 0.4304,
      "step": 3417
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6647602084943867e-05,
      "loss": 0.4346,
      "step": 3418
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6598068879026782e-05,
      "loss": 0.4307,
      "step": 3419
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6548574690141125e-05,
      "loss": 0.4306,
      "step": 3420
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.649911954459576e-05,
      "loss": 0.4277,
      "step": 3421
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.64497034686787e-05,
      "loss": 0.4293,
      "step": 3422
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.64003264886572e-05,
      "loss": 0.4308,
      "step": 3423
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.635098863077776e-05,
      "loss": 0.4276,
      "step": 3424
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6301689921266094e-05,
      "loss": 0.4191,
      "step": 3425
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6252430386327053e-05,
      "loss": 0.4293,
      "step": 3426
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6203210052144735e-05,
      "loss": 0.427,
      "step": 3427
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6154028944882326e-05,
      "loss": 0.4199,
      "step": 3428
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.6104887090682252e-05,
      "loss": 0.4274,
      "step": 3429
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.6055784515665993e-05,
      "loss": 0.4146,
      "step": 3430
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.6006721245934197e-05,
      "loss": 0.4306,
      "step": 3431
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5957697307566586e-05,
      "loss": 0.4283,
      "step": 3432
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5908712726621997e-05,
      "loss": 0.4248,
      "step": 3433
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5859767529138334e-05,
      "loss": 0.4307,
      "step": 3434
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5810861741132586e-05,
      "loss": 0.4281,
      "step": 3435
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5761995388600767e-05,
      "loss": 0.4222,
      "step": 3436
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5713168497517926e-05,
      "loss": 0.4271,
      "step": 3437
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.566438109383821e-05,
      "loss": 0.429,
      "step": 3438
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.561563320349468e-05,
      "loss": 0.4243,
      "step": 3439
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5566924852399443e-05,
      "loss": 0.419,
      "step": 3440
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.551825606644358e-05,
      "loss": 0.4283,
      "step": 3441
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.546962687149713e-05,
      "loss": 0.4241,
      "step": 3442
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.54210372934091e-05,
      "loss": 0.4299,
      "step": 3443
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5372487358007446e-05,
      "loss": 0.4345,
      "step": 3444
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5323977091099038e-05,
      "loss": 0.4262,
      "step": 3445
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5275506518469628e-05,
      "loss": 0.4188,
      "step": 3446
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.522707566588397e-05,
      "loss": 0.4273,
      "step": 3447
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5178684559085608e-05,
      "loss": 0.4235,
      "step": 3448
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5130333223796976e-05,
      "loss": 0.4317,
      "step": 3449
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5082021685719394e-05,
      "loss": 0.4359,
      "step": 3450
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5033749970533015e-05,
      "loss": 0.4335,
      "step": 3451
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4985518103896833e-05,
      "loss": 0.437,
      "step": 3452
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.493732611144862e-05,
      "loss": 0.4349,
      "step": 3453
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4889174018805017e-05,
      "loss": 0.4352,
      "step": 3454
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4841061851561377e-05,
      "loss": 0.4312,
      "step": 3455
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.479298963529194e-05,
      "loss": 0.4324,
      "step": 3456
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4744957395549618e-05,
      "loss": 0.4197,
      "step": 3457
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4696965157866093e-05,
      "loss": 0.4344,
      "step": 3458
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4649012947751806e-05,
      "loss": 0.4241,
      "step": 3459
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4601100790695885e-05,
      "loss": 0.4401,
      "step": 3460
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.455322871216622e-05,
      "loss": 0.4214,
      "step": 3461
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4505396737609355e-05,
      "loss": 0.4255,
      "step": 3462
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4457604892450493e-05,
      "loss": 0.417,
      "step": 3463
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.44098532020936e-05,
      "loss": 0.4307,
      "step": 3464
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4362141691921235e-05,
      "loss": 0.4315,
      "step": 3465
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4314470387294574e-05,
      "loss": 0.4183,
      "step": 3466
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.426683931355347e-05,
      "loss": 0.4292,
      "step": 3467
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4219248496016332e-05,
      "loss": 0.4291,
      "step": 3468
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4171697959980277e-05,
      "loss": 0.4338,
      "step": 3469
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4124187730720917e-05,
      "loss": 0.4242,
      "step": 3470
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4076717833492457e-05,
      "loss": 0.4388,
      "step": 3471
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4029288293527675e-05,
      "loss": 0.4298,
      "step": 3472
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3981899136037878e-05,
      "loss": 0.4314,
      "step": 3473
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3934550386212973e-05,
      "loss": 0.4304,
      "step": 3474
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3887242069221307e-05,
      "loss": 0.4286,
      "step": 3475
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.383997421020977e-05,
      "loss": 0.4271,
      "step": 3476
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.379274683430376e-05,
      "loss": 0.4271,
      "step": 3477
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.374555996660711e-05,
      "loss": 0.4254,
      "step": 3478
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3698413632202165e-05,
      "loss": 0.4299,
      "step": 3479
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.365130785614972e-05,
      "loss": 0.4277,
      "step": 3480
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3604242663488997e-05,
      "loss": 0.4212,
      "step": 3481
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.355721807923761e-05,
      "loss": 0.4461,
      "step": 3482
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3510234128391682e-05,
      "loss": 0.4257,
      "step": 3483
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3463290835925667e-05,
      "loss": 0.421,
      "step": 3484
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.341638822679243e-05,
      "loss": 0.4175,
      "step": 3485
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3369526325923175e-05,
      "loss": 0.4341,
      "step": 3486
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.332270515822752e-05,
      "loss": 0.4311,
      "step": 3487
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.32759247485934e-05,
      "loss": 0.4131,
      "step": 3488
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3229185121887086e-05,
      "loss": 0.4218,
      "step": 3489
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3182486302953177e-05,
      "loss": 0.4249,
      "step": 3490
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3135828316614583e-05,
      "loss": 0.4299,
      "step": 3491
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3089211187672488e-05,
      "loss": 0.4256,
      "step": 3492
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3042634940906405e-05,
      "loss": 0.4374,
      "step": 3493
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2996099601074084e-05,
      "loss": 0.4213,
      "step": 3494
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.294960519291153e-05,
      "loss": 0.4366,
      "step": 3495
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.290315174113299e-05,
      "loss": 0.4297,
      "step": 3496
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.285673927043094e-05,
      "loss": 0.427,
      "step": 3497
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2810367805476075e-05,
      "loss": 0.437,
      "step": 3498
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2764037370917312e-05,
      "loss": 0.4312,
      "step": 3499
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2717747991381734e-05,
      "loss": 0.4163,
      "step": 3500
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2671499691474573e-05,
      "loss": 0.4259,
      "step": 3501
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2625292495779338e-05,
      "loss": 0.4267,
      "step": 3502
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2579126428857555e-05,
      "loss": 0.442,
      "step": 3503
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2533001515248965e-05,
      "loss": 0.447,
      "step": 3504
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2486917779471407e-05,
      "loss": 0.4299,
      "step": 3505
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.244087524602083e-05,
      "loss": 0.429,
      "step": 3506
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.239487393937131e-05,
      "loss": 0.4196,
      "step": 3507
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2348913883974976e-05,
      "loss": 0.425,
      "step": 3508
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2302995104262037e-05,
      "loss": 0.4389,
      "step": 3509
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2257117624640754e-05,
      "loss": 0.4376,
      "step": 3510
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.221128146949749e-05,
      "loss": 0.4283,
      "step": 3511
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2165486663196587e-05,
      "loss": 0.4233,
      "step": 3512
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2119733230080408e-05,
      "loss": 0.4282,
      "step": 3513
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2074021194469342e-05,
      "loss": 0.4266,
      "step": 3514
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.202835058066174e-05,
      "loss": 0.42,
      "step": 3515
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.1982721412934016e-05,
      "loss": 0.4238,
      "step": 3516
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.193713371554048e-05,
      "loss": 0.4376,
      "step": 3517
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1891587512713407e-05,
      "loss": 0.4374,
      "step": 3518
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.184608282866304e-05,
      "loss": 0.4254,
      "step": 3519
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1800619687577495e-05,
      "loss": 0.4235,
      "step": 3520
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.175519811362292e-05,
      "loss": 0.4163,
      "step": 3521
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.170981813094327e-05,
      "loss": 0.4186,
      "step": 3522
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1664479763660407e-05,
      "loss": 0.4345,
      "step": 3523
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1619183035874103e-05,
      "loss": 0.4255,
      "step": 3524
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1573927971661957e-05,
      "loss": 0.4321,
      "step": 3525
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1528714595079468e-05,
      "loss": 0.4241,
      "step": 3526
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1483542930159927e-05,
      "loss": 0.4169,
      "step": 3527
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.143841300091447e-05,
      "loss": 0.4245,
      "step": 3528
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1393324831332097e-05,
      "loss": 0.4274,
      "step": 3529
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1348278445379556e-05,
      "loss": 0.4343,
      "step": 3530
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1303273867001382e-05,
      "loss": 0.4383,
      "step": 3531
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.12583111201199e-05,
      "loss": 0.4336,
      "step": 3532
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.121339022863522e-05,
      "loss": 0.4226,
      "step": 3533
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.116851121642518e-05,
      "loss": 0.4361,
      "step": 3534
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.1123674107345347e-05,
      "loss": 0.4276,
      "step": 3535
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.107887892522905e-05,
      "loss": 0.4311,
      "step": 3536
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.1034125693887274e-05,
      "loss": 0.4327,
      "step": 3537
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0989414437108802e-05,
      "loss": 0.44,
      "step": 3538
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0944745178660007e-05,
      "loss": 0.4309,
      "step": 3539
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.090011794228499e-05,
      "loss": 0.4231,
      "step": 3540
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.08555327517055e-05,
      "loss": 0.4246,
      "step": 3541
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0810989630620926e-05,
      "loss": 0.4278,
      "step": 3542
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0766488602708312e-05,
      "loss": 0.4284,
      "step": 3543
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0722029691622336e-05,
      "loss": 0.4208,
      "step": 3544
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0677612920995247e-05,
      "loss": 0.4233,
      "step": 3545
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.063323831443692e-05,
      "loss": 0.431,
      "step": 3546
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.058890589553486e-05,
      "loss": 0.4316,
      "step": 3547
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0544615687854085e-05,
      "loss": 0.4345,
      "step": 3548
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0500367714937198e-05,
      "loss": 0.4294,
      "step": 3549
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0456162000304348e-05,
      "loss": 0.4312,
      "step": 3550
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0411998567453218e-05,
      "loss": 0.4214,
      "step": 3551
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0367877439859028e-05,
      "loss": 0.4274,
      "step": 3552
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.03237986409745e-05,
      "loss": 0.4333,
      "step": 3553
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0279762194229867e-05,
      "loss": 0.4241,
      "step": 3554
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0235768123032816e-05,
      "loss": 0.4259,
      "step": 3555
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0191816450768586e-05,
      "loss": 0.4255,
      "step": 3556
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.01479072007998e-05,
      "loss": 0.4305,
      "step": 3557
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0104040396466573e-05,
      "loss": 0.4321,
      "step": 3558
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0060216061086436e-05,
      "loss": 0.4238,
      "step": 3559
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.001643421795435e-05,
      "loss": 0.4271,
      "step": 3560
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.9972694890342704e-05,
      "loss": 0.4249,
      "step": 3561
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9928998101501262e-05,
      "loss": 0.4311,
      "step": 3562
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.988534387465719e-05,
      "loss": 0.425,
      "step": 3563
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9841732233015065e-05,
      "loss": 0.4323,
      "step": 3564
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.979816319975678e-05,
      "loss": 0.4412,
      "step": 3565
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9754636798041582e-05,
      "loss": 0.4289,
      "step": 3566
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.971115305100608e-05,
      "loss": 0.4407,
      "step": 3567
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.966771198176417e-05,
      "loss": 0.4262,
      "step": 3568
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9624313613407142e-05,
      "loss": 0.4332,
      "step": 3569
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.958095796900351e-05,
      "loss": 0.4264,
      "step": 3570
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.953764507159911e-05,
      "loss": 0.4271,
      "step": 3571
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9494374944217042e-05,
      "loss": 0.4427,
      "step": 3572
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.945114760985769e-05,
      "loss": 0.4341,
      "step": 3573
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.940796309149866e-05,
      "loss": 0.4296,
      "step": 3574
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9364821412094857e-05,
      "loss": 0.4276,
      "step": 3575
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9321722594578374e-05,
      "loss": 0.4275,
      "step": 3576
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9278666661858503e-05,
      "loss": 0.427,
      "step": 3577
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9235653636821783e-05,
      "loss": 0.4276,
      "step": 3578
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9192683542331913e-05,
      "loss": 0.4278,
      "step": 3579
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.914975640122978e-05,
      "loss": 0.4211,
      "step": 3580
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9106872236333463e-05,
      "loss": 0.432,
      "step": 3581
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.906403107043815e-05,
      "loss": 0.4188,
      "step": 3582
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.90212329263162e-05,
      "loss": 0.415,
      "step": 3583
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8978477826717135e-05,
      "loss": 0.425,
      "step": 3584
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8935765794367556e-05,
      "loss": 0.4299,
      "step": 3585
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.889309685197116e-05,
      "loss": 0.4378,
      "step": 3586
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8850471022208783e-05,
      "loss": 0.4209,
      "step": 3587
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.880788832773831e-05,
      "loss": 0.421,
      "step": 3588
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8765348791194705e-05,
      "loss": 0.4397,
      "step": 3589
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.872285243519001e-05,
      "loss": 0.43,
      "step": 3590
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8680399282313278e-05,
      "loss": 0.4143,
      "step": 3591
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.86379893551306e-05,
      "loss": 0.4263,
      "step": 3592
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8595622676185164e-05,
      "loss": 0.4301,
      "step": 3593
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8553299267997092e-05,
      "loss": 0.4428,
      "step": 3594
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8511019153063524e-05,
      "loss": 0.4186,
      "step": 3595
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8468782353858584e-05,
      "loss": 0.4302,
      "step": 3596
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8426588892833375e-05,
      "loss": 0.4268,
      "step": 3597
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8384438792415982e-05,
      "loss": 0.4271,
      "step": 3598
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8342332075011415e-05,
      "loss": 0.4442,
      "step": 3599
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.830026876300164e-05,
      "loss": 0.432,
      "step": 3600
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.825824887874551e-05,
      "loss": 0.4259,
      "step": 3601
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8216272444578897e-05,
      "loss": 0.4311,
      "step": 3602
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.817433948281446e-05,
      "loss": 0.4156,
      "step": 3603
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.813245001574182e-05,
      "loss": 0.4239,
      "step": 3604
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.809060406562745e-05,
      "loss": 0.4239,
      "step": 3605
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8048801654714688e-05,
      "loss": 0.4257,
      "step": 3606
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.800704280522376e-05,
      "loss": 0.426,
      "step": 3607
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7965327539351695e-05,
      "loss": 0.4297,
      "step": 3608
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7923655879272393e-05,
      "loss": 0.4229,
      "step": 3609
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.788202784713653e-05,
      "loss": 0.4279,
      "step": 3610
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7840443465071653e-05,
      "loss": 0.4236,
      "step": 3611
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.779890275518208e-05,
      "loss": 0.4269,
      "step": 3612
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7757405739548882e-05,
      "loss": 0.4324,
      "step": 3613
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.771595244022993e-05,
      "loss": 0.4271,
      "step": 3614
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.767454287925988e-05,
      "loss": 0.4329,
      "step": 3615
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7633177078650075e-05,
      "loss": 0.4229,
      "step": 3616
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7591855060388686e-05,
      "loss": 0.4155,
      "step": 3617
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7550576846440558e-05,
      "loss": 0.4237,
      "step": 3618
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7509342458747237e-05,
      "loss": 0.4302,
      "step": 3619
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.746815191922697e-05,
      "loss": 0.4283,
      "step": 3620
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7427005249774784e-05,
      "loss": 0.4212,
      "step": 3621
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7385902472262296e-05,
      "loss": 0.4485,
      "step": 3622
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.73448436085378e-05,
      "loss": 0.4246,
      "step": 3623
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.730382868042629e-05,
      "loss": 0.4308,
      "step": 3624
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7262857709729373e-05,
      "loss": 0.4266,
      "step": 3625
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7221930718225298e-05,
      "loss": 0.4394,
      "step": 3626
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.718104772766893e-05,
      "loss": 0.431,
      "step": 3627
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7140208759791753e-05,
      "loss": 0.4384,
      "step": 3628
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.709941383630188e-05,
      "loss": 0.4337,
      "step": 3629
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7058662978883955e-05,
      "loss": 0.4166,
      "step": 3630
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7017956209199247e-05,
      "loss": 0.4206,
      "step": 3631
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6977293548885554e-05,
      "loss": 0.4225,
      "step": 3632
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6936675019557257e-05,
      "loss": 0.4266,
      "step": 3633
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6896100642805246e-05,
      "loss": 0.4206,
      "step": 3634
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6855570440196978e-05,
      "loss": 0.4351,
      "step": 3635
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6815084433276394e-05,
      "loss": 0.422,
      "step": 3636
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6774642643563953e-05,
      "loss": 0.4145,
      "step": 3637
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6734245092556665e-05,
      "loss": 0.4215,
      "step": 3638
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6693891801727945e-05,
      "loss": 0.4122,
      "step": 3639
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6653582792527722e-05,
      "loss": 0.4176,
      "step": 3640
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.661331808638238e-05,
      "loss": 0.4167,
      "step": 3641
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.657309770469473e-05,
      "loss": 0.4319,
      "step": 3642
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.653292166884407e-05,
      "loss": 0.4323,
      "step": 3643
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6492790000186098e-05,
      "loss": 0.4303,
      "step": 3644
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6452702720052925e-05,
      "loss": 0.4188,
      "step": 3645
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6412659849753042e-05,
      "loss": 0.4214,
      "step": 3646
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.637266141057142e-05,
      "loss": 0.422,
      "step": 3647
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6332707423769343e-05,
      "loss": 0.4224,
      "step": 3648
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6292797910584467e-05,
      "loss": 0.4321,
      "step": 3649
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.625293289223082e-05,
      "loss": 0.4211,
      "step": 3650
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.6213112389898788e-05,
      "loss": 0.4328,
      "step": 3651
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.617333642475509e-05,
      "loss": 0.4181,
      "step": 3652
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.6133605017942754e-05,
      "loss": 0.4261,
      "step": 3653
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.6093918190581147e-05,
      "loss": 0.4301,
      "step": 3654
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.6054275963765907e-05,
      "loss": 0.4311,
      "step": 3655
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.6014678358569024e-05,
      "loss": 0.4363,
      "step": 3656
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5975125396038725e-05,
      "loss": 0.4238,
      "step": 3657
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5935617097199494e-05,
      "loss": 0.4314,
      "step": 3658
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5896153483052102e-05,
      "loss": 0.4334,
      "step": 3659
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.585673457457356e-05,
      "loss": 0.4184,
      "step": 3660
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5817360392717118e-05,
      "loss": 0.4159,
      "step": 3661
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.577803095841225e-05,
      "loss": 0.4341,
      "step": 3662
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.57387462925646e-05,
      "loss": 0.4253,
      "step": 3663
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.569950641605613e-05,
      "loss": 0.4274,
      "step": 3664
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5660311349744884e-05,
      "loss": 0.4288,
      "step": 3665
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.562116111446513e-05,
      "loss": 0.4373,
      "step": 3666
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.55820557310273e-05,
      "loss": 0.4349,
      "step": 3667
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.554299522021796e-05,
      "loss": 0.4346,
      "step": 3668
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.55039796027999e-05,
      "loss": 0.4268,
      "step": 3669
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.546500889951198e-05,
      "loss": 0.4165,
      "step": 3670
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.542608313106918e-05,
      "loss": 0.4248,
      "step": 3671
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.538720231816264e-05,
      "loss": 0.4245,
      "step": 3672
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5348366481459565e-05,
      "loss": 0.4335,
      "step": 3673
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5309575641603248e-05,
      "loss": 0.4097,
      "step": 3674
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5270829819213138e-05,
      "loss": 0.4209,
      "step": 3675
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.523212903488468e-05,
      "loss": 0.4357,
      "step": 3676
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5193473309189388e-05,
      "loss": 0.4292,
      "step": 3677
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.515486266267484e-05,
      "loss": 0.4243,
      "step": 3678
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5116297115864642e-05,
      "loss": 0.4211,
      "step": 3679
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.507777668925846e-05,
      "loss": 0.4343,
      "step": 3680
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.5039301403331919e-05,
      "loss": 0.4344,
      "step": 3681
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.50008712785367e-05,
      "loss": 0.4387,
      "step": 3682
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.496248633530043e-05,
      "loss": 0.4286,
      "step": 3683
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4924146594026788e-05,
      "loss": 0.4261,
      "step": 3684
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4885852075095374e-05,
      "loss": 0.4247,
      "step": 3685
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4847602798861759e-05,
      "loss": 0.4302,
      "step": 3686
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4809398785657469e-05,
      "loss": 0.4288,
      "step": 3687
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4771240055789958e-05,
      "loss": 0.4318,
      "step": 3688
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4733126629542638e-05,
      "loss": 0.4241,
      "step": 3689
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.469505852717481e-05,
      "loss": 0.4209,
      "step": 3690
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4657035768921701e-05,
      "loss": 0.4215,
      "step": 3691
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.461905837499441e-05,
      "loss": 0.4219,
      "step": 3692
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4581126365579989e-05,
      "loss": 0.432,
      "step": 3693
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4543239760841287e-05,
      "loss": 0.4251,
      "step": 3694
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4505398580917052e-05,
      "loss": 0.4215,
      "step": 3695
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.446760284592189e-05,
      "loss": 0.4276,
      "step": 3696
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4429852575946235e-05,
      "loss": 0.4098,
      "step": 3697
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4392147791056388e-05,
      "loss": 0.4248,
      "step": 3698
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4354488511294417e-05,
      "loss": 0.4243,
      "step": 3699
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4316874756678244e-05,
      "loss": 0.4166,
      "step": 3700
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4279306547201577e-05,
      "loss": 0.4223,
      "step": 3701
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4241783902833938e-05,
      "loss": 0.4234,
      "step": 3702
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4204306843520586e-05,
      "loss": 0.4225,
      "step": 3703
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.41668753891826e-05,
      "loss": 0.4199,
      "step": 3704
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4129489559716758e-05,
      "loss": 0.4247,
      "step": 3705
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4092149374995634e-05,
      "loss": 0.4228,
      "step": 3706
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4054854854867517e-05,
      "loss": 0.4249,
      "step": 3707
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4017606019156427e-05,
      "loss": 0.4304,
      "step": 3708
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.3980402887662104e-05,
      "loss": 0.4284,
      "step": 3709
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3943245480159973e-05,
      "loss": 0.4268,
      "step": 3710
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3906133816401223e-05,
      "loss": 0.4238,
      "step": 3711
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3869067916112644e-05,
      "loss": 0.4247,
      "step": 3712
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3832047798996727e-05,
      "loss": 0.4281,
      "step": 3713
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3795073484731647e-05,
      "loss": 0.4233,
      "step": 3714
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3758144992971222e-05,
      "loss": 0.4136,
      "step": 3715
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3721262343344876e-05,
      "loss": 0.4194,
      "step": 3716
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3684425555457758e-05,
      "loss": 0.4244,
      "step": 3717
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3647634648890551e-05,
      "loss": 0.4294,
      "step": 3718
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3610889643199565e-05,
      "loss": 0.4321,
      "step": 3719
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3574190557916744e-05,
      "loss": 0.4223,
      "step": 3720
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3537537412549584e-05,
      "loss": 0.4233,
      "step": 3721
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.350093022658121e-05,
      "loss": 0.4377,
      "step": 3722
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.346436901947028e-05,
      "loss": 0.4368,
      "step": 3723
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3427853810651014e-05,
      "loss": 0.4334,
      "step": 3724
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3391384619533198e-05,
      "loss": 0.4236,
      "step": 3725
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3354961465502125e-05,
      "loss": 0.4244,
      "step": 3726
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3318584367918664e-05,
      "loss": 0.4238,
      "step": 3727
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3282253346119144e-05,
      "loss": 0.4296,
      "step": 3728
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3245968419415478e-05,
      "loss": 0.4385,
      "step": 3729
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3209729607095023e-05,
      "loss": 0.4275,
      "step": 3730
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3173536928420627e-05,
      "loss": 0.4243,
      "step": 3731
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3137390402630623e-05,
      "loss": 0.429,
      "step": 3732
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.310129004893882e-05,
      "loss": 0.4348,
      "step": 3733
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3065235886534467e-05,
      "loss": 0.428,
      "step": 3734
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3029227934582277e-05,
      "loss": 0.4369,
      "step": 3735
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2993266212222387e-05,
      "loss": 0.4234,
      "step": 3736
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.295735073857035e-05,
      "loss": 0.4273,
      "step": 3737
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2921481532717184e-05,
      "loss": 0.4196,
      "step": 3738
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2885658613729278e-05,
      "loss": 0.4154,
      "step": 3739
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2849882000648395e-05,
      "loss": 0.4255,
      "step": 3740
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2814151712491729e-05,
      "loss": 0.4231,
      "step": 3741
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2778467768251823e-05,
      "loss": 0.4274,
      "step": 3742
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2742830186896581e-05,
      "loss": 0.4324,
      "step": 3743
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2707238987369286e-05,
      "loss": 0.4254,
      "step": 3744
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2671694188588545e-05,
      "loss": 0.4258,
      "step": 3745
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2636195809448314e-05,
      "loss": 0.4213,
      "step": 3746
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2600743868817854e-05,
      "loss": 0.4249,
      "step": 3747
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2565338385541792e-05,
      "loss": 0.4168,
      "step": 3748
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2529979378440004e-05,
      "loss": 0.4209,
      "step": 3749
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2494666866307691e-05,
      "loss": 0.4236,
      "step": 3750
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2459400867915328e-05,
      "loss": 0.427,
      "step": 3751
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2424181402008672e-05,
      "loss": 0.4185,
      "step": 3752
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.238900848730874e-05,
      "loss": 0.4248,
      "step": 3753
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2353882142511797e-05,
      "loss": 0.4227,
      "step": 3754
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2318802386289386e-05,
      "loss": 0.4183,
      "step": 3755
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2283769237288223e-05,
      "loss": 0.4235,
      "step": 3756
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2248782714130347e-05,
      "loss": 0.4096,
      "step": 3757
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2213842835412926e-05,
      "loss": 0.4305,
      "step": 3758
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2178949619708368e-05,
      "loss": 0.4331,
      "step": 3759
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2144103085564274e-05,
      "loss": 0.4273,
      "step": 3760
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2109303251503434e-05,
      "loss": 0.4253,
      "step": 3761
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2074550136023799e-05,
      "loss": 0.4153,
      "step": 3762
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.203984375759849e-05,
      "loss": 0.4259,
      "step": 3763
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2005184134675829e-05,
      "loss": 0.4244,
      "step": 3764
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1970571285679233e-05,
      "loss": 0.4126,
      "step": 3765
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1936005229007274e-05,
      "loss": 0.4246,
      "step": 3766
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1901485983033645e-05,
      "loss": 0.4081,
      "step": 3767
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1867013566107132e-05,
      "loss": 0.4405,
      "step": 3768
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1832587996551725e-05,
      "loss": 0.4256,
      "step": 3769
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1798209292666407e-05,
      "loss": 0.4302,
      "step": 3770
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1763877472725283e-05,
      "loss": 0.4252,
      "step": 3771
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1729592554977552e-05,
      "loss": 0.4183,
      "step": 3772
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1695354557647487e-05,
      "loss": 0.4329,
      "step": 3773
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1661163498934346e-05,
      "loss": 0.4206,
      "step": 3774
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1627019397012573e-05,
      "loss": 0.4219,
      "step": 3775
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1592922270031536e-05,
      "loss": 0.4245,
      "step": 3776
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1558872136115673e-05,
      "loss": 0.4158,
      "step": 3777
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1524869013364447e-05,
      "loss": 0.43,
      "step": 3778
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1490912919852326e-05,
      "loss": 0.4285,
      "step": 3779
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.145700387362877e-05,
      "loss": 0.4145,
      "step": 3780
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1423141892718258e-05,
      "loss": 0.4231,
      "step": 3781
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.138932699512022e-05,
      "loss": 0.4266,
      "step": 3782
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1355559198809063e-05,
      "loss": 0.4273,
      "step": 3783
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1321838521734207e-05,
      "loss": 0.4165,
      "step": 3784
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1288164981819948e-05,
      "loss": 0.4309,
      "step": 3785
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1254538596965592e-05,
      "loss": 0.43,
      "step": 3786
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1220959385045326e-05,
      "loss": 0.4215,
      "step": 3787
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.118742736390831e-05,
      "loss": 0.4258,
      "step": 3788
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1153942551378582e-05,
      "loss": 0.4315,
      "step": 3789
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1120504965255107e-05,
      "loss": 0.4321,
      "step": 3790
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1087114623311735e-05,
      "loss": 0.4196,
      "step": 3791
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1053771543297198e-05,
      "loss": 0.4261,
      "step": 3792
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1020475742935143e-05,
      "loss": 0.4173,
      "step": 3793
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0987227239924059e-05,
      "loss": 0.4324,
      "step": 3794
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0954026051937282e-05,
      "loss": 0.4352,
      "step": 3795
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.092087219662301e-05,
      "loss": 0.4351,
      "step": 3796
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0887765691604279e-05,
      "loss": 0.4225,
      "step": 3797
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0854706554478966e-05,
      "loss": 0.4373,
      "step": 3798
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0821694802819759e-05,
      "loss": 0.4325,
      "step": 3799
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0788730454174156e-05,
      "loss": 0.4329,
      "step": 3800
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.075581352606444e-05,
      "loss": 0.4267,
      "step": 3801
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0722944035987759e-05,
      "loss": 0.4297,
      "step": 3802
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.069012200141597e-05,
      "loss": 0.423,
      "step": 3803
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.065734743979573e-05,
      "loss": 0.4252,
      "step": 3804
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0624620368548455e-05,
      "loss": 0.4155,
      "step": 3805
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0591940805070332e-05,
      "loss": 0.4229,
      "step": 3806
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.05593087667323e-05,
      "loss": 0.4241,
      "step": 3807
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0526724270880006e-05,
      "loss": 0.4172,
      "step": 3808
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.049418733483385e-05,
      "loss": 0.423,
      "step": 3809
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0461697975888917e-05,
      "loss": 0.4145,
      "step": 3810
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0429256211315074e-05,
      "loss": 0.4202,
      "step": 3811
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0396862058356838e-05,
      "loss": 0.4156,
      "step": 3812
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0364515534233399e-05,
      "loss": 0.4317,
      "step": 3813
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0332216656138683e-05,
      "loss": 0.4156,
      "step": 3814
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0299965441241233e-05,
      "loss": 0.4269,
      "step": 3815
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0267761906684292e-05,
      "loss": 0.4246,
      "step": 3816
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.023560606958578e-05,
      "loss": 0.4364,
      "step": 3817
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.020349794703822e-05,
      "loss": 0.421,
      "step": 3818
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0171437556108777e-05,
      "loss": 0.435,
      "step": 3819
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0139424913839258e-05,
      "loss": 0.4329,
      "step": 3820
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0107460037246075e-05,
      "loss": 0.4214,
      "step": 3821
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0075542943320294e-05,
      "loss": 0.4217,
      "step": 3822
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0043673649027518e-05,
      "loss": 0.4216,
      "step": 3823
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0011852171308e-05,
      "loss": 0.4236,
      "step": 3824
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.98007852707652e-06,
      "loss": 0.4233,
      "step": 3825
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.948352733222466e-06,
      "loss": 0.4155,
      "step": 3826
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.916674806609804e-06,
      "loss": 0.4178,
      "step": 3827
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.885044764077012e-06,
      "loss": 0.4365,
      "step": 3828
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.853462622437148e-06,
      "loss": 0.4306,
      "step": 3829
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.821928398477809e-06,
      "loss": 0.4244,
      "step": 3830
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.790442108961117e-06,
      "loss": 0.4279,
      "step": 3831
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.759003770623698e-06,
      "loss": 0.4206,
      "step": 3832
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.727613400176705e-06,
      "loss": 0.4203,
      "step": 3833
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.696271014305792e-06,
      "loss": 0.4277,
      "step": 3834
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.664976629671097e-06,
      "loss": 0.4376,
      "step": 3835
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.63373026290726e-06,
      "loss": 0.4284,
      "step": 3836
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.602531930623381e-06,
      "loss": 0.4376,
      "step": 3837
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.571381649403021e-06,
      "loss": 0.4325,
      "step": 3838
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.540279435804233e-06,
      "loss": 0.4258,
      "step": 3839
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.509225306359503e-06,
      "loss": 0.4265,
      "step": 3840
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.47821927757574e-06,
      "loss": 0.4137,
      "step": 3841
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.447261365934302e-06,
      "loss": 0.4172,
      "step": 3842
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.416351587890948e-06,
      "loss": 0.4251,
      "step": 3843
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.385489959875903e-06,
      "loss": 0.4268,
      "step": 3844
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.354676498293735e-06,
      "loss": 0.4234,
      "step": 3845
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.323911219523451e-06,
      "loss": 0.4235,
      "step": 3846
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.293194139918404e-06,
      "loss": 0.4192,
      "step": 3847
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.262525275806411e-06,
      "loss": 0.4299,
      "step": 3848
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.231904643489575e-06,
      "loss": 0.4133,
      "step": 3849
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.2013322592444e-06,
      "loss": 0.4334,
      "step": 3850
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.170808139321718e-06,
      "loss": 0.4334,
      "step": 3851
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.14033229994674e-06,
      "loss": 0.4301,
      "step": 3852
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.109904757318988e-06,
      "loss": 0.4327,
      "step": 3853
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.07952552761232e-06,
      "loss": 0.4135,
      "step": 3854
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.04919462697491e-06,
      "loss": 0.4303,
      "step": 3855
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.01891207152923e-06,
      "loss": 0.4262,
      "step": 3856
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.988677877372108e-06,
      "loss": 0.4216,
      "step": 3857
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.958492060574598e-06,
      "loss": 0.4201,
      "step": 3858
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.92835463718209e-06,
      "loss": 0.4313,
      "step": 3859
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.898265623214208e-06,
      "loss": 0.4235,
      "step": 3860
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.868225034664868e-06,
      "loss": 0.4324,
      "step": 3861
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.83823288750224e-06,
      "loss": 0.4233,
      "step": 3862
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.808289197668752e-06,
      "loss": 0.4304,
      "step": 3863
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.778393981081046e-06,
      "loss": 0.4136,
      "step": 3864
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.748547253630057e-06,
      "loss": 0.428,
      "step": 3865
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.718749031180884e-06,
      "loss": 0.4251,
      "step": 3866
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.68899932957289e-06,
      "loss": 0.4331,
      "step": 3867
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.65929816461959e-06,
      "loss": 0.4209,
      "step": 3868
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.629645552108744e-06,
      "loss": 0.4171,
      "step": 3869
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.600041507802314e-06,
      "loss": 0.4234,
      "step": 3870
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.57048604743641e-06,
      "loss": 0.4296,
      "step": 3871
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.540979186721331e-06,
      "loss": 0.4366,
      "step": 3872
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.511520941341544e-06,
      "loss": 0.4286,
      "step": 3873
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.482111326955666e-06,
      "loss": 0.4256,
      "step": 3874
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.452750359196493e-06,
      "loss": 0.4265,
      "step": 3875
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.423438053670918e-06,
      "loss": 0.4288,
      "step": 3876
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.39417442596001e-06,
      "loss": 0.4203,
      "step": 3877
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.36495949161894e-06,
      "loss": 0.4315,
      "step": 3878
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.335793266176995e-06,
      "loss": 0.4275,
      "step": 3879
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.306675765137561e-06,
      "loss": 0.4237,
      "step": 3880
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.27760700397816e-06,
      "loss": 0.4287,
      "step": 3881
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.248586998150387e-06,
      "loss": 0.426,
      "step": 3882
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.219615763079891e-06,
      "loss": 0.427,
      "step": 3883
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.190693314166453e-06,
      "loss": 0.4277,
      "step": 3884
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.161819666783888e-06,
      "loss": 0.4299,
      "step": 3885
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.132994836280073e-06,
      "loss": 0.4273,
      "step": 3886
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.10421883797694e-06,
      "loss": 0.4356,
      "step": 3887
    },
    {
      "epoch": 2.63,
      "learning_rate": 8.075491687170455e-06,
      "loss": 0.4225,
      "step": 3888
    },
    {
      "epoch": 2.63,
      "learning_rate": 8.046813399130637e-06,
      "loss": 0.4219,
      "step": 3889
    },
    {
      "epoch": 2.63,
      "learning_rate": 8.018183989101524e-06,
      "loss": 0.4332,
      "step": 3890
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.989603472301155e-06,
      "loss": 0.429,
      "step": 3891
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.961071863921598e-06,
      "loss": 0.4364,
      "step": 3892
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.932589179128936e-06,
      "loss": 0.4237,
      "step": 3893
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.904155433063221e-06,
      "loss": 0.4183,
      "step": 3894
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.875770640838509e-06,
      "loss": 0.431,
      "step": 3895
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.847434817542821e-06,
      "loss": 0.427,
      "step": 3896
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.81914797823815e-06,
      "loss": 0.4231,
      "step": 3897
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.790910137960462e-06,
      "loss": 0.4227,
      "step": 3898
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.762721311719678e-06,
      "loss": 0.4299,
      "step": 3899
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.73458151449965e-06,
      "loss": 0.4342,
      "step": 3900
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.706490761258157e-06,
      "loss": 0.4329,
      "step": 3901
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.678449066926963e-06,
      "loss": 0.4218,
      "step": 3902
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.650456446411714e-06,
      "loss": 0.4204,
      "step": 3903
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.622512914591973e-06,
      "loss": 0.423,
      "step": 3904
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.594618486321215e-06,
      "loss": 0.4226,
      "step": 3905
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.5667731764268e-06,
      "loss": 0.4259,
      "step": 3906
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.538976999710001e-06,
      "loss": 0.4401,
      "step": 3907
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.511229970945977e-06,
      "loss": 0.4275,
      "step": 3908
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.483532104883728e-06,
      "loss": 0.4265,
      "step": 3909
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.455883416246156e-06,
      "loss": 0.4352,
      "step": 3910
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.4282839197299835e-06,
      "loss": 0.4387,
      "step": 3911
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.400733630005863e-06,
      "loss": 0.4197,
      "step": 3912
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.373232561718202e-06,
      "loss": 0.4197,
      "step": 3913
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.345780729485285e-06,
      "loss": 0.4314,
      "step": 3914
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.318378147899241e-06,
      "loss": 0.42,
      "step": 3915
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.291024831525961e-06,
      "loss": 0.4327,
      "step": 3916
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.263720794905227e-06,
      "loss": 0.43,
      "step": 3917
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.2364660525505835e-06,
      "loss": 0.4115,
      "step": 3918
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.2092606189493535e-06,
      "loss": 0.421,
      "step": 3919
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.1821045085626815e-06,
      "loss": 0.4334,
      "step": 3920
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.154997735825486e-06,
      "loss": 0.4319,
      "step": 3921
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.1279403151464645e-06,
      "loss": 0.4207,
      "step": 3922
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.1009322609080775e-06,
      "loss": 0.4344,
      "step": 3923
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.073973587466532e-06,
      "loss": 0.43,
      "step": 3924
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.047064309151797e-06,
      "loss": 0.4293,
      "step": 3925
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.020204440267597e-06,
      "loss": 0.4265,
      "step": 3926
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.99339399509138e-06,
      "loss": 0.4256,
      "step": 3927
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.966632987874311e-06,
      "loss": 0.4346,
      "step": 3928
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.939921432841279e-06,
      "loss": 0.4219,
      "step": 3929
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.913259344190926e-06,
      "loss": 0.4241,
      "step": 3930
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.88664673609557e-06,
      "loss": 0.4295,
      "step": 3931
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.860083622701208e-06,
      "loss": 0.4309,
      "step": 3932
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.833570018127566e-06,
      "loss": 0.4253,
      "step": 3933
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.807105936468017e-06,
      "loss": 0.4266,
      "step": 3934
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.78069139178964e-06,
      "loss": 0.4274,
      "step": 3935
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.75432639813316e-06,
      "loss": 0.4164,
      "step": 3936
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.728010969512999e-06,
      "loss": 0.4218,
      "step": 3937
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.7017451199171576e-06,
      "loss": 0.4306,
      "step": 3938
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.67552886330739e-06,
      "loss": 0.4218,
      "step": 3939
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.6493622136190035e-06,
      "loss": 0.4102,
      "step": 3940
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.6232451847609735e-06,
      "loss": 0.434,
      "step": 3941
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.597177790615872e-06,
      "loss": 0.4338,
      "step": 3942
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.571160045039925e-06,
      "loss": 0.4171,
      "step": 3943
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.54519196186294e-06,
      "loss": 0.4295,
      "step": 3944
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.5192735548883385e-06,
      "loss": 0.4321,
      "step": 3945
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.493404837893124e-06,
      "loss": 0.4194,
      "step": 3946
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.467585824627887e-06,
      "loss": 0.4325,
      "step": 3947
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.441816528816835e-06,
      "loss": 0.4317,
      "step": 3948
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.416096964157714e-06,
      "loss": 0.4281,
      "step": 3949
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.390427144321831e-06,
      "loss": 0.4229,
      "step": 3950
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.36480708295406e-06,
      "loss": 0.4194,
      "step": 3951
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.3392367936728315e-06,
      "loss": 0.4284,
      "step": 3952
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.3137162900701155e-06,
      "loss": 0.4262,
      "step": 3953
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.288245585711416e-06,
      "loss": 0.4412,
      "step": 3954
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.26282469413576e-06,
      "loss": 0.4114,
      "step": 3955
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.237453628855694e-06,
      "loss": 0.421,
      "step": 3956
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.2121324033573294e-06,
      "loss": 0.4212,
      "step": 3957
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.1868610311002106e-06,
      "loss": 0.429,
      "step": 3958
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.161639525517427e-06,
      "loss": 0.4287,
      "step": 3959
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.136467900015552e-06,
      "loss": 0.4257,
      "step": 3960
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.111346167974641e-06,
      "loss": 0.4251,
      "step": 3961
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.0862743427482324e-06,
      "loss": 0.4322,
      "step": 3962
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.0612524376633336e-06,
      "loss": 0.4226,
      "step": 3963
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.036280466020405e-06,
      "loss": 0.4283,
      "step": 3964
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.011358441093407e-06,
      "loss": 0.4208,
      "step": 3965
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.986486376129708e-06,
      "loss": 0.4299,
      "step": 3966
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.9616642843501326e-06,
      "loss": 0.4268,
      "step": 3967
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.936892178948939e-06,
      "loss": 0.4252,
      "step": 3968
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.912170073093803e-06,
      "loss": 0.414,
      "step": 3969
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.887497979925882e-06,
      "loss": 0.4304,
      "step": 3970
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.862875912559685e-06,
      "loss": 0.4336,
      "step": 3971
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.838303884083151e-06,
      "loss": 0.4296,
      "step": 3972
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.8137819075576135e-06,
      "loss": 0.4292,
      "step": 3973
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.7893099960178e-06,
      "loss": 0.4197,
      "step": 3974
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.76488816247186e-06,
      "loss": 0.4224,
      "step": 3975
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.740516419901276e-06,
      "loss": 0.4287,
      "step": 3976
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.716194781260942e-06,
      "loss": 0.4299,
      "step": 3977
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.691923259479093e-06,
      "loss": 0.4317,
      "step": 3978
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.667701867457331e-06,
      "loss": 0.4316,
      "step": 3979
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.643530618070625e-06,
      "loss": 0.4235,
      "step": 3980
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.619409524167274e-06,
      "loss": 0.4304,
      "step": 3981
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.5953385985689114e-06,
      "loss": 0.4269,
      "step": 3982
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.571317854070523e-06,
      "loss": 0.4235,
      "step": 3983
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.54734730344042e-06,
      "loss": 0.4312,
      "step": 3984
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.52342695942023e-06,
      "loss": 0.4227,
      "step": 3985
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.499556834724873e-06,
      "loss": 0.4224,
      "step": 3986
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.475736942042598e-06,
      "loss": 0.4214,
      "step": 3987
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.451967294034943e-06,
      "loss": 0.4314,
      "step": 3988
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.428247903336736e-06,
      "loss": 0.4255,
      "step": 3989
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.404578782556102e-06,
      "loss": 0.4265,
      "step": 3990
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.380959944274444e-06,
      "loss": 0.4247,
      "step": 3991
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.3573914010464074e-06,
      "loss": 0.4261,
      "step": 3992
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.333873165399938e-06,
      "loss": 0.4221,
      "step": 3993
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.3104052498362476e-06,
      "loss": 0.4201,
      "step": 3994
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.286987666829768e-06,
      "loss": 0.4198,
      "step": 3995
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.263620428828186e-06,
      "loss": 0.4392,
      "step": 3996
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.240303548252446e-06,
      "loss": 0.4327,
      "step": 3997
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.217037037496686e-06,
      "loss": 0.4202,
      "step": 3998
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.193820908928315e-06,
      "loss": 0.4236,
      "step": 3999
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.170655174887939e-06,
      "loss": 0.4176,
      "step": 4000
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.147539847689364e-06,
      "loss": 0.4271,
      "step": 4001
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.124474939619617e-06,
      "loss": 0.4231,
      "step": 4002
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.1014604629389455e-06,
      "loss": 0.4232,
      "step": 4003
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.078496429880753e-06,
      "loss": 0.4296,
      "step": 4004
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.0555828526516525e-06,
      "loss": 0.4179,
      "step": 4005
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.032719743431414e-06,
      "loss": 0.4243,
      "step": 4006
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.009907114373013e-06,
      "loss": 0.4365,
      "step": 4007
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.987144977602565e-06,
      "loss": 0.4295,
      "step": 4008
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.964433345219355e-06,
      "loss": 0.4325,
      "step": 4009
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.941772229295838e-06,
      "loss": 0.423,
      "step": 4010
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.919161641877579e-06,
      "loss": 0.4237,
      "step": 4011
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.89660159498333e-06,
      "loss": 0.4322,
      "step": 4012
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.8740921006049545e-06,
      "loss": 0.4165,
      "step": 4013
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.851633170707437e-06,
      "loss": 0.4119,
      "step": 4014
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.8292248172289056e-06,
      "loss": 0.4225,
      "step": 4015
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.8068670520805655e-06,
      "loss": 0.4335,
      "step": 4016
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.784559887146789e-06,
      "loss": 0.4209,
      "step": 4017
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.762303334285001e-06,
      "loss": 0.416,
      "step": 4018
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.740097405325761e-06,
      "loss": 0.4191,
      "step": 4019
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.717942112072693e-06,
      "loss": 0.4272,
      "step": 4020
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.695837466302499e-06,
      "loss": 0.4356,
      "step": 4021
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.673783479764993e-06,
      "loss": 0.4288,
      "step": 4022
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.651780164183039e-06,
      "loss": 0.4273,
      "step": 4023
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.629827531252573e-06,
      "loss": 0.44,
      "step": 4024
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.607925592642592e-06,
      "loss": 0.4197,
      "step": 4025
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.586074359995119e-06,
      "loss": 0.4226,
      "step": 4026
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.56427384492526e-06,
      "loss": 0.4256,
      "step": 4027
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.542524059021147e-06,
      "loss": 0.4361,
      "step": 4028
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.520825013843944e-06,
      "loss": 0.4324,
      "step": 4029
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.4991767209278665e-06,
      "loss": 0.4201,
      "step": 4030
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.477579191780113e-06,
      "loss": 0.4196,
      "step": 4031
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.45603243788093e-06,
      "loss": 0.4147,
      "step": 4032
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.43453647068357e-06,
      "loss": 0.4215,
      "step": 4033
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.413091301614258e-06,
      "loss": 0.4274,
      "step": 4034
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.391696942072265e-06,
      "loss": 0.4286,
      "step": 4035
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.370353403429828e-06,
      "loss": 0.4154,
      "step": 4036
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.34906069703217e-06,
      "loss": 0.4147,
      "step": 4037
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.327818834197484e-06,
      "loss": 0.4161,
      "step": 4038
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.306627826216969e-06,
      "loss": 0.4189,
      "step": 4039
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.285487684354772e-06,
      "loss": 0.4152,
      "step": 4040
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.264398419848003e-06,
      "loss": 0.4129,
      "step": 4041
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.243360043906719e-06,
      "loss": 0.4317,
      "step": 4042
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.222372567713939e-06,
      "loss": 0.4279,
      "step": 4043
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.201436002425618e-06,
      "loss": 0.4226,
      "step": 4044
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.180550359170677e-06,
      "loss": 0.4246,
      "step": 4045
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.159715649050922e-06,
      "loss": 0.4266,
      "step": 4046
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.138931883141106e-06,
      "loss": 0.4301,
      "step": 4047
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.118199072488927e-06,
      "loss": 0.4347,
      "step": 4048
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.0975172281149775e-06,
      "loss": 0.4358,
      "step": 4049
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.076886361012744e-06,
      "loss": 0.4229,
      "step": 4050
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.056306482148631e-06,
      "loss": 0.4227,
      "step": 4051
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.035777602461943e-06,
      "loss": 0.4237,
      "step": 4052
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.015299732864886e-06,
      "loss": 0.4345,
      "step": 4053
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.994872884242506e-06,
      "loss": 0.4216,
      "step": 4054
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.974497067452787e-06,
      "loss": 0.4266,
      "step": 4055
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.95417229332653e-06,
      "loss": 0.4131,
      "step": 4056
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.933898572667471e-06,
      "loss": 0.4197,
      "step": 4057
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.913675916252157e-06,
      "loss": 0.4256,
      "step": 4058
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.89350433482999e-06,
      "loss": 0.4249,
      "step": 4059
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.873383839123257e-06,
      "loss": 0.4204,
      "step": 4060
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.853314439827071e-06,
      "loss": 0.4276,
      "step": 4061
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.833296147609366e-06,
      "loss": 0.4263,
      "step": 4062
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.8133289731109524e-06,
      "loss": 0.425,
      "step": 4063
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.7934129269454277e-06,
      "loss": 0.4298,
      "step": 4064
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.7735480196992466e-06,
      "loss": 0.4202,
      "step": 4065
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.753734261931663e-06,
      "loss": 0.4186,
      "step": 4066
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.733971664174729e-06,
      "loss": 0.4217,
      "step": 4067
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.71426023693332e-06,
      "loss": 0.4169,
      "step": 4068
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.6945999906851102e-06,
      "loss": 0.4186,
      "step": 4069
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.6749909358805736e-06,
      "loss": 0.4314,
      "step": 4070
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.655433082942972e-06,
      "loss": 0.4207,
      "step": 4071
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.6359264422683224e-06,
      "loss": 0.4246,
      "step": 4072
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.6164710242254653e-06,
      "loss": 0.421,
      "step": 4073
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5970668391559713e-06,
      "loss": 0.4374,
      "step": 4074
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5777138973741908e-06,
      "loss": 0.4192,
      "step": 4075
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5584122091672723e-06,
      "loss": 0.4259,
      "step": 4076
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5391617847950753e-06,
      "loss": 0.4233,
      "step": 4077
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5199626344902148e-06,
      "loss": 0.4234,
      "step": 4078
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.500814768458072e-06,
      "loss": 0.4153,
      "step": 4079
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.48171819687676e-06,
      "loss": 0.4189,
      "step": 4080
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.462672929897104e-06,
      "loss": 0.4171,
      "step": 4081
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.4436789776426948e-06,
      "loss": 0.42,
      "step": 4082
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.4247363502098227e-06,
      "loss": 0.4255,
      "step": 4083
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.4058450576674894e-06,
      "loss": 0.4355,
      "step": 4084
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.3870051100574395e-06,
      "loss": 0.4172,
      "step": 4085
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.368216517394107e-06,
      "loss": 0.4353,
      "step": 4086
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.3494792896646143e-06,
      "loss": 0.4259,
      "step": 4087
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.330793436828816e-06,
      "loss": 0.4181,
      "step": 4088
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.312158968819212e-06,
      "loss": 0.424,
      "step": 4089
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2935758955410347e-06,
      "loss": 0.426,
      "step": 4090
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2750442268721705e-06,
      "loss": 0.4232,
      "step": 4091
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2565639726631735e-06,
      "loss": 0.4156,
      "step": 4092
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2381351427372975e-06,
      "loss": 0.4348,
      "step": 4093
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2197577468904395e-06,
      "loss": 0.4229,
      "step": 4094
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.2014317948911632e-06,
      "loss": 0.4276,
      "step": 4095
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.1831572964806988e-06,
      "loss": 0.42,
      "step": 4096
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.1649342613728983e-06,
      "loss": 0.4228,
      "step": 4097
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.1467626992542797e-06,
      "loss": 0.4289,
      "step": 4098
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.1286426197839948e-06,
      "loss": 0.4225,
      "step": 4099
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.110574032593827e-06,
      "loss": 0.4354,
      "step": 4100
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0925569472882055e-06,
      "loss": 0.4191,
      "step": 4101
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0745913734441355e-06,
      "loss": 0.4356,
      "step": 4102
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.056677320611323e-06,
      "loss": 0.4234,
      "step": 4103
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0388147983120064e-06,
      "loss": 0.4289,
      "step": 4104
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0210038160410902e-06,
      "loss": 0.4289,
      "step": 4105
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.003244383266035e-06,
      "loss": 0.4242,
      "step": 4106
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.985536509426934e-06,
      "loss": 0.4271,
      "step": 4107
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.9678802039364685e-06,
      "loss": 0.4295,
      "step": 4108
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.950275476179898e-06,
      "loss": 0.4216,
      "step": 4109
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.9327223355150703e-06,
      "loss": 0.4276,
      "step": 4110
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.9152207912724217e-06,
      "loss": 0.431,
      "step": 4111
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.897770852754944e-06,
      "loss": 0.426,
      "step": 4112
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.880372529238229e-06,
      "loss": 0.4185,
      "step": 4113
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.8630258299703782e-06,
      "loss": 0.4206,
      "step": 4114
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.8457307641721166e-06,
      "loss": 0.4323,
      "step": 4115
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.8284873410366785e-06,
      "loss": 0.4052,
      "step": 4116
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.811295569729844e-06,
      "loss": 0.4203,
      "step": 4117
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.79415545938998e-06,
      "loss": 0.4258,
      "step": 4118
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.777067019127966e-06,
      "loss": 0.4227,
      "step": 4119
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.760030258027202e-06,
      "loss": 0.4127,
      "step": 4120
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7430451851436447e-06,
      "loss": 0.4174,
      "step": 4121
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.726111809505738e-06,
      "loss": 0.4278,
      "step": 4122
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7092301401145158e-06,
      "loss": 0.4268,
      "step": 4123
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.6924001859434553e-06,
      "loss": 0.4224,
      "step": 4124
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.6756219559385785e-06,
      "loss": 0.4312,
      "step": 4125
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.6588954590183957e-06,
      "loss": 0.4249,
      "step": 4126
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.6422207040739612e-06,
      "loss": 0.4367,
      "step": 4127
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.6255976999687627e-06,
      "loss": 0.4251,
      "step": 4128
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.60902645553881e-06,
      "loss": 0.4171,
      "step": 4129
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.592506979592624e-06,
      "loss": 0.4244,
      "step": 4130
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.5760392809111798e-06,
      "loss": 0.4294,
      "step": 4131
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.559623368247921e-06,
      "loss": 0.4207,
      "step": 4132
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.5432592503288e-06,
      "loss": 0.4213,
      "step": 4133
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.5269469358521927e-06,
      "loss": 0.4437,
      "step": 4134
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.5106864334889847e-06,
      "loss": 0.4261,
      "step": 4135
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.4944777518824735e-06,
      "loss": 0.4312,
      "step": 4136
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.4783208996484564e-06,
      "loss": 0.4164,
      "step": 4137
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.462215885375141e-06,
      "loss": 0.4267,
      "step": 4138
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.4461627176232237e-06,
      "loss": 0.426,
      "step": 4139
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.4301614049258013e-06,
      "loss": 0.4254,
      "step": 4140
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.414211955788415e-06,
      "loss": 0.4268,
      "step": 4141
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.3983143786890704e-06,
      "loss": 0.4289,
      "step": 4142
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.3824686820781538e-06,
      "loss": 0.4162,
      "step": 4143
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.366674874378505e-06,
      "loss": 0.4128,
      "step": 4144
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.350932963985375e-06,
      "loss": 0.4244,
      "step": 4145
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.335242959266415e-06,
      "loss": 0.4273,
      "step": 4146
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.3196048685616977e-06,
      "loss": 0.4316,
      "step": 4147
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.30401870018373e-06,
      "loss": 0.4209,
      "step": 4148
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.288484462417351e-06,
      "loss": 0.4352,
      "step": 4149
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.273002163519855e-06,
      "loss": 0.4352,
      "step": 4150
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2575718117209044e-06,
      "loss": 0.4367,
      "step": 4151
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.24219341522256e-06,
      "loss": 0.4235,
      "step": 4152
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2268669821992383e-06,
      "loss": 0.4388,
      "step": 4153
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.211592520797767e-06,
      "loss": 0.4182,
      "step": 4154
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.19637003913733e-06,
      "loss": 0.4234,
      "step": 4155
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1811995453094867e-06,
      "loss": 0.4225,
      "step": 4156
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1660810473781543e-06,
      "loss": 0.4189,
      "step": 4157
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.151014553379649e-06,
      "loss": 0.4216,
      "step": 4158
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1360000713225746e-06,
      "loss": 0.4264,
      "step": 4159
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.121037609187959e-06,
      "loss": 0.4195,
      "step": 4160
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.106127174929129e-06,
      "loss": 0.4318,
      "step": 4161
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0912687764717775e-06,
      "loss": 0.4235,
      "step": 4162
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0764624217139318e-06,
      "loss": 0.4304,
      "step": 4163
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0617081185259512e-06,
      "loss": 0.4219,
      "step": 4164
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0470058747505516e-06,
      "loss": 0.4276,
      "step": 4165
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.032355698202748e-06,
      "loss": 0.4292,
      "step": 4166
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.017757596669889e-06,
      "loss": 0.4187,
      "step": 4167
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0032115779116346e-06,
      "loss": 0.4298,
      "step": 4168
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.988717649659966e-06,
      "loss": 0.4247,
      "step": 4169
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.974275819619198e-06,
      "loss": 0.4247,
      "step": 4170
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9598860954659237e-06,
      "loss": 0.4171,
      "step": 4171
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9455484848490357e-06,
      "loss": 0.4284,
      "step": 4172
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.931262995389738e-06,
      "loss": 0.4193,
      "step": 4173
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.917029634681544e-06,
      "loss": 0.4258,
      "step": 4174
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9028484102902145e-06,
      "loss": 0.4356,
      "step": 4175
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8887193297538518e-06,
      "loss": 0.4288,
      "step": 4176
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8746424005828045e-06,
      "loss": 0.4267,
      "step": 4177
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.860617630259709e-06,
      "loss": 0.4217,
      "step": 4178
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8466450262394796e-06,
      "loss": 0.4238,
      "step": 4179
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8327245959492978e-06,
      "loss": 0.4247,
      "step": 4180
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8188563467885999e-06,
      "loss": 0.4141,
      "step": 4181
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8050402861291227e-06,
      "loss": 0.4317,
      "step": 4182
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.7912764213148136e-06,
      "loss": 0.4279,
      "step": 4183
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7775647596619093e-06,
      "loss": 0.4175,
      "step": 4184
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7639053084588908e-06,
      "loss": 0.4274,
      "step": 4185
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7502980749664722e-06,
      "loss": 0.4228,
      "step": 4186
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7367430664176455e-06,
      "loss": 0.4313,
      "step": 4187
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7232402900175915e-06,
      "loss": 0.4281,
      "step": 4188
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.709789752943769e-06,
      "loss": 0.4218,
      "step": 4189
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6963914623458477e-06,
      "loss": 0.4264,
      "step": 4190
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6830454253457416e-06,
      "loss": 0.4286,
      "step": 4191
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6697516490375654e-06,
      "loss": 0.4307,
      "step": 4192
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.656510140487666e-06,
      "loss": 0.4197,
      "step": 4193
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6433209067346355e-06,
      "loss": 0.4283,
      "step": 4194
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6301839547892328e-06,
      "loss": 0.4237,
      "step": 4195
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6170992916344497e-06,
      "loss": 0.424,
      "step": 4196
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6040669242254892e-06,
      "loss": 0.423,
      "step": 4197
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.5910868594897322e-06,
      "loss": 0.4335,
      "step": 4198
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.578159104326793e-06,
      "loss": 0.4182,
      "step": 4199
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5652836656084413e-06,
      "loss": 0.4172,
      "step": 4200
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5524605501786693e-06,
      "loss": 0.4292,
      "step": 4201
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5396897648536358e-06,
      "loss": 0.4226,
      "step": 4202
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5269713164216992e-06,
      "loss": 0.423,
      "step": 4203
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5143052116434075e-06,
      "loss": 0.4193,
      "step": 4204
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5016914572514417e-06,
      "loss": 0.4167,
      "step": 4205
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4891300599507053e-06,
      "loss": 0.4349,
      "step": 4206
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4766210264182234e-06,
      "loss": 0.4297,
      "step": 4207
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4641643633032443e-06,
      "loss": 0.4216,
      "step": 4208
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4517600772271267e-06,
      "loss": 0.4256,
      "step": 4209
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4394081747834076e-06,
      "loss": 0.4214,
      "step": 4210
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4271086625377905e-06,
      "loss": 0.4153,
      "step": 4211
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4148615470281123e-06,
      "loss": 0.4247,
      "step": 4212
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4026668347643768e-06,
      "loss": 0.4271,
      "step": 4213
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3905245322287096e-06,
      "loss": 0.4273,
      "step": 4214
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.378434645875404e-06,
      "loss": 0.4274,
      "step": 4215
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3663971821308629e-06,
      "loss": 0.4181,
      "step": 4216
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3544121473936467e-06,
      "loss": 0.426,
      "step": 4217
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3424795480344365e-06,
      "loss": 0.4168,
      "step": 4218
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.330599390396059e-06,
      "loss": 0.4227,
      "step": 4219
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3187716807934293e-06,
      "loss": 0.4311,
      "step": 4220
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3069964255136069e-06,
      "loss": 0.4219,
      "step": 4221
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2952736308157632e-06,
      "loss": 0.4263,
      "step": 4222
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2836033029312134e-06,
      "loss": 0.4244,
      "step": 4223
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2719854480633286e-06,
      "loss": 0.438,
      "step": 4224
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2604200723876357e-06,
      "loss": 0.4281,
      "step": 4225
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2489071820517396e-06,
      "loss": 0.4227,
      "step": 4226
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2374467831753555e-06,
      "loss": 0.4135,
      "step": 4227
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2260388818502888e-06,
      "loss": 0.4189,
      "step": 4228
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.214683484140433e-06,
      "loss": 0.4284,
      "step": 4229
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.2033805960818267e-06,
      "loss": 0.4315,
      "step": 4230
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1921302236825193e-06,
      "loss": 0.4358,
      "step": 4231
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.180932372922705e-06,
      "loss": 0.4372,
      "step": 4232
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1697870497546337e-06,
      "loss": 0.4307,
      "step": 4233
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1586942601026218e-06,
      "loss": 0.409,
      "step": 4234
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.147654009863086e-06,
      "loss": 0.4324,
      "step": 4235
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1366663049045213e-06,
      "loss": 0.4207,
      "step": 4236
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.125731151067455e-06,
      "loss": 0.4167,
      "step": 4237
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1148485541645049e-06,
      "loss": 0.4216,
      "step": 4238
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1040185199803653e-06,
      "loss": 0.4275,
      "step": 4239
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0932410542717653e-06,
      "loss": 0.4181,
      "step": 4240
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0825161627675107e-06,
      "loss": 0.4325,
      "step": 4241
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0718438511684414e-06,
      "loss": 0.4242,
      "step": 4242
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0612241251474643e-06,
      "loss": 0.4315,
      "step": 4243
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0506569903495188e-06,
      "loss": 0.4138,
      "step": 4244
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0401424523916126e-06,
      "loss": 0.4255,
      "step": 4245
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0296805168627853e-06,
      "loss": 0.4249,
      "step": 4246
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0192711893240891e-06,
      "loss": 0.4252,
      "step": 4247
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0089144753086532e-06,
      "loss": 0.4271,
      "step": 4248
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.986103803216297e-07,
      "loss": 0.4291,
      "step": 4249
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.883589098401701e-07,
      "loss": 0.4228,
      "step": 4250
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.78160069313494e-07,
      "loss": 0.4414,
      "step": 4251
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.6801386416282e-07,
      "loss": 0.4232,
      "step": 4252
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.579202997814118e-07,
      "loss": 0.4209,
      "step": 4253
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.478793815345111e-07,
      "loss": 0.4246,
      "step": 4254
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.378911147594038e-07,
      "loss": 0.4205,
      "step": 4255
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.279555047654098e-07,
      "loss": 0.4235,
      "step": 4256
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.180725568338044e-07,
      "loss": 0.4285,
      "step": 4257
    },
    {
      "epoch": 2.88,
      "learning_rate": 9.082422762179188e-07,
      "loss": 0.4222,
      "step": 4258
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.98464668143062e-07,
      "loss": 0.4193,
      "step": 4259
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.887397378065543e-07,
      "loss": 0.4234,
      "step": 4260
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.790674903777163e-07,
      "loss": 0.4285,
      "step": 4261
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.694479309978576e-07,
      "loss": 0.4166,
      "step": 4262
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.598810647802991e-07,
      "loss": 0.4255,
      "step": 4263
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.503668968103174e-07,
      "loss": 0.4266,
      "step": 4264
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.409054321452115e-07,
      "loss": 0.4233,
      "step": 4265
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.314966758142584e-07,
      "loss": 0.4277,
      "step": 4266
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.221406328187021e-07,
      "loss": 0.4135,
      "step": 4267
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.128373081317864e-07,
      "loss": 0.428,
      "step": 4268
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.035867066987002e-07,
      "loss": 0.4248,
      "step": 4269
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.943888334366656e-07,
      "loss": 0.4369,
      "step": 4270
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.852436932348162e-07,
      "loss": 0.4219,
      "step": 4271
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.761512909542967e-07,
      "loss": 0.4219,
      "step": 4272
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.671116314281967e-07,
      "loss": 0.4193,
      "step": 4273
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.581247194615615e-07,
      "loss": 0.4202,
      "step": 4274
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.491905598314363e-07,
      "loss": 0.4273,
      "step": 4275
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.403091572867781e-07,
      "loss": 0.4198,
      "step": 4276
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.31480516548555e-07,
      "loss": 0.4221,
      "step": 4277
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.227046423096351e-07,
      "loss": 0.4356,
      "step": 4278
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.139815392348648e-07,
      "loss": 0.4262,
      "step": 4279
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.053112119610461e-07,
      "loss": 0.4236,
      "step": 4280
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.966936650969258e-07,
      "loss": 0.4208,
      "step": 4281
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.881289032231619e-07,
      "loss": 0.4214,
      "step": 4282
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.796169308924128e-07,
      "loss": 0.42,
      "step": 4283
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.711577526292146e-07,
      "loss": 0.4209,
      "step": 4284
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.627513729300927e-07,
      "loss": 0.43,
      "step": 4285
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.543977962634729e-07,
      "loss": 0.4206,
      "step": 4286
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.460970270697364e-07,
      "loss": 0.4265,
      "step": 4287
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.378490697611761e-07,
      "loss": 0.4074,
      "step": 4288
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.296539287220182e-07,
      "loss": 0.4223,
      "step": 4289
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.215116083084227e-07,
      "loss": 0.4306,
      "step": 4290
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.134221128484607e-07,
      "loss": 0.4238,
      "step": 4291
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.053854466421372e-07,
      "loss": 0.4379,
      "step": 4292
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.97401613961357e-07,
      "loss": 0.4247,
      "step": 4293
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.894706190499588e-07,
      "loss": 0.4295,
      "step": 4294
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.815924661236816e-07,
      "loss": 0.4219,
      "step": 4295
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.737671593701977e-07,
      "loss": 0.4275,
      "step": 4296
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.659947029490575e-07,
      "loss": 0.4287,
      "step": 4297
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.582751009917453e-07,
      "loss": 0.4212,
      "step": 4298
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.506083576016341e-07,
      "loss": 0.4274,
      "step": 4299
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.429944768540085e-07,
      "loss": 0.4241,
      "step": 4300
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.354334627960534e-07,
      "loss": 0.4263,
      "step": 4301
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.279253194468425e-07,
      "loss": 0.424,
      "step": 4302
    },
    {
      "epoch": 2.91,
      "learning_rate": 5.204700507973615e-07,
      "loss": 0.4172,
      "step": 4303
    },
    {
      "epoch": 2.91,
      "learning_rate": 5.130676608104845e-07,
      "loss": 0.4309,
      "step": 4304
    },
    {
      "epoch": 2.91,
      "learning_rate": 5.057181534209643e-07,
      "loss": 0.4303,
      "step": 4305
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.984215325354646e-07,
      "loss": 0.4234,
      "step": 4306
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.911778020325275e-07,
      "loss": 0.4182,
      "step": 4307
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.839869657625728e-07,
      "loss": 0.4188,
      "step": 4308
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.768490275479209e-07,
      "loss": 0.4307,
      "step": 4309
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.6976399118277003e-07,
      "loss": 0.4289,
      "step": 4310
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.627318604331743e-07,
      "loss": 0.4256,
      "step": 4311
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.557526390370992e-07,
      "loss": 0.4303,
      "step": 4312
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.488263307043772e-07,
      "loss": 0.4237,
      "step": 4313
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.4195293911669655e-07,
      "loss": 0.4363,
      "step": 4314
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.3513246792763476e-07,
      "loss": 0.437,
      "step": 4315
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.2836492076264725e-07,
      "loss": 0.4157,
      "step": 4316
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.2165030121902313e-07,
      "loss": 0.4357,
      "step": 4317
    },
    {
      "epoch": 2.92,
      "learning_rate": 4.14988612865963e-07,
      "loss": 0.4206,
      "step": 4318
    },
    {
      "epoch": 2.92,
      "learning_rate": 4.0837985924448984e-07,
      "loss": 0.4187,
      "step": 4319
    },
    {
      "epoch": 2.92,
      "learning_rate": 4.0182404386752695e-07,
      "loss": 0.4162,
      "step": 4320
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.9532117021982006e-07,
      "loss": 0.4204,
      "step": 4321
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.888712417579932e-07,
      "loss": 0.427,
      "step": 4322
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.8247426191054814e-07,
      "loss": 0.4202,
      "step": 4323
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.7613023407779836e-07,
      "loss": 0.4292,
      "step": 4324
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.6983916163193524e-07,
      "loss": 0.4257,
      "step": 4325
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.636010479170171e-07,
      "loss": 0.4242,
      "step": 4326
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.574158962489027e-07,
      "loss": 0.4265,
      "step": 4327
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.51283709915351e-07,
      "loss": 0.4211,
      "step": 4328
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.452044921759434e-07,
      "loss": 0.4284,
      "step": 4329
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.3917824626209517e-07,
      "loss": 0.4275,
      "step": 4330
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.3320497537708826e-07,
      "loss": 0.4221,
      "step": 4331
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.272846826960385e-07,
      "loss": 0.4228,
      "step": 4332
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.214173713658841e-07,
      "loss": 0.4242,
      "step": 4333
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.1560304450541924e-07,
      "loss": 0.4198,
      "step": 4334
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.0984170520526045e-07,
      "loss": 0.4306,
      "step": 4335
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.041333565278803e-07,
      "loss": 0.4269,
      "step": 4336
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.984780015075517e-07,
      "loss": 0.4281,
      "step": 4337
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.928756431504143e-07,
      "loss": 0.4185,
      "step": 4338
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.8732628443440824e-07,
      "loss": 0.4294,
      "step": 4339
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.818299283093184e-07,
      "loss": 0.4283,
      "step": 4340
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.7638657769674116e-07,
      "loss": 0.4228,
      "step": 4341
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.7099623549012855e-07,
      "loss": 0.4195,
      "step": 4342
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.6565890455471087e-07,
      "loss": 0.4137,
      "step": 4343
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.6037458772757426e-07,
      "loss": 0.4302,
      "step": 4344
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.551432878176163e-07,
      "loss": 0.4312,
      "step": 4345
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.4996500760554597e-07,
      "loss": 0.4298,
      "step": 4346
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.448397498438948e-07,
      "loss": 0.4225,
      "step": 4347
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.3976751725700576e-07,
      "loss": 0.432,
      "step": 4348
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.3474831254105545e-07,
      "loss": 0.4229,
      "step": 4349
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.2978213836400975e-07,
      "loss": 0.422,
      "step": 4350
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.2486899736564593e-07,
      "loss": 0.4193,
      "step": 4351
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.2000889215757492e-07,
      "loss": 0.4182,
      "step": 4352
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.152018253231858e-07,
      "loss": 0.4296,
      "step": 4353
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.1044779941770121e-07,
      "loss": 0.4353,
      "step": 4354
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.0574681696814425e-07,
      "loss": 0.411,
      "step": 4355
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.0109888047331604e-07,
      "loss": 0.427,
      "step": 4356
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.9650399240386252e-07,
      "loss": 0.4287,
      "step": 4357
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.9196215520219663e-07,
      "loss": 0.4189,
      "step": 4358
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.8747337128256492e-07,
      "loss": 0.421,
      "step": 4359
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.8303764303096994e-07,
      "loss": 0.4216,
      "step": 4360
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.7865497280525888e-07,
      "loss": 0.4214,
      "step": 4361
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.7432536293504609e-07,
      "loss": 0.4254,
      "step": 4362
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.7004881572174614e-07,
      "loss": 0.4212,
      "step": 4363
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.6582533343857397e-07,
      "loss": 0.4259,
      "step": 4364
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.6165491833053382e-07,
      "loss": 0.4228,
      "step": 4365
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.5753757261443013e-07,
      "loss": 0.4343,
      "step": 4366
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.5347329847884563e-07,
      "loss": 0.4112,
      "step": 4367
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4946209808416324e-07,
      "loss": 0.4216,
      "step": 4368
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.45503973562533e-07,
      "loss": 0.4214,
      "step": 4369
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4159892701792742e-07,
      "loss": 0.4129,
      "step": 4370
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.3774696052607505e-07,
      "loss": 0.4165,
      "step": 4371
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.3394807613451575e-07,
      "loss": 0.429,
      "step": 4372
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.302022758625454e-07,
      "loss": 0.4289,
      "step": 4373
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.2650956170124906e-07,
      "loss": 0.4137,
      "step": 4374
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.228699356135232e-07,
      "loss": 0.425,
      "step": 4375
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.1928339953399815e-07,
      "loss": 0.4283,
      "step": 4376
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.1574995536912658e-07,
      "loss": 0.414,
      "step": 4377
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.1226960499712836e-07,
      "loss": 0.4283,
      "step": 4378
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0884235026796807e-07,
      "loss": 0.4125,
      "step": 4379
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0546819300343292e-07,
      "loss": 0.4205,
      "step": 4380
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0214713499706597e-07,
      "loss": 0.4301,
      "step": 4381
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.887917801417734e-08,
      "loss": 0.4313,
      "step": 4382
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.56643237918664e-08,
      "loss": 0.4242,
      "step": 4383
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.250257403899954e-08,
      "loss": 0.4242,
      "step": 4384
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.939393043622124e-08,
      "loss": 0.423,
      "step": 4385
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.633839463593196e-08,
      "loss": 0.4316,
      "step": 4386
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.333596826229917e-08,
      "loss": 0.4229,
      "step": 4387
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.038665291130177e-08,
      "loss": 0.4256,
      "step": 4388
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.749045015064127e-08,
      "loss": 0.4362,
      "step": 4389
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.464736151979734e-08,
      "loss": 0.4228,
      "step": 4390
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.185738853003887e-08,
      "loss": 0.4334,
      "step": 4391
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.912053266436846e-08,
      "loss": 0.423,
      "step": 4392
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.643679537757797e-08,
      "loss": 0.4218,
      "step": 4393
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.380617809621514e-08,
      "loss": 0.4183,
      "step": 4394
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.12286822186059e-08,
      "loss": 0.4246,
      "step": 4395
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.870430911480984e-08,
      "loss": 0.4241,
      "step": 4396
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.6233060126664735e-08,
      "loss": 0.4214,
      "step": 4397
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.381493656778647e-08,
      "loss": 0.4261,
      "step": 4398
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.144993972352463e-08,
      "loss": 0.4188,
      "step": 4399
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.9138070850995866e-08,
      "loss": 0.4305,
      "step": 4400
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.6879331179094934e-08,
      "loss": 0.4276,
      "step": 4401
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.467372190846142e-08,
      "loss": 0.4124,
      "step": 4402
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.252124421147974e-08,
      "loss": 0.4267,
      "step": 4403
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.0421899232323536e-08,
      "loss": 0.4327,
      "step": 4404
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.8375688086889075e-08,
      "loss": 0.4274,
      "step": 4405
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.638261186286185e-08,
      "loss": 0.4232,
      "step": 4406
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.4442671619661085e-08,
      "loss": 0.4277,
      "step": 4407
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.255586838847302e-08,
      "loss": 0.4163,
      "step": 4408
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.072220317221763e-08,
      "loss": 0.4299,
      "step": 4409
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.8941676945604125e-08,
      "loss": 0.4331,
      "step": 4410
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.7214290655075414e-08,
      "loss": 0.4278,
      "step": 4411
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.5540045218819253e-08,
      "loss": 0.4308,
      "step": 4412
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.3918941526790418e-08,
      "loss": 0.437,
      "step": 4413
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.2350980440688508e-08,
      "loss": 0.4215,
      "step": 4414
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.083616279398015e-08,
      "loss": 0.417,
      "step": 4415
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.93744893918546e-08,
      "loss": 0.4062,
      "step": 4416
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.796596101127923e-08,
      "loss": 0.4185,
      "step": 4417
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.6610578400966246e-08,
      "loss": 0.4223,
      "step": 4418
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.530834228137268e-08,
      "loss": 0.4235,
      "step": 4419
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.4059253344689271e-08,
      "loss": 0.424,
      "step": 4420
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.286331225489601e-08,
      "loss": 0.4219,
      "step": 4421
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.172051964767329e-08,
      "loss": 0.4128,
      "step": 4422
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.0630876130501843e-08,
      "loss": 0.4204,
      "step": 4423
    },
    {
      "epoch": 2.99,
      "learning_rate": 9.594382282585023e-09,
      "loss": 0.4282,
      "step": 4424
    },
    {
      "epoch": 2.99,
      "learning_rate": 8.611038654848802e-09,
      "loss": 0.425,
      "step": 4425
    },
    {
      "epoch": 2.99,
      "learning_rate": 7.680845770008383e-09,
      "loss": 0.4156,
      "step": 4426
    },
    {
      "epoch": 2.99,
      "learning_rate": 6.8038041225237936e-09,
      "loss": 0.4245,
      "step": 4427
    },
    {
      "epoch": 2.99,
      "learning_rate": 5.979914178566581e-09,
      "loss": 0.4199,
      "step": 4428
    },
    {
      "epoch": 2.99,
      "learning_rate": 5.209176376086422e-09,
      "loss": 0.4259,
      "step": 4429
    },
    {
      "epoch": 2.99,
      "learning_rate": 4.4915911247778165e-09,
      "loss": 0.4219,
      "step": 4430
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.827158806068986e-09,
      "loss": 0.4194,
      "step": 4431
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.215879773132979e-09,
      "loss": 0.4185,
      "step": 4432
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.6577543509209713e-09,
      "loss": 0.4227,
      "step": 4433
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.152782836084555e-09,
      "loss": 0.4368,
      "step": 4434
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.7009654970534528e-09,
      "loss": 0.4215,
      "step": 4435
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.3023025739800077e-09,
      "loss": 0.4122,
      "step": 4436
    },
    {
      "epoch": 3.0,
      "learning_rate": 9.567942787946926e-10,
      "loss": 0.4246,
      "step": 4437
    },
    {
      "epoch": 3.0,
      "learning_rate": 6.644407951394982e-10,
      "loss": 0.4138,
      "step": 4438
    },
    {
      "epoch": 3.0,
      "learning_rate": 4.252422784123411e-10,
      "loss": 0.4132,
      "step": 4439
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.3919885577816657e-10,
      "loss": 0.4211,
      "step": 4440
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.0631062611343723e-10,
      "loss": 0.4465,
      "step": 4441
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.6577660061644084e-11,
      "loss": 0.4218,
      "step": 4442
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0,
      "loss": 0.3894,
      "step": 4443
    },
    {
      "epoch": 3.0,
      "step": 4443,
      "total_flos": 4.0785724049138385e+19,
      "train_loss": 0.5103372382857417,
      "train_runtime": 42339.0568,
      "train_samples_per_second": 6.712,
      "train_steps_per_second": 0.105
    }
  ],
  "max_steps": 4443,
  "num_train_epochs": 3,
  "total_flos": 4.0785724049138385e+19,
  "trial_name": null,
  "trial_params": null
}