{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 5352,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 0.0,
      "loss": 9.646,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0,
      "loss": 9.7719,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0,
      "loss": 9.7988,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0,
      "loss": 9.6969,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.000000000000001e-08,
      "loss": 9.4957,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.2000000000000002e-07,
      "loss": 9.5051,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.8e-07,
      "loss": 9.7647,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.4000000000000003e-07,
      "loss": 9.7394,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.0000000000000004e-07,
      "loss": 10.1559,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.0000000000000004e-07,
      "loss": 9.3623,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.6e-07,
      "loss": 9.2681,
      "step": 11
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.2e-07,
      "loss": 9.475,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.2e-07,
      "loss": 9.7685,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.800000000000001e-07,
      "loss": 10.4204,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.4e-07,
      "loss": 9.1608,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.000000000000001e-07,
      "loss": 9.1745,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.599999999999999e-07,
      "loss": 9.2533,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.2e-07,
      "loss": 9.4772,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.799999999999999e-07,
      "loss": 9.6937,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.4e-07,
      "loss": 9.5748,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 9e-07,
      "loss": 9.1705,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.600000000000001e-07,
      "loss": 9.665,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.0200000000000002e-06,
      "loss": 9.9372,
      "step": 23
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.08e-06,
      "loss": 8.8675,
      "step": 24
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.14e-06,
      "loss": 9.0426,
      "step": 25
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2000000000000002e-06,
      "loss": 9.063,
      "step": 26
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.26e-06,
      "loss": 9.3062,
      "step": 27
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3199999999999999e-06,
      "loss": 8.9558,
      "step": 28
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.38e-06,
      "loss": 9.1951,
      "step": 29
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.44e-06,
      "loss": 8.5517,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5e-06,
      "loss": 8.7724,
      "step": 31
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5599999999999999e-06,
      "loss": 9.0144,
      "step": 32
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.62e-06,
      "loss": 8.7829,
      "step": 33
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.68e-06,
      "loss": 8.4442,
      "step": 34
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.74e-06,
      "loss": 9.2078,
      "step": 35
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.8e-06,
      "loss": 8.8228,
      "step": 36
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.86e-06,
      "loss": 8.5485,
      "step": 37
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9200000000000003e-06,
      "loss": 8.9114,
      "step": 38
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.98e-06,
      "loss": 8.6764,
      "step": 39
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.0400000000000004e-06,
      "loss": 8.2901,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.1000000000000002e-06,
      "loss": 8.5247,
      "step": 41
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.16e-06,
      "loss": 8.3651,
      "step": 42
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.22e-06,
      "loss": 8.3897,
      "step": 43
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.28e-06,
      "loss": 8.4195,
      "step": 44
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.34e-06,
      "loss": 8.7698,
      "step": 45
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.4000000000000003e-06,
      "loss": 8.3036,
      "step": 46
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.46e-06,
      "loss": 8.4639,
      "step": 47
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.52e-06,
      "loss": 8.727,
      "step": 48
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.58e-06,
      "loss": 8.4404,
      "step": 49
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.6399999999999997e-06,
      "loss": 8.586,
      "step": 50
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.7e-06,
      "loss": 7.952,
      "step": 51
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.76e-06,
      "loss": 7.9831,
      "step": 52
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.82e-06,
      "loss": 7.8591,
      "step": 53
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.88e-06,
      "loss": 7.7584,
      "step": 54
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.9400000000000002e-06,
      "loss": 7.9994,
      "step": 55
    },
    {
      "epoch": 0.03,
      "learning_rate": 3e-06,
      "loss": 7.721,
      "step": 56
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.06e-06,
      "loss": 7.6421,
      "step": 57
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.1199999999999998e-06,
      "loss": 7.6342,
      "step": 58
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.18e-06,
      "loss": 7.5409,
      "step": 59
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.24e-06,
      "loss": 7.675,
      "step": 60
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.3e-06,
      "loss": 7.7044,
      "step": 61
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.36e-06,
      "loss": 7.6713,
      "step": 62
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.4200000000000003e-06,
      "loss": 7.7013,
      "step": 63
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.48e-06,
      "loss": 7.5207,
      "step": 64
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.54e-06,
      "loss": 7.5308,
      "step": 65
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.6e-06,
      "loss": 7.4321,
      "step": 66
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.66e-06,
      "loss": 7.549,
      "step": 67
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.72e-06,
      "loss": 7.5311,
      "step": 68
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.7800000000000002e-06,
      "loss": 7.4505,
      "step": 69
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.8400000000000005e-06,
      "loss": 7.4192,
      "step": 70
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.9e-06,
      "loss": 7.4737,
      "step": 71
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.96e-06,
      "loss": 7.3745,
      "step": 72
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.0200000000000005e-06,
      "loss": 7.4593,
      "step": 73
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.080000000000001e-06,
      "loss": 7.3116,
      "step": 74
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.14e-06,
      "loss": 7.5823,
      "step": 75
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.2000000000000004e-06,
      "loss": 7.4017,
      "step": 76
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.26e-06,
      "loss": 7.4774,
      "step": 77
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.32e-06,
      "loss": 7.2905,
      "step": 78
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.3799999999999996e-06,
      "loss": 7.3126,
      "step": 79
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.44e-06,
      "loss": 7.3237,
      "step": 80
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.5e-06,
      "loss": 7.4401,
      "step": 81
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.56e-06,
      "loss": 7.564,
      "step": 82
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.62e-06,
      "loss": 7.2217,
      "step": 83
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.68e-06,
      "loss": 7.2945,
      "step": 84
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.74e-06,
      "loss": 7.342,
      "step": 85
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.800000000000001e-06,
      "loss": 7.3562,
      "step": 86
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.86e-06,
      "loss": 7.4173,
      "step": 87
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.92e-06,
      "loss": 7.3108,
      "step": 88
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.980000000000001e-06,
      "loss": 7.3819,
      "step": 89
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.04e-06,
      "loss": 7.1405,
      "step": 90
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.1e-06,
      "loss": 7.464,
      "step": 91
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.16e-06,
      "loss": 7.0621,
      "step": 92
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.22e-06,
      "loss": 7.2455,
      "step": 93
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.279999999999999e-06,
      "loss": 7.5043,
      "step": 94
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.34e-06,
      "loss": 7.4358,
      "step": 95
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.4e-06,
      "loss": 7.3645,
      "step": 96
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.46e-06,
      "loss": 7.2957,
      "step": 97
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.52e-06,
      "loss": 7.1162,
      "step": 98
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.58e-06,
      "loss": 7.4271,
      "step": 99
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.64e-06,
      "loss": 7.4022,
      "step": 100
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.7000000000000005e-06,
      "loss": 7.2946,
      "step": 101
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.76e-06,
      "loss": 7.243,
      "step": 102
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.82e-06,
      "loss": 7.2317,
      "step": 103
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.8800000000000005e-06,
      "loss": 7.1866,
      "step": 104
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.940000000000001e-06,
      "loss": 7.1308,
      "step": 105
    },
    {
      "epoch": 0.06,
      "learning_rate": 6e-06,
      "loss": 7.083,
      "step": 106
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.0600000000000004e-06,
      "loss": 7.5203,
      "step": 107
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.12e-06,
      "loss": 7.2079,
      "step": 108
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.18e-06,
      "loss": 7.055,
      "step": 109
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.2399999999999995e-06,
      "loss": 7.2252,
      "step": 110
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.3e-06,
      "loss": 7.0642,
      "step": 111
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.36e-06,
      "loss": 6.9843,
      "step": 112
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.42e-06,
      "loss": 7.1753,
      "step": 113
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.48e-06,
      "loss": 7.2549,
      "step": 114
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.54e-06,
      "loss": 7.1127,
      "step": 115
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.6e-06,
      "loss": 7.1639,
      "step": 116
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.660000000000001e-06,
      "loss": 7.2538,
      "step": 117
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.72e-06,
      "loss": 7.0184,
      "step": 118
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.78e-06,
      "loss": 7.1171,
      "step": 119
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.840000000000001e-06,
      "loss": 7.1383,
      "step": 120
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.900000000000001e-06,
      "loss": 7.3203,
      "step": 121
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.96e-06,
      "loss": 7.2452,
      "step": 122
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.0200000000000006e-06,
      "loss": 7.2143,
      "step": 123
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.08e-06,
      "loss": 7.2271,
      "step": 124
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.14e-06,
      "loss": 6.9322,
      "step": 125
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.2e-06,
      "loss": 7.4055,
      "step": 126
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.26e-06,
      "loss": 7.0597,
      "step": 127
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.32e-06,
      "loss": 7.1323,
      "step": 128
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.3800000000000005e-06,
      "loss": 6.9954,
      "step": 129
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.44e-06,
      "loss": 7.0255,
      "step": 130
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.5e-06,
      "loss": 7.3312,
      "step": 131
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.5600000000000005e-06,
      "loss": 7.0925,
      "step": 132
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.62e-06,
      "loss": 7.251,
      "step": 133
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.680000000000001e-06,
      "loss": 7.036,
      "step": 134
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.74e-06,
      "loss": 6.9941,
      "step": 135
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.8e-06,
      "loss": 7.2761,
      "step": 136
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.860000000000001e-06,
      "loss": 7.1411,
      "step": 137
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.92e-06,
      "loss": 7.0528,
      "step": 138
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.98e-06,
      "loss": 7.2624,
      "step": 139
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.040000000000001e-06,
      "loss": 7.1532,
      "step": 140
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.1e-06,
      "loss": 7.3406,
      "step": 141
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.160000000000001e-06,
      "loss": 7.2517,
      "step": 142
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.220000000000001e-06,
      "loss": 7.0044,
      "step": 143
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.28e-06,
      "loss": 7.1148,
      "step": 144
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.340000000000001e-06,
      "loss": 7.2629,
      "step": 145
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.400000000000001e-06,
      "loss": 7.1883,
      "step": 146
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.459999999999999e-06,
      "loss": 7.1461,
      "step": 147
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.52e-06,
      "loss": 7.122,
      "step": 148
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.58e-06,
      "loss": 7.0611,
      "step": 149
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.64e-06,
      "loss": 6.3844,
      "step": 150
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.7e-06,
      "loss": 7.0732,
      "step": 151
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.759999999999999e-06,
      "loss": 7.0725,
      "step": 152
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.82e-06,
      "loss": 6.9844,
      "step": 153
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.88e-06,
      "loss": 7.2963,
      "step": 154
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.939999999999999e-06,
      "loss": 6.9017,
      "step": 155
    },
    {
      "epoch": 0.09,
      "learning_rate": 9e-06,
      "loss": 7.2435,
      "step": 156
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.06e-06,
      "loss": 7.0353,
      "step": 157
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.12e-06,
      "loss": 7.06,
      "step": 158
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.18e-06,
      "loss": 7.0771,
      "step": 159
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.24e-06,
      "loss": 7.2243,
      "step": 160
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.3e-06,
      "loss": 7.1683,
      "step": 161
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.36e-06,
      "loss": 7.3092,
      "step": 162
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.42e-06,
      "loss": 7.0984,
      "step": 163
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.48e-06,
      "loss": 7.1257,
      "step": 164
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.54e-06,
      "loss": 7.1048,
      "step": 165
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.600000000000001e-06,
      "loss": 7.1907,
      "step": 166
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.66e-06,
      "loss": 7.151,
      "step": 167
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.72e-06,
      "loss": 6.9363,
      "step": 168
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.780000000000001e-06,
      "loss": 6.87,
      "step": 169
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.84e-06,
      "loss": 7.205,
      "step": 170
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.9e-06,
      "loss": 7.1592,
      "step": 171
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.960000000000001e-06,
      "loss": 7.0488,
      "step": 172
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.002e-05,
      "loss": 7.058,
      "step": 173
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.008e-05,
      "loss": 7.0036,
      "step": 174
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0140000000000001e-05,
      "loss": 7.171,
      "step": 175
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.02e-05,
      "loss": 6.921,
      "step": 176
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0260000000000002e-05,
      "loss": 6.9625,
      "step": 177
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.032e-05,
      "loss": 7.0435,
      "step": 178
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0379999999999999e-05,
      "loss": 6.9003,
      "step": 179
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.044e-05,
      "loss": 7.1111,
      "step": 180
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.05e-05,
      "loss": 6.9381,
      "step": 181
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0559999999999999e-05,
      "loss": 6.9341,
      "step": 182
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.062e-05,
      "loss": 7.1557,
      "step": 183
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.068e-05,
      "loss": 6.8587,
      "step": 184
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.074e-05,
      "loss": 7.0471,
      "step": 185
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.08e-05,
      "loss": 6.9722,
      "step": 186
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.086e-05,
      "loss": 6.8958,
      "step": 187
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.092e-05,
      "loss": 7.0297,
      "step": 188
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.098e-05,
      "loss": 6.9514,
      "step": 189
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.104e-05,
      "loss": 6.9733,
      "step": 190
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.11e-05,
      "loss": 7.1154,
      "step": 191
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.116e-05,
      "loss": 6.8412,
      "step": 192
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1220000000000001e-05,
      "loss": 6.8898,
      "step": 193
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.128e-05,
      "loss": 7.0222,
      "step": 194
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.134e-05,
      "loss": 7.0209,
      "step": 195
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1400000000000001e-05,
      "loss": 7.0365,
      "step": 196
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.146e-05,
      "loss": 6.9405,
      "step": 197
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.152e-05,
      "loss": 6.7527,
      "step": 198
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1580000000000001e-05,
      "loss": 6.7106,
      "step": 199
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.164e-05,
      "loss": 6.4889,
      "step": 200
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1700000000000001e-05,
      "loss": 7.2304,
      "step": 201
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1760000000000001e-05,
      "loss": 6.998,
      "step": 202
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.182e-05,
      "loss": 7.3014,
      "step": 203
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1880000000000001e-05,
      "loss": 6.9186,
      "step": 204
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1940000000000001e-05,
      "loss": 6.9083,
      "step": 205
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2e-05,
      "loss": 7.0765,
      "step": 206
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2060000000000001e-05,
      "loss": 7.053,
      "step": 207
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2120000000000001e-05,
      "loss": 6.9603,
      "step": 208
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2180000000000002e-05,
      "loss": 7.0472,
      "step": 209
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.224e-05,
      "loss": 7.1937,
      "step": 210
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2299999999999999e-05,
      "loss": 7.0072,
      "step": 211
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.236e-05,
      "loss": 7.1505,
      "step": 212
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.242e-05,
      "loss": 7.0382,
      "step": 213
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2479999999999999e-05,
      "loss": 6.852,
      "step": 214
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.254e-05,
      "loss": 7.1235,
      "step": 215
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.26e-05,
      "loss": 6.8629,
      "step": 216
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2659999999999999e-05,
      "loss": 7.0027,
      "step": 217
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.272e-05,
      "loss": 6.9188,
      "step": 218
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.278e-05,
      "loss": 6.9759,
      "step": 219
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.284e-05,
      "loss": 7.0635,
      "step": 220
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.29e-05,
      "loss": 7.1394,
      "step": 221
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.296e-05,
      "loss": 7.1949,
      "step": 222
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.302e-05,
      "loss": 6.8955,
      "step": 223
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.308e-05,
      "loss": 7.0308,
      "step": 224
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.314e-05,
      "loss": 7.0852,
      "step": 225
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.32e-05,
      "loss": 7.2942,
      "step": 226
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.326e-05,
      "loss": 7.2081,
      "step": 227
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3320000000000001e-05,
      "loss": 7.1173,
      "step": 228
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.338e-05,
      "loss": 6.9345,
      "step": 229
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.344e-05,
      "loss": 6.9924,
      "step": 230
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3500000000000001e-05,
      "loss": 6.9912,
      "step": 231
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.356e-05,
      "loss": 7.097,
      "step": 232
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.362e-05,
      "loss": 7.2127,
      "step": 233
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3680000000000001e-05,
      "loss": 7.0553,
      "step": 234
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.374e-05,
      "loss": 6.888,
      "step": 235
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3800000000000002e-05,
      "loss": 7.2717,
      "step": 236
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3860000000000001e-05,
      "loss": 7.0714,
      "step": 237
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.392e-05,
      "loss": 7.0867,
      "step": 238
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3980000000000002e-05,
      "loss": 6.9838,
      "step": 239
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.4040000000000001e-05,
      "loss": 7.0201,
      "step": 240
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4099999999999999e-05,
      "loss": 7.032,
      "step": 241
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.416e-05,
      "loss": 6.9762,
      "step": 242
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.422e-05,
      "loss": 6.8247,
      "step": 243
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.428e-05,
      "loss": 6.9969,
      "step": 244
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.434e-05,
      "loss": 6.9705,
      "step": 245
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.44e-05,
      "loss": 7.0128,
      "step": 246
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.446e-05,
      "loss": 6.5521,
      "step": 247
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.452e-05,
      "loss": 6.8975,
      "step": 248
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.458e-05,
      "loss": 6.6274,
      "step": 249
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.464e-05,
      "loss": 6.543,
      "step": 250
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.47e-05,
      "loss": 7.0635,
      "step": 251
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4760000000000001e-05,
      "loss": 7.3753,
      "step": 252
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.482e-05,
      "loss": 7.0203,
      "step": 253
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.488e-05,
      "loss": 7.1315,
      "step": 254
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4940000000000001e-05,
      "loss": 7.0209,
      "step": 255
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.5e-05,
      "loss": 7.025,
      "step": 256
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.506e-05,
      "loss": 7.0403,
      "step": 257
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.5120000000000001e-05,
      "loss": 6.9659,
      "step": 258
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.518e-05,
      "loss": 7.2237,
      "step": 259
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.524e-05,
      "loss": 6.742,
      "step": 260
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.53e-05,
      "loss": 6.8603,
      "step": 261
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5360000000000002e-05,
      "loss": 6.907,
      "step": 262
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.542e-05,
      "loss": 7.1284,
      "step": 263
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.548e-05,
      "loss": 7.0301,
      "step": 264
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.554e-05,
      "loss": 6.8978,
      "step": 265
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.56e-05,
      "loss": 6.7717,
      "step": 266
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5660000000000003e-05,
      "loss": 7.0333,
      "step": 267
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5720000000000002e-05,
      "loss": 6.9698,
      "step": 268
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.578e-05,
      "loss": 7.0326,
      "step": 269
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.584e-05,
      "loss": 6.8031,
      "step": 270
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.59e-05,
      "loss": 6.8608,
      "step": 271
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.596e-05,
      "loss": 6.8752,
      "step": 272
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.6020000000000002e-05,
      "loss": 6.7441,
      "step": 273
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.6080000000000002e-05,
      "loss": 6.8664,
      "step": 274
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.614e-05,
      "loss": 6.9774,
      "step": 275
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.62e-05,
      "loss": 6.9911,
      "step": 276
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.626e-05,
      "loss": 6.9141,
      "step": 277
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6320000000000003e-05,
      "loss": 7.0444,
      "step": 278
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6380000000000002e-05,
      "loss": 6.828,
      "step": 279
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6440000000000002e-05,
      "loss": 6.9379,
      "step": 280
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.65e-05,
      "loss": 6.8366,
      "step": 281
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.656e-05,
      "loss": 6.9912,
      "step": 282
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6620000000000004e-05,
      "loss": 6.9299,
      "step": 283
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6680000000000003e-05,
      "loss": 6.5692,
      "step": 284
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6740000000000002e-05,
      "loss": 6.9797,
      "step": 285
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6800000000000002e-05,
      "loss": 6.5681,
      "step": 286
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.686e-05,
      "loss": 6.7236,
      "step": 287
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6919999999999997e-05,
      "loss": 6.9405,
      "step": 288
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.698e-05,
      "loss": 6.5858,
      "step": 289
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.704e-05,
      "loss": 6.9379,
      "step": 290
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.71e-05,
      "loss": 6.8069,
      "step": 291
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.716e-05,
      "loss": 6.9986,
      "step": 292
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.7219999999999998e-05,
      "loss": 6.8355,
      "step": 293
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.728e-05,
      "loss": 6.8072,
      "step": 294
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.734e-05,
      "loss": 6.7032,
      "step": 295
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.74e-05,
      "loss": 6.5432,
      "step": 296
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.746e-05,
      "loss": 6.5114,
      "step": 297
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7519999999999998e-05,
      "loss": 6.8482,
      "step": 298
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.758e-05,
      "loss": 6.436,
      "step": 299
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.764e-05,
      "loss": 6.2729,
      "step": 300
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.77e-05,
      "loss": 6.6906,
      "step": 301
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.776e-05,
      "loss": 6.5058,
      "step": 302
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.782e-05,
      "loss": 6.585,
      "step": 303
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7879999999999998e-05,
      "loss": 6.4205,
      "step": 304
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.794e-05,
      "loss": 6.5869,
      "step": 305
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8e-05,
      "loss": 6.5081,
      "step": 306
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.806e-05,
      "loss": 6.6771,
      "step": 307
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.812e-05,
      "loss": 6.7533,
      "step": 308
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.818e-05,
      "loss": 6.5612,
      "step": 309
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.824e-05,
      "loss": 6.7425,
      "step": 310
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.83e-05,
      "loss": 6.518,
      "step": 311
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.836e-05,
      "loss": 6.6526,
      "step": 312
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.842e-05,
      "loss": 6.5643,
      "step": 313
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.848e-05,
      "loss": 6.5001,
      "step": 314
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.854e-05,
      "loss": 6.5695,
      "step": 315
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.86e-05,
      "loss": 6.4166,
      "step": 316
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.866e-05,
      "loss": 6.3553,
      "step": 317
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.872e-05,
      "loss": 6.3397,
      "step": 318
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.878e-05,
      "loss": 6.5771,
      "step": 319
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.884e-05,
      "loss": 6.4769,
      "step": 320
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8900000000000002e-05,
      "loss": 6.6914,
      "step": 321
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.896e-05,
      "loss": 6.6381,
      "step": 322
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.902e-05,
      "loss": 6.6176,
      "step": 323
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.908e-05,
      "loss": 6.4186,
      "step": 324
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.914e-05,
      "loss": 6.267,
      "step": 325
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9200000000000003e-05,
      "loss": 6.5408,
      "step": 326
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9260000000000002e-05,
      "loss": 6.3241,
      "step": 327
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.932e-05,
      "loss": 6.363,
      "step": 328
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.938e-05,
      "loss": 6.2959,
      "step": 329
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.944e-05,
      "loss": 6.5393,
      "step": 330
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.95e-05,
      "loss": 6.3424,
      "step": 331
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9560000000000002e-05,
      "loss": 6.472,
      "step": 332
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9620000000000002e-05,
      "loss": 6.5025,
      "step": 333
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.968e-05,
      "loss": 6.4152,
      "step": 334
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.974e-05,
      "loss": 6.302,
      "step": 335
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.98e-05,
      "loss": 6.2684,
      "step": 336
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9860000000000003e-05,
      "loss": 6.3401,
      "step": 337
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9920000000000002e-05,
      "loss": 6.0776,
      "step": 338
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9980000000000002e-05,
      "loss": 6.1896,
      "step": 339
    },
    {
      "epoch": 0.19,
      "learning_rate": 2.004e-05,
      "loss": 6.3508,
      "step": 340
    },
    {
      "epoch": 0.19,
      "learning_rate": 2.01e-05,
      "loss": 6.0735,
      "step": 341
    },
    {
      "epoch": 0.19,
      "learning_rate": 2.016e-05,
      "loss": 6.3727,
      "step": 342
    },
    {
      "epoch": 0.19,
      "learning_rate": 2.0220000000000003e-05,
      "loss": 6.3666,
      "step": 343
    },
    {
      "epoch": 0.19,
      "learning_rate": 2.0280000000000002e-05,
      "loss": 6.3386,
      "step": 344
    },
    {
      "epoch": 0.19,
      "learning_rate": 2.0340000000000002e-05,
      "loss": 6.0668,
      "step": 345
    },
    {
      "epoch": 0.19,
      "learning_rate": 2.04e-05,
      "loss": 6.2498,
      "step": 346
    },
    {
      "epoch": 0.19,
      "learning_rate": 2.046e-05,
      "loss": 6.7105,
      "step": 347
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.0520000000000003e-05,
      "loss": 6.4332,
      "step": 348
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.0580000000000003e-05,
      "loss": 5.7855,
      "step": 349
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.064e-05,
      "loss": 5.6302,
      "step": 350
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.07e-05,
      "loss": 6.3813,
      "step": 351
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.0759999999999998e-05,
      "loss": 6.3583,
      "step": 352
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.082e-05,
      "loss": 6.3118,
      "step": 353
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.088e-05,
      "loss": 6.2486,
      "step": 354
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.094e-05,
      "loss": 6.0309,
      "step": 355
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.1e-05,
      "loss": 6.275,
      "step": 356
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.1059999999999998e-05,
      "loss": 6.3483,
      "step": 357
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.1119999999999998e-05,
      "loss": 6.199,
      "step": 358
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.118e-05,
      "loss": 6.405,
      "step": 359
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.124e-05,
      "loss": 6.2315,
      "step": 360
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.13e-05,
      "loss": 6.0119,
      "step": 361
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.136e-05,
      "loss": 5.9327,
      "step": 362
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.1419999999999998e-05,
      "loss": 6.0674,
      "step": 363
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.148e-05,
      "loss": 6.0941,
      "step": 364
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.154e-05,
      "loss": 6.1867,
      "step": 365
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.16e-05,
      "loss": 6.2712,
      "step": 366
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.166e-05,
      "loss": 6.0772,
      "step": 367
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.172e-05,
      "loss": 6.1193,
      "step": 368
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.178e-05,
      "loss": 6.1658,
      "step": 369
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.184e-05,
      "loss": 5.9765,
      "step": 370
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.19e-05,
      "loss": 6.0289,
      "step": 371
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.196e-05,
      "loss": 6.1845,
      "step": 372
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.202e-05,
      "loss": 6.1622,
      "step": 373
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.208e-05,
      "loss": 5.9383,
      "step": 374
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.214e-05,
      "loss": 5.8226,
      "step": 375
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.22e-05,
      "loss": 6.0426,
      "step": 376
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.226e-05,
      "loss": 5.9302,
      "step": 377
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.232e-05,
      "loss": 5.8862,
      "step": 378
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.238e-05,
      "loss": 6.0889,
      "step": 379
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.2440000000000002e-05,
      "loss": 6.127,
      "step": 380
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.25e-05,
      "loss": 5.9815,
      "step": 381
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.256e-05,
      "loss": 5.8049,
      "step": 382
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.262e-05,
      "loss": 5.9601,
      "step": 383
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.268e-05,
      "loss": 5.9896,
      "step": 384
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.274e-05,
      "loss": 6.1103,
      "step": 385
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.2800000000000002e-05,
      "loss": 6.2715,
      "step": 386
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.286e-05,
      "loss": 6.1132,
      "step": 387
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.292e-05,
      "loss": 5.7827,
      "step": 388
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.298e-05,
      "loss": 5.7706,
      "step": 389
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.304e-05,
      "loss": 6.1264,
      "step": 390
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.3100000000000002e-05,
      "loss": 5.8546,
      "step": 391
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.3160000000000002e-05,
      "loss": 6.4934,
      "step": 392
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.322e-05,
      "loss": 6.1479,
      "step": 393
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.328e-05,
      "loss": 5.9958,
      "step": 394
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.334e-05,
      "loss": 5.9835,
      "step": 395
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.3400000000000003e-05,
      "loss": 5.7532,
      "step": 396
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.3460000000000002e-05,
      "loss": 5.8927,
      "step": 397
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.3520000000000002e-05,
      "loss": 6.1253,
      "step": 398
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.358e-05,
      "loss": 6.1784,
      "step": 399
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.364e-05,
      "loss": 6.0252,
      "step": 400
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.37e-05,
      "loss": 6.6853,
      "step": 401
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.3760000000000003e-05,
      "loss": 6.2031,
      "step": 402
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.3820000000000002e-05,
      "loss": 5.9138,
      "step": 403
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.3880000000000002e-05,
      "loss": 6.1915,
      "step": 404
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.394e-05,
      "loss": 5.8955,
      "step": 405
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.4e-05,
      "loss": 6.0869,
      "step": 406
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.4060000000000003e-05,
      "loss": 5.9179,
      "step": 407
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.4120000000000003e-05,
      "loss": 6.0656,
      "step": 408
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.4180000000000002e-05,
      "loss": 5.9288,
      "step": 409
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.4240000000000002e-05,
      "loss": 5.8978,
      "step": 410
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.43e-05,
      "loss": 6.0034,
      "step": 411
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.4360000000000004e-05,
      "loss": 5.7767,
      "step": 412
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.442e-05,
      "loss": 6.1283,
      "step": 413
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.448e-05,
      "loss": 5.8503,
      "step": 414
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.454e-05,
      "loss": 6.0095,
      "step": 415
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.4599999999999998e-05,
      "loss": 5.9851,
      "step": 416
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.4659999999999998e-05,
      "loss": 5.8617,
      "step": 417
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.472e-05,
      "loss": 5.9014,
      "step": 418
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.478e-05,
      "loss": 5.8312,
      "step": 419
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.484e-05,
      "loss": 6.0186,
      "step": 420
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.49e-05,
      "loss": 5.8552,
      "step": 421
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.4959999999999998e-05,
      "loss": 5.8374,
      "step": 422
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.502e-05,
      "loss": 5.7043,
      "step": 423
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.508e-05,
      "loss": 6.1189,
      "step": 424
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.514e-05,
      "loss": 6.0517,
      "step": 425
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.52e-05,
      "loss": 5.8418,
      "step": 426
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.526e-05,
      "loss": 5.7581,
      "step": 427
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.5319999999999998e-05,
      "loss": 5.9097,
      "step": 428
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.538e-05,
      "loss": 5.9328,
      "step": 429
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.544e-05,
      "loss": 5.812,
      "step": 430
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.55e-05,
      "loss": 5.7009,
      "step": 431
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.556e-05,
      "loss": 5.8661,
      "step": 432
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.562e-05,
      "loss": 5.6769,
      "step": 433
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.568e-05,
      "loss": 5.9462,
      "step": 434
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.574e-05,
      "loss": 5.8893,
      "step": 435
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.58e-05,
      "loss": 5.6401,
      "step": 436
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.586e-05,
      "loss": 5.8101,
      "step": 437
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.592e-05,
      "loss": 5.9193,
      "step": 438
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.5980000000000002e-05,
      "loss": 5.7365,
      "step": 439
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.604e-05,
      "loss": 5.8024,
      "step": 440
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.61e-05,
      "loss": 5.7961,
      "step": 441
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.616e-05,
      "loss": 5.8636,
      "step": 442
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.622e-05,
      "loss": 5.7268,
      "step": 443
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.628e-05,
      "loss": 5.662,
      "step": 444
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.6340000000000002e-05,
      "loss": 5.7899,
      "step": 445
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.64e-05,
      "loss": 6.2986,
      "step": 446
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.646e-05,
      "loss": 5.8928,
      "step": 447
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.652e-05,
      "loss": 6.0612,
      "step": 448
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.658e-05,
      "loss": 6.0148,
      "step": 449
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.6640000000000002e-05,
      "loss": 5.3174,
      "step": 450
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.6700000000000002e-05,
      "loss": 6.0291,
      "step": 451
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.676e-05,
      "loss": 5.883,
      "step": 452
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.682e-05,
      "loss": 5.9288,
      "step": 453
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.688e-05,
      "loss": 5.5972,
      "step": 454
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.6940000000000003e-05,
      "loss": 5.6117,
      "step": 455
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.7000000000000002e-05,
      "loss": 5.8273,
      "step": 456
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.7060000000000002e-05,
      "loss": 5.4896,
      "step": 457
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.712e-05,
      "loss": 5.6234,
      "step": 458
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.718e-05,
      "loss": 5.5151,
      "step": 459
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.724e-05,
      "loss": 5.9104,
      "step": 460
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.7300000000000003e-05,
      "loss": 5.56,
      "step": 461
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.7360000000000002e-05,
      "loss": 5.5886,
      "step": 462
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.7420000000000002e-05,
      "loss": 5.6127,
      "step": 463
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.748e-05,
      "loss": 5.7957,
      "step": 464
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.754e-05,
      "loss": 5.5925,
      "step": 465
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.7600000000000003e-05,
      "loss": 5.7286,
      "step": 466
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.7660000000000003e-05,
      "loss": 5.703,
      "step": 467
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.7720000000000002e-05,
      "loss": 5.6674,
      "step": 468
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.778e-05,
      "loss": 5.4795,
      "step": 469
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.784e-05,
      "loss": 5.6715,
      "step": 470
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.79e-05,
      "loss": 5.738,
      "step": 471
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.7960000000000003e-05,
      "loss": 5.5129,
      "step": 472
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.8020000000000003e-05,
      "loss": 5.6156,
      "step": 473
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.8080000000000002e-05,
      "loss": 5.5239,
      "step": 474
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.8139999999999998e-05,
      "loss": 5.5943,
      "step": 475
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.8199999999999998e-05,
      "loss": 5.6577,
      "step": 476
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.826e-05,
      "loss": 5.7205,
      "step": 477
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.832e-05,
      "loss": 5.5524,
      "step": 478
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.838e-05,
      "loss": 5.7494,
      "step": 479
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.844e-05,
      "loss": 5.7085,
      "step": 480
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.8499999999999998e-05,
      "loss": 5.6814,
      "step": 481
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.856e-05,
      "loss": 5.8119,
      "step": 482
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.862e-05,
      "loss": 5.9045,
      "step": 483
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.868e-05,
      "loss": 5.5896,
      "step": 484
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.874e-05,
      "loss": 5.6487,
      "step": 485
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.88e-05,
      "loss": 5.7455,
      "step": 486
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.8859999999999998e-05,
      "loss": 5.5234,
      "step": 487
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.892e-05,
      "loss": 5.8607,
      "step": 488
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.898e-05,
      "loss": 5.6522,
      "step": 489
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.904e-05,
      "loss": 5.6137,
      "step": 490
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.91e-05,
      "loss": 5.4497,
      "step": 491
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.916e-05,
      "loss": 5.6573,
      "step": 492
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.922e-05,
      "loss": 5.7641,
      "step": 493
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.928e-05,
      "loss": 5.7829,
      "step": 494
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.934e-05,
      "loss": 5.4454,
      "step": 495
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.94e-05,
      "loss": 5.7315,
      "step": 496
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.946e-05,
      "loss": 5.2833,
      "step": 497
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.9520000000000002e-05,
      "loss": 5.8994,
      "step": 498
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.958e-05,
      "loss": 5.4753,
      "step": 499
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.964e-05,
      "loss": 5.3825,
      "step": 500
    },
    {
      "epoch": 0.28,
      "eval_loss": 5.539823055267334,
      "eval_runtime": 535.5475,
      "eval_samples_per_second": 4.933,
      "eval_steps_per_second": 0.618,
      "eval_wer": 0.9601563585134035,
      "step": 500
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.97e-05,
      "loss": 5.9641,
      "step": 501
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.976e-05,
      "loss": 5.8708,
      "step": 502
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.982e-05,
      "loss": 5.6421,
      "step": 503
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.9880000000000002e-05,
      "loss": 5.6836,
      "step": 504
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.994e-05,
      "loss": 5.4648,
      "step": 505
    },
    {
      "epoch": 0.28,
      "learning_rate": 3e-05,
      "loss": 5.883,
      "step": 506
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.9993816982687552e-05,
      "loss": 5.5796,
      "step": 507
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.9987633965375106e-05,
      "loss": 5.5719,
      "step": 508
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9981450948062654e-05,
      "loss": 5.6836,
      "step": 509
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9975267930750208e-05,
      "loss": 5.7018,
      "step": 510
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.996908491343776e-05,
      "loss": 5.7354,
      "step": 511
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.996290189612531e-05,
      "loss": 5.5162,
      "step": 512
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.995671887881286e-05,
      "loss": 5.5504,
      "step": 513
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9950535861500412e-05,
      "loss": 5.4056,
      "step": 514
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9944352844187963e-05,
      "loss": 5.5184,
      "step": 515
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9938169826875517e-05,
      "loss": 5.741,
      "step": 516
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.993198680956307e-05,
      "loss": 5.5557,
      "step": 517
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.992580379225062e-05,
      "loss": 5.7338,
      "step": 518
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.991962077493817e-05,
      "loss": 5.5007,
      "step": 519
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.991343775762572e-05,
      "loss": 5.4319,
      "step": 520
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9907254740313276e-05,
      "loss": 5.282,
      "step": 521
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9901071723000827e-05,
      "loss": 5.4889,
      "step": 522
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9894888705688374e-05,
      "loss": 5.3306,
      "step": 523
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.988870568837593e-05,
      "loss": 5.6146,
      "step": 524
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.988252267106348e-05,
      "loss": 5.6235,
      "step": 525
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9876339653751034e-05,
      "loss": 5.5,
      "step": 526
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.987015663643858e-05,
      "loss": 5.6816,
      "step": 527
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9863973619126133e-05,
      "loss": 5.358,
      "step": 528
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9857790601813687e-05,
      "loss": 5.6113,
      "step": 529
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9851607584501238e-05,
      "loss": 5.6772,
      "step": 530
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.984542456718879e-05,
      "loss": 5.2315,
      "step": 531
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.983924154987634e-05,
      "loss": 5.3005,
      "step": 532
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.983305853256389e-05,
      "loss": 5.2843,
      "step": 533
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9826875515251445e-05,
      "loss": 5.3441,
      "step": 534
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9820692497938996e-05,
      "loss": 5.6496,
      "step": 535
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9814509480626544e-05,
      "loss": 5.4859,
      "step": 536
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9808326463314098e-05,
      "loss": 5.2861,
      "step": 537
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.980214344600165e-05,
      "loss": 5.4684,
      "step": 538
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9795960428689203e-05,
      "loss": 5.5343,
      "step": 539
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9789777411376754e-05,
      "loss": 5.5708,
      "step": 540
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9783594394064302e-05,
      "loss": 5.3512,
      "step": 541
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9777411376751856e-05,
      "loss": 5.6482,
      "step": 542
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9771228359439407e-05,
      "loss": 5.3037,
      "step": 543
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.976504534212696e-05,
      "loss": 5.4302,
      "step": 544
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.975886232481451e-05,
      "loss": 5.065,
      "step": 545
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.975267930750206e-05,
      "loss": 5.1237,
      "step": 546
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9746496290189615e-05,
      "loss": 5.7082,
      "step": 547
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9740313272877166e-05,
      "loss": 5.4364,
      "step": 548
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9734130255564717e-05,
      "loss": 5.5296,
      "step": 549
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9727947238252268e-05,
      "loss": 5.2537,
      "step": 550
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.972176422093982e-05,
      "loss": 6.0571,
      "step": 551
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.971558120362737e-05,
      "loss": 5.96,
      "step": 552
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9709398186314924e-05,
      "loss": 5.6144,
      "step": 553
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9703215169002475e-05,
      "loss": 5.5449,
      "step": 554
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9697032151690026e-05,
      "loss": 5.4027,
      "step": 555
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9690849134377577e-05,
      "loss": 5.5616,
      "step": 556
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9684666117065128e-05,
      "loss": 5.7222,
      "step": 557
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9678483099752682e-05,
      "loss": 5.4871,
      "step": 558
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.967230008244023e-05,
      "loss": 5.2447,
      "step": 559
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.966611706512778e-05,
      "loss": 5.5694,
      "step": 560
    },
    {
      "epoch": 0.31,
      "learning_rate": 2.9659934047815335e-05,
      "loss": 5.5309,
      "step": 561
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9653751030502886e-05,
      "loss": 5.3443,
      "step": 562
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.964756801319044e-05,
      "loss": 5.3349,
      "step": 563
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9641384995877988e-05,
      "loss": 5.7504,
      "step": 564
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.963520197856554e-05,
      "loss": 5.5106,
      "step": 565
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9629018961253094e-05,
      "loss": 5.4355,
      "step": 566
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9622835943940644e-05,
      "loss": 5.4441,
      "step": 567
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9616652926628195e-05,
      "loss": 5.5074,
      "step": 568
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9610469909315746e-05,
      "loss": 5.4754,
      "step": 569
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9604286892003297e-05,
      "loss": 5.5495,
      "step": 570
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9598103874690852e-05,
      "loss": 5.3079,
      "step": 571
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9591920857378403e-05,
      "loss": 5.4742,
      "step": 572
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.958573784006595e-05,
      "loss": 5.2292,
      "step": 573
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9579554822753505e-05,
      "loss": 5.184,
      "step": 574
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9573371805441056e-05,
      "loss": 5.2926,
      "step": 575
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.956718878812861e-05,
      "loss": 5.3267,
      "step": 576
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9561005770816158e-05,
      "loss": 5.4642,
      "step": 577
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.955482275350371e-05,
      "loss": 5.1962,
      "step": 578
    },
    {
      "epoch": 0.32,
      "learning_rate": 2.9548639736191263e-05,
      "loss": 5.5664,
      "step": 579
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9542456718878814e-05,
      "loss": 5.6008,
      "step": 580
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9536273701566365e-05,
      "loss": 5.3059,
      "step": 581
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9530090684253916e-05,
      "loss": 5.3557,
      "step": 582
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9523907666941467e-05,
      "loss": 5.166,
      "step": 583
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.951772464962902e-05,
      "loss": 5.6136,
      "step": 584
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9511541632316572e-05,
      "loss": 5.2976,
      "step": 585
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9505358615004123e-05,
      "loss": 5.0328,
      "step": 586
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9499175597691674e-05,
      "loss": 5.4383,
      "step": 587
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9492992580379225e-05,
      "loss": 5.5318,
      "step": 588
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9486809563066776e-05,
      "loss": 5.4578,
      "step": 589
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.948062654575433e-05,
      "loss": 5.6787,
      "step": 590
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9474443528441878e-05,
      "loss": 5.2692,
      "step": 591
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9468260511129433e-05,
      "loss": 5.429,
      "step": 592
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9462077493816984e-05,
      "loss": 5.0683,
      "step": 593
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9455894476504535e-05,
      "loss": 5.5062,
      "step": 594
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.944971145919209e-05,
      "loss": 5.2751,
      "step": 595
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9443528441879636e-05,
      "loss": 5.5247,
      "step": 596
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.9437345424567187e-05,
      "loss": 5.3127,
      "step": 597
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9431162407254742e-05,
      "loss": 4.9918,
      "step": 598
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9424979389942293e-05,
      "loss": 5.3875,
      "step": 599
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9418796372629844e-05,
      "loss": 5.4605,
      "step": 600
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9412613355317395e-05,
      "loss": 5.5951,
      "step": 601
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9406430338004946e-05,
      "loss": 5.7858,
      "step": 602
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.94002473206925e-05,
      "loss": 5.5544,
      "step": 603
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.939406430338005e-05,
      "loss": 5.4941,
      "step": 604
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9387881286067602e-05,
      "loss": 5.2258,
      "step": 605
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9381698268755153e-05,
      "loss": 5.2396,
      "step": 606
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9375515251442704e-05,
      "loss": 5.5031,
      "step": 607
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.936933223413026e-05,
      "loss": 5.2805,
      "step": 608
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9363149216817806e-05,
      "loss": 5.5403,
      "step": 609
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9356966199505357e-05,
      "loss": 5.1783,
      "step": 610
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.935078318219291e-05,
      "loss": 5.5835,
      "step": 611
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9344600164880462e-05,
      "loss": 5.3616,
      "step": 612
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9338417147568017e-05,
      "loss": 5.3444,
      "step": 613
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9332234130255564e-05,
      "loss": 5.4895,
      "step": 614
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.9326051112943115e-05,
      "loss": 5.2369,
      "step": 615
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.931986809563067e-05,
      "loss": 5.2288,
      "step": 616
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.931368507831822e-05,
      "loss": 5.3693,
      "step": 617
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.930750206100577e-05,
      "loss": 5.0981,
      "step": 618
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9301319043693323e-05,
      "loss": 5.2756,
      "step": 619
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9295136026380874e-05,
      "loss": 5.4271,
      "step": 620
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9288953009068428e-05,
      "loss": 5.3978,
      "step": 621
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.928276999175598e-05,
      "loss": 5.2228,
      "step": 622
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9276586974443527e-05,
      "loss": 5.3207,
      "step": 623
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.927040395713108e-05,
      "loss": 5.6184,
      "step": 624
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9264220939818632e-05,
      "loss": 5.3545,
      "step": 625
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9258037922506183e-05,
      "loss": 5.3439,
      "step": 626
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9251854905193737e-05,
      "loss": 5.2222,
      "step": 627
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9245671887881285e-05,
      "loss": 5.6089,
      "step": 628
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.923948887056884e-05,
      "loss": 5.6198,
      "step": 629
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.923330585325639e-05,
      "loss": 5.299,
      "step": 630
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.922712283594394e-05,
      "loss": 5.4093,
      "step": 631
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9220939818631492e-05,
      "loss": 5.2287,
      "step": 632
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9214756801319043e-05,
      "loss": 5.1944,
      "step": 633
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.9208573784006597e-05,
      "loss": 5.3845,
      "step": 634
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.920239076669415e-05,
      "loss": 5.3426,
      "step": 635
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.91962077493817e-05,
      "loss": 5.1357,
      "step": 636
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.919002473206925e-05,
      "loss": 5.2192,
      "step": 637
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.91838417147568e-05,
      "loss": 5.0221,
      "step": 638
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.9177658697444352e-05,
      "loss": 5.3339,
      "step": 639
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.9171475680131907e-05,
      "loss": 5.4661,
      "step": 640
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.9165292662819458e-05,
      "loss": 5.1834,
      "step": 641
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.915910964550701e-05,
      "loss": 5.3965,
      "step": 642
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.915292662819456e-05,
      "loss": 5.1956,
      "step": 643
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.914674361088211e-05,
      "loss": 5.2349,
      "step": 644
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.9140560593569665e-05,
      "loss": 4.9527,
      "step": 645
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.9134377576257213e-05,
      "loss": 5.1617,
      "step": 646
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.9128194558944764e-05,
      "loss": 5.4295,
      "step": 647
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.9122011541632318e-05,
      "loss": 5.3213,
      "step": 648
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.911582852431987e-05,
      "loss": 5.4255,
      "step": 649
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.9109645507007423e-05,
      "loss": 5.2061,
      "step": 650
    },
    {
      "epoch": 0.36,
      "learning_rate": 2.910346248969497e-05,
      "loss": 5.7266,
      "step": 651
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9097279472382522e-05,
      "loss": 5.3865,
      "step": 652
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9091096455070076e-05,
      "loss": 5.3109,
      "step": 653
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9084913437757627e-05,
      "loss": 5.3632,
      "step": 654
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9078730420445175e-05,
      "loss": 5.0781,
      "step": 655
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.907254740313273e-05,
      "loss": 5.2196,
      "step": 656
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.906636438582028e-05,
      "loss": 5.115,
      "step": 657
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9060181368507835e-05,
      "loss": 5.4551,
      "step": 658
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9053998351195386e-05,
      "loss": 5.2483,
      "step": 659
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9047815333882933e-05,
      "loss": 5.3254,
      "step": 660
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9041632316570488e-05,
      "loss": 5.4975,
      "step": 661
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.903544929925804e-05,
      "loss": 5.3125,
      "step": 662
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9029266281945593e-05,
      "loss": 5.3223,
      "step": 663
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.902308326463314e-05,
      "loss": 5.286,
      "step": 664
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.901690024732069e-05,
      "loss": 5.1097,
      "step": 665
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9010717230008246e-05,
      "loss": 5.2701,
      "step": 666
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.9004534212695797e-05,
      "loss": 5.3522,
      "step": 667
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.8998351195383348e-05,
      "loss": 5.1436,
      "step": 668
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.89921681780709e-05,
      "loss": 5.2097,
      "step": 669
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.898598516075845e-05,
      "loss": 5.2523,
      "step": 670
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8979802143446004e-05,
      "loss": 5.3753,
      "step": 671
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8973619126133555e-05,
      "loss": 5.2596,
      "step": 672
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8967436108821106e-05,
      "loss": 5.1452,
      "step": 673
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8961253091508657e-05,
      "loss": 5.102,
      "step": 674
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8955070074196208e-05,
      "loss": 5.1355,
      "step": 675
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.894888705688376e-05,
      "loss": 5.2513,
      "step": 676
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8942704039571313e-05,
      "loss": 5.0586,
      "step": 677
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.893652102225886e-05,
      "loss": 5.3843,
      "step": 678
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8930338004946415e-05,
      "loss": 5.102,
      "step": 679
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8924154987633966e-05,
      "loss": 5.257,
      "step": 680
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8917971970321517e-05,
      "loss": 5.2693,
      "step": 681
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.891178895300907e-05,
      "loss": 5.2065,
      "step": 682
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.890560593569662e-05,
      "loss": 5.1752,
      "step": 683
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.889942291838417e-05,
      "loss": 5.2305,
      "step": 684
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8893239901071725e-05,
      "loss": 5.2815,
      "step": 685
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8887056883759276e-05,
      "loss": 5.2455,
      "step": 686
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8880873866446827e-05,
      "loss": 5.1127,
      "step": 687
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8874690849134378e-05,
      "loss": 5.1398,
      "step": 688
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.886850783182193e-05,
      "loss": 5.3219,
      "step": 689
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8862324814509483e-05,
      "loss": 5.2657,
      "step": 690
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8856141797197034e-05,
      "loss": 5.0926,
      "step": 691
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8849958779884585e-05,
      "loss": 5.2779,
      "step": 692
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8843775762572136e-05,
      "loss": 5.316,
      "step": 693
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8837592745259687e-05,
      "loss": 5.2978,
      "step": 694
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.883140972794724e-05,
      "loss": 5.2909,
      "step": 695
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.882522671063479e-05,
      "loss": 5.2528,
      "step": 696
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.881904369332234e-05,
      "loss": 4.907,
      "step": 697
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8812860676009894e-05,
      "loss": 4.9649,
      "step": 698
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8806677658697445e-05,
      "loss": 5.1816,
      "step": 699
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8800494641385e-05,
      "loss": 4.9101,
      "step": 700
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8794311624072547e-05,
      "loss": 5.2272,
      "step": 701
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8788128606760098e-05,
      "loss": 5.1887,
      "step": 702
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8781945589447652e-05,
      "loss": 5.2119,
      "step": 703
    },
    {
      "epoch": 0.39,
      "learning_rate": 2.8775762572135203e-05,
      "loss": 5.3781,
      "step": 704
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8769579554822754e-05,
      "loss": 5.0271,
      "step": 705
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8763396537510305e-05,
      "loss": 5.2187,
      "step": 706
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8757213520197856e-05,
      "loss": 5.1961,
      "step": 707
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.875103050288541e-05,
      "loss": 5.2395,
      "step": 708
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8744847485572962e-05,
      "loss": 5.0505,
      "step": 709
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.873866446826051e-05,
      "loss": 5.2914,
      "step": 710
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8732481450948064e-05,
      "loss": 5.1359,
      "step": 711
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8726298433635615e-05,
      "loss": 5.1107,
      "step": 712
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8720115416323166e-05,
      "loss": 5.1161,
      "step": 713
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.871393239901072e-05,
      "loss": 5.5007,
      "step": 714
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8707749381698268e-05,
      "loss": 5.2493,
      "step": 715
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8701566364385822e-05,
      "loss": 5.0354,
      "step": 716
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8695383347073373e-05,
      "loss": 5.0705,
      "step": 717
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8689200329760924e-05,
      "loss": 5.3052,
      "step": 718
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8683017312448475e-05,
      "loss": 5.0939,
      "step": 719
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8676834295136026e-05,
      "loss": 5.2313,
      "step": 720
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.8670651277823577e-05,
      "loss": 5.286,
      "step": 721
    },
    {
      "epoch": 0.4,
      "learning_rate": 2.866446826051113e-05,
      "loss": 5.3997,
      "step": 722
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8658285243198682e-05,
      "loss": 5.2993,
      "step": 723
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8652102225886233e-05,
      "loss": 5.4062,
      "step": 724
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8645919208573784e-05,
      "loss": 5.2287,
      "step": 725
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8639736191261335e-05,
      "loss": 5.1159,
      "step": 726
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.863355317394889e-05,
      "loss": 4.9039,
      "step": 727
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.862737015663644e-05,
      "loss": 5.2914,
      "step": 728
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.862118713932399e-05,
      "loss": 5.133,
      "step": 729
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8615004122011542e-05,
      "loss": 5.1351,
      "step": 730
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8608821104699093e-05,
      "loss": 5.2261,
      "step": 731
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8602638087386648e-05,
      "loss": 5.0873,
      "step": 732
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8596455070074195e-05,
      "loss": 5.2245,
      "step": 733
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8590272052761746e-05,
      "loss": 5.233,
      "step": 734
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.85840890354493e-05,
      "loss": 5.1569,
      "step": 735
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8577906018136852e-05,
      "loss": 5.1381,
      "step": 736
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8571723000824406e-05,
      "loss": 5.1032,
      "step": 737
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8565539983511954e-05,
      "loss": 5.0614,
      "step": 738
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.8559356966199505e-05,
      "loss": 5.0468,
      "step": 739
    },
    {
      "epoch": 0.41,
      "learning_rate": 2.855317394888706e-05,
      "loss": 5.1049,
      "step": 740
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.854699093157461e-05,
      "loss": 5.2208,
      "step": 741
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8540807914262158e-05,
      "loss": 5.1479,
      "step": 742
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8534624896949712e-05,
      "loss": 5.1076,
      "step": 743
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8528441879637263e-05,
      "loss": 5.1219,
      "step": 744
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8522258862324817e-05,
      "loss": 5.4071,
      "step": 745
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.851607584501237e-05,
      "loss": 4.934,
      "step": 746
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8509892827699916e-05,
      "loss": 5.1473,
      "step": 747
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.850370981038747e-05,
      "loss": 5.0009,
      "step": 748
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.849752679307502e-05,
      "loss": 4.9874,
      "step": 749
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8491343775762572e-05,
      "loss": 5.044,
      "step": 750
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8485160758450123e-05,
      "loss": 5.3151,
      "step": 751
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8478977741137674e-05,
      "loss": 5.3359,
      "step": 752
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.847279472382523e-05,
      "loss": 5.259,
      "step": 753
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.846661170651278e-05,
      "loss": 5.0733,
      "step": 754
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.846042868920033e-05,
      "loss": 5.1571,
      "step": 755
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.845424567188788e-05,
      "loss": 5.3085,
      "step": 756
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8448062654575433e-05,
      "loss": 5.1497,
      "step": 757
    },
    {
      "epoch": 0.42,
      "learning_rate": 2.8441879637262987e-05,
      "loss": 5.128,
      "step": 758
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.8435696619950538e-05,
      "loss": 5.0527,
      "step": 759
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.842951360263809e-05,
      "loss": 5.0522,
      "step": 760
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.842333058532564e-05,
      "loss": 5.4062,
      "step": 761
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.841714756801319e-05,
      "loss": 4.9889,
      "step": 762
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.8410964550700742e-05,
      "loss": 5.278,
      "step": 763
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.8404781533388296e-05,
      "loss": 4.9941,
      "step": 764
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.8398598516075844e-05,
      "loss": 4.9369,
      "step": 765
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.8392415498763398e-05,
      "loss": 4.9573,
      "step": 766
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.838623248145095e-05,
      "loss": 5.1981,
      "step": 767
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.83800494641385e-05,
      "loss": 5.0608,
      "step": 768
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.8373866446826054e-05,
      "loss": 5.3507,
      "step": 769
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.8367683429513602e-05,
      "loss": 5.0425,
      "step": 770
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.8361500412201153e-05,
      "loss": 5.0674,
      "step": 771
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.8355317394888707e-05,
      "loss": 5.2775,
      "step": 772
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.834913437757626e-05,
      "loss": 5.1424,
      "step": 773
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.834295136026381e-05,
      "loss": 5.1323,
      "step": 774
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.833676834295136e-05,
      "loss": 5.2102,
      "step": 775
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.833058532563891e-05,
      "loss": 5.0493,
      "step": 776
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8324402308326466e-05,
      "loss": 4.9052,
      "step": 777
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8318219291014017e-05,
      "loss": 5.3509,
      "step": 778
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8312036273701564e-05,
      "loss": 5.161,
      "step": 779
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.830585325638912e-05,
      "loss": 5.1868,
      "step": 780
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.829967023907667e-05,
      "loss": 5.1628,
      "step": 781
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8293487221764224e-05,
      "loss": 5.1855,
      "step": 782
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.828730420445177e-05,
      "loss": 5.1084,
      "step": 783
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8281121187139323e-05,
      "loss": 5.124,
      "step": 784
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8274938169826877e-05,
      "loss": 4.9428,
      "step": 785
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8268755152514428e-05,
      "loss": 5.1226,
      "step": 786
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.826257213520198e-05,
      "loss": 4.7852,
      "step": 787
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.825638911788953e-05,
      "loss": 5.231,
      "step": 788
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.825020610057708e-05,
      "loss": 5.1106,
      "step": 789
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8244023083264635e-05,
      "loss": 5.1685,
      "step": 790
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8237840065952186e-05,
      "loss": 5.1574,
      "step": 791
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8231657048639737e-05,
      "loss": 4.9205,
      "step": 792
    },
    {
      "epoch": 0.44,
      "learning_rate": 2.8225474031327288e-05,
      "loss": 5.1017,
      "step": 793
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.821929101401484e-05,
      "loss": 5.0593,
      "step": 794
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8213107996702394e-05,
      "loss": 5.0787,
      "step": 795
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8206924979389944e-05,
      "loss": 5.1005,
      "step": 796
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8200741962077492e-05,
      "loss": 4.6095,
      "step": 797
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8194558944765046e-05,
      "loss": 5.2869,
      "step": 798
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8188375927452597e-05,
      "loss": 5.0375,
      "step": 799
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.818219291014015e-05,
      "loss": 5.0567,
      "step": 800
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8176009892827703e-05,
      "loss": 5.3824,
      "step": 801
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.816982687551525e-05,
      "loss": 5.4143,
      "step": 802
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8163643858202805e-05,
      "loss": 5.2348,
      "step": 803
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8157460840890356e-05,
      "loss": 5.1777,
      "step": 804
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8151277823577907e-05,
      "loss": 5.3406,
      "step": 805
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8145094806265458e-05,
      "loss": 5.2961,
      "step": 806
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.813891178895301e-05,
      "loss": 5.0327,
      "step": 807
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.813272877164056e-05,
      "loss": 4.8477,
      "step": 808
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8126545754328114e-05,
      "loss": 5.2276,
      "step": 809
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8120362737015665e-05,
      "loss": 5.2742,
      "step": 810
    },
    {
      "epoch": 0.45,
      "learning_rate": 2.8114179719703216e-05,
      "loss": 5.1928,
      "step": 811
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8107996702390767e-05,
      "loss": 5.184,
      "step": 812
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8101813685078318e-05,
      "loss": 5.1633,
      "step": 813
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8095630667765872e-05,
      "loss": 5.005,
      "step": 814
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.808944765045342e-05,
      "loss": 5.0074,
      "step": 815
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8083264633140974e-05,
      "loss": 5.1074,
      "step": 816
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8077081615828525e-05,
      "loss": 5.0794,
      "step": 817
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8070898598516076e-05,
      "loss": 5.2228,
      "step": 818
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.806471558120363e-05,
      "loss": 5.1719,
      "step": 819
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8058532563891178e-05,
      "loss": 5.1692,
      "step": 820
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.805234954657873e-05,
      "loss": 5.2093,
      "step": 821
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8046166529266284e-05,
      "loss": 4.8718,
      "step": 822
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8039983511953835e-05,
      "loss": 5.266,
      "step": 823
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.803380049464139e-05,
      "loss": 5.2395,
      "step": 824
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8027617477328936e-05,
      "loss": 4.9265,
      "step": 825
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8021434460016487e-05,
      "loss": 5.058,
      "step": 826
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8015251442704042e-05,
      "loss": 5.1117,
      "step": 827
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.8009068425391593e-05,
      "loss": 4.967,
      "step": 828
    },
    {
      "epoch": 0.46,
      "learning_rate": 2.800288540807914e-05,
      "loss": 5.0131,
      "step": 829
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7996702390766695e-05,
      "loss": 5.159,
      "step": 830
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7990519373454246e-05,
      "loss": 5.3765,
      "step": 831
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.79843363561418e-05,
      "loss": 5.3227,
      "step": 832
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.797815333882935e-05,
      "loss": 5.1404,
      "step": 833
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.79719703215169e-05,
      "loss": 4.7971,
      "step": 834
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7965787304204453e-05,
      "loss": 4.9082,
      "step": 835
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7959604286892004e-05,
      "loss": 5.043,
      "step": 836
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7953421269579555e-05,
      "loss": 5.1535,
      "step": 837
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7947238252267106e-05,
      "loss": 4.9707,
      "step": 838
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7941055234954657e-05,
      "loss": 4.9261,
      "step": 839
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.793487221764221e-05,
      "loss": 5.0195,
      "step": 840
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7928689200329762e-05,
      "loss": 5.0883,
      "step": 841
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7922506183017313e-05,
      "loss": 4.9,
      "step": 842
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7916323165704864e-05,
      "loss": 5.0288,
      "step": 843
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7910140148392415e-05,
      "loss": 5.2844,
      "step": 844
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.7903957131079966e-05,
      "loss": 4.9546,
      "step": 845
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.789777411376752e-05,
      "loss": 4.8847,
      "step": 846
    },
    {
      "epoch": 0.47,
      "learning_rate": 2.789159109645507e-05,
      "loss": 4.7989,
      "step": 847
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7885408079142623e-05,
      "loss": 5.1362,
      "step": 848
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7879225061830174e-05,
      "loss": 4.9939,
      "step": 849
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7873042044517725e-05,
      "loss": 4.8436,
      "step": 850
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.786685902720528e-05,
      "loss": 5.2751,
      "step": 851
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7860676009892827e-05,
      "loss": 5.3211,
      "step": 852
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.785449299258038e-05,
      "loss": 5.0414,
      "step": 853
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7848309975267932e-05,
      "loss": 5.1702,
      "step": 854
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7842126957955483e-05,
      "loss": 5.2231,
      "step": 855
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7835943940643037e-05,
      "loss": 5.018,
      "step": 856
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7829760923330585e-05,
      "loss": 4.9726,
      "step": 857
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7823577906018136e-05,
      "loss": 5.2033,
      "step": 858
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.781739488870569e-05,
      "loss": 5.1099,
      "step": 859
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.781121187139324e-05,
      "loss": 5.1553,
      "step": 860
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7805028854080792e-05,
      "loss": 4.8557,
      "step": 861
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7798845836768343e-05,
      "loss": 4.9122,
      "step": 862
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7792662819455894e-05,
      "loss": 4.918,
      "step": 863
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.778647980214345e-05,
      "loss": 4.9421,
      "step": 864
    },
    {
      "epoch": 0.48,
      "learning_rate": 2.7780296784831e-05,
      "loss": 5.051,
      "step": 865
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7774113767518547e-05,
      "loss": 5.1311,
      "step": 866
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.77679307502061e-05,
      "loss": 5.0165,
      "step": 867
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7761747732893652e-05,
      "loss": 5.1272,
      "step": 868
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7755564715581207e-05,
      "loss": 5.1127,
      "step": 869
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7749381698268754e-05,
      "loss": 5.1328,
      "step": 870
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7743198680956305e-05,
      "loss": 5.1025,
      "step": 871
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.773701566364386e-05,
      "loss": 5.0042,
      "step": 872
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.773083264633141e-05,
      "loss": 5.1284,
      "step": 873
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.772464962901896e-05,
      "loss": 4.9826,
      "step": 874
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7718466611706513e-05,
      "loss": 5.2926,
      "step": 875
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7712283594394064e-05,
      "loss": 5.0009,
      "step": 876
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7706100577081618e-05,
      "loss": 5.4075,
      "step": 877
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.769991755976917e-05,
      "loss": 5.0222,
      "step": 878
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.769373454245672e-05,
      "loss": 5.0253,
      "step": 879
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.768755152514427e-05,
      "loss": 5.027,
      "step": 880
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7681368507831822e-05,
      "loss": 5.1263,
      "step": 881
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7675185490519376e-05,
      "loss": 4.8239,
      "step": 882
    },
    {
      "epoch": 0.49,
      "learning_rate": 2.7669002473206927e-05,
      "loss": 5.1572,
      "step": 883
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.7662819455894475e-05,
      "loss": 5.0263,
      "step": 884
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.765663643858203e-05,
      "loss": 5.1692,
      "step": 885
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.765045342126958e-05,
      "loss": 5.2101,
      "step": 886
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.764427040395713e-05,
      "loss": 5.2766,
      "step": 887
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.7638087386644686e-05,
      "loss": 5.018,
      "step": 888
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.7631904369332233e-05,
      "loss": 4.9988,
      "step": 889
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.7625721352019788e-05,
      "loss": 5.0667,
      "step": 890
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.761953833470734e-05,
      "loss": 5.0805,
      "step": 891
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.761335531739489e-05,
      "loss": 5.077,
      "step": 892
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.760717230008244e-05,
      "loss": 5.2535,
      "step": 893
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.760098928276999e-05,
      "loss": 5.0759,
      "step": 894
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.7594806265457542e-05,
      "loss": 5.0184,
      "step": 895
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.7588623248145097e-05,
      "loss": 4.9982,
      "step": 896
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.7582440230832648e-05,
      "loss": 5.0284,
      "step": 897
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.75762572135202e-05,
      "loss": 4.8308,
      "step": 898
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.757007419620775e-05,
      "loss": 4.6554,
      "step": 899
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.75638911788953e-05,
      "loss": 4.9265,
      "step": 900
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7557708161582855e-05,
      "loss": 5.15,
      "step": 901
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7551525144270403e-05,
      "loss": 5.2692,
      "step": 902
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7545342126957954e-05,
      "loss": 5.3622,
      "step": 903
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7539159109645508e-05,
      "loss": 4.8174,
      "step": 904
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.753297609233306e-05,
      "loss": 5.2661,
      "step": 905
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7526793075020613e-05,
      "loss": 5.3023,
      "step": 906
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.752061005770816e-05,
      "loss": 5.0262,
      "step": 907
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7514427040395712e-05,
      "loss": 5.0233,
      "step": 908
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7508244023083266e-05,
      "loss": 5.1232,
      "step": 909
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7502061005770817e-05,
      "loss": 5.0099,
      "step": 910
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7495877988458368e-05,
      "loss": 5.1516,
      "step": 911
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.748969497114592e-05,
      "loss": 5.0001,
      "step": 912
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.748351195383347e-05,
      "loss": 4.9148,
      "step": 913
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7477328936521025e-05,
      "loss": 5.2425,
      "step": 914
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7471145919208576e-05,
      "loss": 4.997,
      "step": 915
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7464962901896123e-05,
      "loss": 5.0948,
      "step": 916
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.7458779884583678e-05,
      "loss": 4.9313,
      "step": 917
    },
    {
      "epoch": 0.51,
      "learning_rate": 2.745259686727123e-05,
      "loss": 5.2418,
      "step": 918
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7446413849958783e-05,
      "loss": 4.9694,
      "step": 919
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7440230832646334e-05,
      "loss": 4.6992,
      "step": 920
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.743404781533388e-05,
      "loss": 5.1115,
      "step": 921
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7427864798021436e-05,
      "loss": 4.7085,
      "step": 922
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7421681780708987e-05,
      "loss": 4.9233,
      "step": 923
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7415498763396538e-05,
      "loss": 5.134,
      "step": 924
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.740931574608409e-05,
      "loss": 5.0508,
      "step": 925
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.740313272877164e-05,
      "loss": 5.3092,
      "step": 926
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7396949711459194e-05,
      "loss": 4.9989,
      "step": 927
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7390766694146745e-05,
      "loss": 4.8421,
      "step": 928
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7384583676834296e-05,
      "loss": 4.9247,
      "step": 929
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7378400659521847e-05,
      "loss": 5.0924,
      "step": 930
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7372217642209398e-05,
      "loss": 5.1182,
      "step": 931
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.736603462489695e-05,
      "loss": 5.0649,
      "step": 932
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7359851607584503e-05,
      "loss": 4.8861,
      "step": 933
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.735366859027205e-05,
      "loss": 5.1063,
      "step": 934
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7347485572959605e-05,
      "loss": 5.2409,
      "step": 935
    },
    {
      "epoch": 0.52,
      "learning_rate": 2.7341302555647156e-05,
      "loss": 5.1278,
      "step": 936
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7335119538334707e-05,
      "loss": 4.883,
      "step": 937
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7328936521022262e-05,
      "loss": 4.7839,
      "step": 938
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.732275350370981e-05,
      "loss": 4.6731,
      "step": 939
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.731657048639736e-05,
      "loss": 5.0536,
      "step": 940
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7310387469084915e-05,
      "loss": 5.0602,
      "step": 941
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7304204451772466e-05,
      "loss": 5.0551,
      "step": 942
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.729802143446002e-05,
      "loss": 5.0497,
      "step": 943
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7291838417147568e-05,
      "loss": 4.9595,
      "step": 944
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.728565539983512e-05,
      "loss": 5.0736,
      "step": 945
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7279472382522673e-05,
      "loss": 5.0899,
      "step": 946
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7273289365210224e-05,
      "loss": 4.5167,
      "step": 947
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7267106347897775e-05,
      "loss": 4.9615,
      "step": 948
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7260923330585326e-05,
      "loss": 5.442,
      "step": 949
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7254740313272877e-05,
      "loss": 4.4003,
      "step": 950
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.724855729596043e-05,
      "loss": 5.2226,
      "step": 951
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7242374278647982e-05,
      "loss": 5.1868,
      "step": 952
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.723619126133553e-05,
      "loss": 5.2227,
      "step": 953
    },
    {
      "epoch": 0.53,
      "learning_rate": 2.7230008244023084e-05,
      "loss": 4.9726,
      "step": 954
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7223825226710635e-05,
      "loss": 5.1792,
      "step": 955
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.721764220939819e-05,
      "loss": 4.8851,
      "step": 956
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7211459192085737e-05,
      "loss": 5.2344,
      "step": 957
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7205276174773288e-05,
      "loss": 4.987,
      "step": 958
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7199093157460842e-05,
      "loss": 5.2296,
      "step": 959
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7192910140148393e-05,
      "loss": 4.8593,
      "step": 960
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7186727122835944e-05,
      "loss": 4.8314,
      "step": 961
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7180544105523495e-05,
      "loss": 5.0007,
      "step": 962
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7174361088211046e-05,
      "loss": 4.9187,
      "step": 963
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.71681780708986e-05,
      "loss": 4.9405,
      "step": 964
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7161995053586152e-05,
      "loss": 4.9523,
      "step": 965
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7155812036273703e-05,
      "loss": 4.8908,
      "step": 966
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7149629018961254e-05,
      "loss": 5.0855,
      "step": 967
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7143446001648805e-05,
      "loss": 5.046,
      "step": 968
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7137262984336356e-05,
      "loss": 5.0262,
      "step": 969
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.713107996702391e-05,
      "loss": 4.9047,
      "step": 970
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7124896949711458e-05,
      "loss": 4.9759,
      "step": 971
    },
    {
      "epoch": 0.54,
      "learning_rate": 2.7118713932399012e-05,
      "loss": 4.8084,
      "step": 972
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7112530915086563e-05,
      "loss": 5.0045,
      "step": 973
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7106347897774114e-05,
      "loss": 5.1883,
      "step": 974
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.710016488046167e-05,
      "loss": 4.9863,
      "step": 975
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7093981863149216e-05,
      "loss": 4.8026,
      "step": 976
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.708779884583677e-05,
      "loss": 5.3387,
      "step": 977
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.708161582852432e-05,
      "loss": 4.9244,
      "step": 978
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7075432811211872e-05,
      "loss": 4.8267,
      "step": 979
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7069249793899423e-05,
      "loss": 4.7606,
      "step": 980
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7063066776586974e-05,
      "loss": 4.5663,
      "step": 981
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7056883759274525e-05,
      "loss": 4.7708,
      "step": 982
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.705070074196208e-05,
      "loss": 5.1102,
      "step": 983
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.704451772464963e-05,
      "loss": 5.0061,
      "step": 984
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.703833470733718e-05,
      "loss": 4.8819,
      "step": 985
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7032151690024733e-05,
      "loss": 4.9976,
      "step": 986
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7025968672712284e-05,
      "loss": 5.1626,
      "step": 987
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7019785655399838e-05,
      "loss": 5.039,
      "step": 988
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7013602638087385e-05,
      "loss": 5.1844,
      "step": 989
    },
    {
      "epoch": 0.55,
      "learning_rate": 2.7007419620774936e-05,
      "loss": 5.0152,
      "step": 990
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.700123660346249e-05,
      "loss": 4.8583,
      "step": 991
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6995053586150042e-05,
      "loss": 5.3737,
      "step": 992
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6988870568837596e-05,
      "loss": 4.9754,
      "step": 993
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6982687551525144e-05,
      "loss": 5.0693,
      "step": 994
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6976504534212695e-05,
      "loss": 5.1031,
      "step": 995
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.697032151690025e-05,
      "loss": 4.2616,
      "step": 996
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.69641384995878e-05,
      "loss": 4.6983,
      "step": 997
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.695795548227535e-05,
      "loss": 5.5624,
      "step": 998
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6951772464962902e-05,
      "loss": 5.0936,
      "step": 999
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6945589447650453e-05,
      "loss": 4.6164,
      "step": 1000
    },
    {
      "epoch": 0.56,
      "eval_loss": 4.887108325958252,
      "eval_runtime": 666.7746,
      "eval_samples_per_second": 3.962,
      "eval_steps_per_second": 0.496,
      "eval_wer": 0.9780542492608687,
      "step": 1000
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6939406430338007e-05,
      "loss": 5.4145,
      "step": 1001
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.693322341302556e-05,
      "loss": 5.0703,
      "step": 1002
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6927040395713106e-05,
      "loss": 5.145,
      "step": 1003
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.692085737840066e-05,
      "loss": 5.0573,
      "step": 1004
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.691467436108821e-05,
      "loss": 4.9777,
      "step": 1005
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6908491343775766e-05,
      "loss": 5.1226,
      "step": 1006
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.6902308326463317e-05,
      "loss": 4.8744,
      "step": 1007
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.6896125309150864e-05,
      "loss": 4.998,
      "step": 1008
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.688994229183842e-05,
      "loss": 4.9256,
      "step": 1009
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.688375927452597e-05,
      "loss": 4.9134,
      "step": 1010
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.687757625721352e-05,
      "loss": 4.9615,
      "step": 1011
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.687139323990107e-05,
      "loss": 5.0331,
      "step": 1012
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.6865210222588623e-05,
      "loss": 5.0727,
      "step": 1013
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.6859027205276177e-05,
      "loss": 5.1496,
      "step": 1014
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.6852844187963728e-05,
      "loss": 5.3237,
      "step": 1015
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.684666117065128e-05,
      "loss": 4.9186,
      "step": 1016
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.684047815333883e-05,
      "loss": 5.0623,
      "step": 1017
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.683429513602638e-05,
      "loss": 4.9726,
      "step": 1018
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.6828112118713932e-05,
      "loss": 5.0387,
      "step": 1019
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.6821929101401486e-05,
      "loss": 5.0381,
      "step": 1020
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.6815746084089034e-05,
      "loss": 5.0538,
      "step": 1021
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.6809563066776588e-05,
      "loss": 4.9868,
      "step": 1022
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.680338004946414e-05,
      "loss": 4.9661,
      "step": 1023
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.679719703215169e-05,
      "loss": 5.0088,
      "step": 1024
    },
    {
      "epoch": 0.57,
      "learning_rate": 2.6791014014839245e-05,
      "loss": 4.9791,
      "step": 1025
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6784830997526792e-05,
      "loss": 4.7638,
      "step": 1026
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6778647980214343e-05,
      "loss": 5.0471,
      "step": 1027
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6772464962901897e-05,
      "loss": 5.0738,
      "step": 1028
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.676628194558945e-05,
      "loss": 4.9639,
      "step": 1029
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6760098928277003e-05,
      "loss": 4.9713,
      "step": 1030
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.675391591096455e-05,
      "loss": 4.8309,
      "step": 1031
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.67477328936521e-05,
      "loss": 4.8095,
      "step": 1032
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6741549876339656e-05,
      "loss": 4.8943,
      "step": 1033
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6735366859027207e-05,
      "loss": 4.7352,
      "step": 1034
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6729183841714758e-05,
      "loss": 5.0218,
      "step": 1035
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.672300082440231e-05,
      "loss": 5.0986,
      "step": 1036
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.671681780708986e-05,
      "loss": 4.8847,
      "step": 1037
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6710634789777414e-05,
      "loss": 4.9132,
      "step": 1038
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6704451772464965e-05,
      "loss": 5.1066,
      "step": 1039
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6698268755152513e-05,
      "loss": 4.9849,
      "step": 1040
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6692085737840067e-05,
      "loss": 5.1408,
      "step": 1041
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6685902720527618e-05,
      "loss": 5.054,
      "step": 1042
    },
    {
      "epoch": 0.58,
      "learning_rate": 2.6679719703215172e-05,
      "loss": 5.0559,
      "step": 1043
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.667353668590272e-05,
      "loss": 4.8387,
      "step": 1044
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.666735366859027e-05,
      "loss": 4.9909,
      "step": 1045
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6661170651277825e-05,
      "loss": 5.0078,
      "step": 1046
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6654987633965376e-05,
      "loss": 4.8245,
      "step": 1047
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6648804616652927e-05,
      "loss": 4.6872,
      "step": 1048
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6642621599340478e-05,
      "loss": 4.7753,
      "step": 1049
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.663643858202803e-05,
      "loss": 5.4226,
      "step": 1050
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6630255564715584e-05,
      "loss": 5.5609,
      "step": 1051
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6624072547403135e-05,
      "loss": 5.0961,
      "step": 1052
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6617889530090682e-05,
      "loss": 4.8832,
      "step": 1053
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6611706512778237e-05,
      "loss": 4.8612,
      "step": 1054
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6605523495465787e-05,
      "loss": 4.9932,
      "step": 1055
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.659934047815334e-05,
      "loss": 4.9811,
      "step": 1056
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6593157460840893e-05,
      "loss": 5.1003,
      "step": 1057
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.658697444352844e-05,
      "loss": 4.9067,
      "step": 1058
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6580791426215995e-05,
      "loss": 5.0643,
      "step": 1059
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6574608408903546e-05,
      "loss": 4.9875,
      "step": 1060
    },
    {
      "epoch": 0.59,
      "learning_rate": 2.6568425391591097e-05,
      "loss": 4.7926,
      "step": 1061
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.656224237427865e-05,
      "loss": 5.0273,
      "step": 1062
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.65560593569662e-05,
      "loss": 5.1285,
      "step": 1063
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.654987633965375e-05,
      "loss": 4.6824,
      "step": 1064
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6543693322341304e-05,
      "loss": 5.0264,
      "step": 1065
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6537510305028855e-05,
      "loss": 4.7743,
      "step": 1066
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6531327287716406e-05,
      "loss": 5.1941,
      "step": 1067
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6525144270403957e-05,
      "loss": 4.8287,
      "step": 1068
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6518961253091508e-05,
      "loss": 5.0058,
      "step": 1069
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6512778235779062e-05,
      "loss": 4.8511,
      "step": 1070
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6506595218466613e-05,
      "loss": 4.8902,
      "step": 1071
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6500412201154164e-05,
      "loss": 4.8018,
      "step": 1072
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6494229183841715e-05,
      "loss": 4.8636,
      "step": 1073
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6488046166529266e-05,
      "loss": 4.8673,
      "step": 1074
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.648186314921682e-05,
      "loss": 4.9064,
      "step": 1075
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6475680131904368e-05,
      "loss": 5.0842,
      "step": 1076
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.646949711459192e-05,
      "loss": 4.9505,
      "step": 1077
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6463314097279474e-05,
      "loss": 4.8346,
      "step": 1078
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.6457131079967025e-05,
      "loss": 4.9686,
      "step": 1079
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.645094806265458e-05,
      "loss": 4.9598,
      "step": 1080
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6444765045342127e-05,
      "loss": 5.0809,
      "step": 1081
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6438582028029678e-05,
      "loss": 4.8935,
      "step": 1082
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6432399010717232e-05,
      "loss": 4.7081,
      "step": 1083
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6426215993404783e-05,
      "loss": 4.8564,
      "step": 1084
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6420032976092334e-05,
      "loss": 4.7156,
      "step": 1085
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6413849958779885e-05,
      "loss": 5.1152,
      "step": 1086
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6407666941467436e-05,
      "loss": 5.2206,
      "step": 1087
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.640148392415499e-05,
      "loss": 4.9948,
      "step": 1088
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.639530090684254e-05,
      "loss": 5.1111,
      "step": 1089
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.638911788953009e-05,
      "loss": 4.7541,
      "step": 1090
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6382934872217643e-05,
      "loss": 4.8942,
      "step": 1091
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6376751854905194e-05,
      "loss": 5.2073,
      "step": 1092
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6370568837592745e-05,
      "loss": 5.0348,
      "step": 1093
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.63643858202803e-05,
      "loss": 4.7234,
      "step": 1094
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6358202802967847e-05,
      "loss": 4.9039,
      "step": 1095
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.63520197856554e-05,
      "loss": 4.9428,
      "step": 1096
    },
    {
      "epoch": 0.61,
      "learning_rate": 2.6345836768342952e-05,
      "loss": 5.114,
      "step": 1097
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6339653751030503e-05,
      "loss": 5.5488,
      "step": 1098
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6333470733718054e-05,
      "loss": 4.8797,
      "step": 1099
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6327287716405605e-05,
      "loss": 4.8563,
      "step": 1100
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.632110469909316e-05,
      "loss": 4.9274,
      "step": 1101
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.631492168178071e-05,
      "loss": 5.1975,
      "step": 1102
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.630873866446826e-05,
      "loss": 5.1286,
      "step": 1103
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6302555647155813e-05,
      "loss": 4.8664,
      "step": 1104
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6296372629843364e-05,
      "loss": 4.7776,
      "step": 1105
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6290189612530915e-05,
      "loss": 4.8617,
      "step": 1106
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.628400659521847e-05,
      "loss": 4.7462,
      "step": 1107
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6277823577906017e-05,
      "loss": 4.9115,
      "step": 1108
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.627164056059357e-05,
      "loss": 5.0474,
      "step": 1109
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6265457543281122e-05,
      "loss": 4.9914,
      "step": 1110
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6259274525968673e-05,
      "loss": 4.6716,
      "step": 1111
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6253091508656227e-05,
      "loss": 5.1713,
      "step": 1112
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6246908491343775e-05,
      "loss": 4.6925,
      "step": 1113
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.6240725474031326e-05,
      "loss": 5.1356,
      "step": 1114
    },
    {
      "epoch": 0.62,
      "learning_rate": 2.623454245671888e-05,
      "loss": 5.1046,
      "step": 1115
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.622835943940643e-05,
      "loss": 4.8736,
      "step": 1116
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6222176422093986e-05,
      "loss": 4.8859,
      "step": 1117
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6215993404781533e-05,
      "loss": 4.9622,
      "step": 1118
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6209810387469084e-05,
      "loss": 4.867,
      "step": 1119
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.620362737015664e-05,
      "loss": 4.8798,
      "step": 1120
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.619744435284419e-05,
      "loss": 4.995,
      "step": 1121
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6191261335531737e-05,
      "loss": 4.918,
      "step": 1122
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.618507831821929e-05,
      "loss": 4.7501,
      "step": 1123
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6178895300906842e-05,
      "loss": 4.805,
      "step": 1124
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6172712283594397e-05,
      "loss": 4.9769,
      "step": 1125
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6166529266281948e-05,
      "loss": 4.7333,
      "step": 1126
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6160346248969495e-05,
      "loss": 4.8167,
      "step": 1127
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.615416323165705e-05,
      "loss": 5.1666,
      "step": 1128
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.61479802143446e-05,
      "loss": 4.7766,
      "step": 1129
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6141797197032155e-05,
      "loss": 5.1319,
      "step": 1130
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6135614179719703e-05,
      "loss": 5.1379,
      "step": 1131
    },
    {
      "epoch": 0.63,
      "learning_rate": 2.6129431162407254e-05,
      "loss": 5.0557,
      "step": 1132
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.6123248145094808e-05,
      "loss": 4.8742,
      "step": 1133
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.611706512778236e-05,
      "loss": 4.9773,
      "step": 1134
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.611088211046991e-05,
      "loss": 5.0956,
      "step": 1135
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.610469909315746e-05,
      "loss": 4.9144,
      "step": 1136
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.6098516075845012e-05,
      "loss": 4.8839,
      "step": 1137
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.6092333058532566e-05,
      "loss": 4.9693,
      "step": 1138
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.6086150041220117e-05,
      "loss": 4.898,
      "step": 1139
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.6079967023907665e-05,
      "loss": 4.7814,
      "step": 1140
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.607378400659522e-05,
      "loss": 4.9437,
      "step": 1141
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.606760098928277e-05,
      "loss": 5.2085,
      "step": 1142
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.606141797197032e-05,
      "loss": 5.0101,
      "step": 1143
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.6055234954657876e-05,
      "loss": 5.0839,
      "step": 1144
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.6049051937345423e-05,
      "loss": 4.8346,
      "step": 1145
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.6042868920032978e-05,
      "loss": 5.0354,
      "step": 1146
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.603668590272053e-05,
      "loss": 4.9891,
      "step": 1147
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.603050288540808e-05,
      "loss": 4.6912,
      "step": 1148
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.6024319868095634e-05,
      "loss": 4.5245,
      "step": 1149
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.601813685078318e-05,
      "loss": 4.6477,
      "step": 1150
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.6011953833470732e-05,
      "loss": 4.9735,
      "step": 1151
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.6005770816158287e-05,
      "loss": 5.1681,
      "step": 1152
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5999587798845838e-05,
      "loss": 4.9979,
      "step": 1153
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.599340478153339e-05,
      "loss": 4.977,
      "step": 1154
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.598722176422094e-05,
      "loss": 5.0564,
      "step": 1155
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.598103874690849e-05,
      "loss": 4.9594,
      "step": 1156
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5974855729596045e-05,
      "loss": 4.849,
      "step": 1157
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5968672712283596e-05,
      "loss": 4.9423,
      "step": 1158
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5962489694971147e-05,
      "loss": 4.8406,
      "step": 1159
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5956306677658698e-05,
      "loss": 4.7667,
      "step": 1160
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.595012366034625e-05,
      "loss": 5.074,
      "step": 1161
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5943940643033803e-05,
      "loss": 4.8976,
      "step": 1162
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.593775762572135e-05,
      "loss": 4.8479,
      "step": 1163
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5931574608408902e-05,
      "loss": 4.6801,
      "step": 1164
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5925391591096456e-05,
      "loss": 4.9137,
      "step": 1165
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5919208573784007e-05,
      "loss": 5.0982,
      "step": 1166
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.5913025556471562e-05,
      "loss": 5.0113,
      "step": 1167
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.590684253915911e-05,
      "loss": 4.5693,
      "step": 1168
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.590065952184666e-05,
      "loss": 5.0139,
      "step": 1169
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5894476504534215e-05,
      "loss": 5.0103,
      "step": 1170
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5888293487221766e-05,
      "loss": 4.6969,
      "step": 1171
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5882110469909317e-05,
      "loss": 4.6319,
      "step": 1172
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5875927452596868e-05,
      "loss": 5.1122,
      "step": 1173
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.586974443528442e-05,
      "loss": 4.7995,
      "step": 1174
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5863561417971973e-05,
      "loss": 4.8364,
      "step": 1175
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5857378400659524e-05,
      "loss": 4.9522,
      "step": 1176
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.585119538334707e-05,
      "loss": 4.8988,
      "step": 1177
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5845012366034626e-05,
      "loss": 4.8777,
      "step": 1178
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5838829348722177e-05,
      "loss": 4.5769,
      "step": 1179
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5832646331409728e-05,
      "loss": 4.9207,
      "step": 1180
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5826463314097282e-05,
      "loss": 4.8155,
      "step": 1181
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.582028029678483e-05,
      "loss": 4.7215,
      "step": 1182
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5814097279472384e-05,
      "loss": 4.7038,
      "step": 1183
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5807914262159935e-05,
      "loss": 4.8427,
      "step": 1184
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5801731244847486e-05,
      "loss": 5.33,
      "step": 1185
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.5795548227535037e-05,
      "loss": 4.6942,
      "step": 1186
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5789365210222588e-05,
      "loss": 4.8991,
      "step": 1187
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.578318219291014e-05,
      "loss": 4.8864,
      "step": 1188
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5776999175597693e-05,
      "loss": 4.7414,
      "step": 1189
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5770816158285244e-05,
      "loss": 4.6266,
      "step": 1190
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5764633140972795e-05,
      "loss": 4.7421,
      "step": 1191
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5758450123660346e-05,
      "loss": 4.7772,
      "step": 1192
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5752267106347897e-05,
      "loss": 5.08,
      "step": 1193
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5746084089035452e-05,
      "loss": 4.9543,
      "step": 1194
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5739901071723e-05,
      "loss": 4.7403,
      "step": 1195
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5733718054410554e-05,
      "loss": 4.6435,
      "step": 1196
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5727535037098105e-05,
      "loss": 4.6797,
      "step": 1197
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5721352019785656e-05,
      "loss": 4.859,
      "step": 1198
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.571516900247321e-05,
      "loss": 4.842,
      "step": 1199
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5708985985160758e-05,
      "loss": 4.4854,
      "step": 1200
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.570280296784831e-05,
      "loss": 5.1505,
      "step": 1201
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5696619950535863e-05,
      "loss": 4.8358,
      "step": 1202
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5690436933223414e-05,
      "loss": 4.9472,
      "step": 1203
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.568425391591097e-05,
      "loss": 4.7851,
      "step": 1204
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5678070898598516e-05,
      "loss": 4.6222,
      "step": 1205
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5671887881286067e-05,
      "loss": 4.9022,
      "step": 1206
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.566570486397362e-05,
      "loss": 5.0827,
      "step": 1207
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5659521846661172e-05,
      "loss": 4.9679,
      "step": 1208
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.565333882934872e-05,
      "loss": 4.8899,
      "step": 1209
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5647155812036274e-05,
      "loss": 4.6589,
      "step": 1210
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5640972794723825e-05,
      "loss": 4.9692,
      "step": 1211
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.563478977741138e-05,
      "loss": 4.8652,
      "step": 1212
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.562860676009893e-05,
      "loss": 4.9574,
      "step": 1213
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5622423742786478e-05,
      "loss": 4.9088,
      "step": 1214
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5616240725474033e-05,
      "loss": 4.8291,
      "step": 1215
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5610057708161584e-05,
      "loss": 4.8859,
      "step": 1216
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5603874690849135e-05,
      "loss": 4.767,
      "step": 1217
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5597691673536685e-05,
      "loss": 4.6147,
      "step": 1218
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5591508656224236e-05,
      "loss": 4.7684,
      "step": 1219
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.558532563891179e-05,
      "loss": 5.0167,
      "step": 1220
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5579142621599342e-05,
      "loss": 4.7613,
      "step": 1221
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5572959604286893e-05,
      "loss": 4.8122,
      "step": 1222
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5566776586974444e-05,
      "loss": 5.0041,
      "step": 1223
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5560593569661995e-05,
      "loss": 4.7878,
      "step": 1224
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.555441055234955e-05,
      "loss": 4.7803,
      "step": 1225
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.55482275350371e-05,
      "loss": 4.8905,
      "step": 1226
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5542044517724648e-05,
      "loss": 4.9528,
      "step": 1227
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5535861500412202e-05,
      "loss": 5.0063,
      "step": 1228
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5529678483099753e-05,
      "loss": 4.764,
      "step": 1229
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5523495465787304e-05,
      "loss": 5.1096,
      "step": 1230
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.551731244847486e-05,
      "loss": 4.87,
      "step": 1231
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5511129431162406e-05,
      "loss": 4.6875,
      "step": 1232
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.550494641384996e-05,
      "loss": 4.9332,
      "step": 1233
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.549876339653751e-05,
      "loss": 4.8026,
      "step": 1234
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5492580379225062e-05,
      "loss": 5.1795,
      "step": 1235
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5486397361912617e-05,
      "loss": 4.8931,
      "step": 1236
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5480214344600164e-05,
      "loss": 4.8291,
      "step": 1237
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.5474031327287715e-05,
      "loss": 4.6316,
      "step": 1238
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.546784830997527e-05,
      "loss": 4.9901,
      "step": 1239
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.546166529266282e-05,
      "loss": 4.9102,
      "step": 1240
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.545548227535037e-05,
      "loss": 4.8858,
      "step": 1241
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5449299258037923e-05,
      "loss": 5.2103,
      "step": 1242
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5443116240725474e-05,
      "loss": 4.8011,
      "step": 1243
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5436933223413028e-05,
      "loss": 4.8575,
      "step": 1244
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.543075020610058e-05,
      "loss": 4.5216,
      "step": 1245
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5424567188788127e-05,
      "loss": 5.0342,
      "step": 1246
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.541838417147568e-05,
      "loss": 4.8097,
      "step": 1247
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5412201154163232e-05,
      "loss": 4.8448,
      "step": 1248
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5406018136850786e-05,
      "loss": 4.8977,
      "step": 1249
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5399835119538334e-05,
      "loss": 4.8569,
      "step": 1250
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5393652102225885e-05,
      "loss": 5.1505,
      "step": 1251
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.538746908491344e-05,
      "loss": 4.8778,
      "step": 1252
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.538128606760099e-05,
      "loss": 5.1084,
      "step": 1253
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.537510305028854e-05,
      "loss": 5.094,
      "step": 1254
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5368920032976092e-05,
      "loss": 5.1851,
      "step": 1255
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5362737015663643e-05,
      "loss": 4.8388,
      "step": 1256
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.5356553998351197e-05,
      "loss": 4.7799,
      "step": 1257
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.535037098103875e-05,
      "loss": 4.7009,
      "step": 1258
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.5344187963726296e-05,
      "loss": 4.9216,
      "step": 1259
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.533800494641385e-05,
      "loss": 5.1708,
      "step": 1260
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.53318219291014e-05,
      "loss": 4.7236,
      "step": 1261
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.5325638911788956e-05,
      "loss": 4.8985,
      "step": 1262
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.5319455894476507e-05,
      "loss": 4.9122,
      "step": 1263
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.5313272877164054e-05,
      "loss": 4.7738,
      "step": 1264
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.530708985985161e-05,
      "loss": 4.842,
      "step": 1265
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.530090684253916e-05,
      "loss": 4.9568,
      "step": 1266
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.529472382522671e-05,
      "loss": 4.6103,
      "step": 1267
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.5288540807914265e-05,
      "loss": 4.9317,
      "step": 1268
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.5282357790601813e-05,
      "loss": 4.97,
      "step": 1269
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.5276174773289367e-05,
      "loss": 4.7459,
      "step": 1270
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.5269991755976918e-05,
      "loss": 4.7313,
      "step": 1271
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.526380873866447e-05,
      "loss": 4.8117,
      "step": 1272
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.525762572135202e-05,
      "loss": 4.9075,
      "step": 1273
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.525144270403957e-05,
      "loss": 4.5964,
      "step": 1274
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.5245259686727122e-05,
      "loss": 4.9061,
      "step": 1275
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5239076669414676e-05,
      "loss": 5.1522,
      "step": 1276
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5232893652102227e-05,
      "loss": 5.0761,
      "step": 1277
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5226710634789778e-05,
      "loss": 4.9107,
      "step": 1278
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.522052761747733e-05,
      "loss": 5.0614,
      "step": 1279
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.521434460016488e-05,
      "loss": 4.914,
      "step": 1280
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5208161582852435e-05,
      "loss": 4.8465,
      "step": 1281
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5201978565539982e-05,
      "loss": 4.7844,
      "step": 1282
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5195795548227537e-05,
      "loss": 5.0141,
      "step": 1283
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5189612530915088e-05,
      "loss": 4.8255,
      "step": 1284
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.518342951360264e-05,
      "loss": 4.7903,
      "step": 1285
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5177246496290193e-05,
      "loss": 4.8954,
      "step": 1286
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.517106347897774e-05,
      "loss": 4.8847,
      "step": 1287
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.516488046166529e-05,
      "loss": 4.7944,
      "step": 1288
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5158697444352846e-05,
      "loss": 4.9238,
      "step": 1289
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.5152514427040397e-05,
      "loss": 4.5289,
      "step": 1290
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.514633140972795e-05,
      "loss": 4.9529,
      "step": 1291
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.51401483924155e-05,
      "loss": 4.8474,
      "step": 1292
    },
    {
      "epoch": 0.72,
      "learning_rate": 2.513396537510305e-05,
      "loss": 5.1621,
      "step": 1293
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5127782357790604e-05,
      "loss": 4.5221,
      "step": 1294
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5121599340478155e-05,
      "loss": 4.659,
      "step": 1295
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5115416323165703e-05,
      "loss": 4.7044,
      "step": 1296
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5109233305853257e-05,
      "loss": 4.5416,
      "step": 1297
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5103050288540808e-05,
      "loss": 4.2229,
      "step": 1298
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5096867271228362e-05,
      "loss": 4.9191,
      "step": 1299
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5090684253915913e-05,
      "loss": 4.2715,
      "step": 1300
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.508450123660346e-05,
      "loss": 5.0503,
      "step": 1301
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5078318219291015e-05,
      "loss": 5.1075,
      "step": 1302
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5072135201978566e-05,
      "loss": 4.9359,
      "step": 1303
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5065952184666117e-05,
      "loss": 5.0719,
      "step": 1304
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5059769167353668e-05,
      "loss": 4.7327,
      "step": 1305
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.505358615004122e-05,
      "loss": 5.0169,
      "step": 1306
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5047403132728774e-05,
      "loss": 4.7854,
      "step": 1307
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5041220115416325e-05,
      "loss": 4.798,
      "step": 1308
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5035037098103876e-05,
      "loss": 4.7651,
      "step": 1309
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5028854080791427e-05,
      "loss": 4.6903,
      "step": 1310
    },
    {
      "epoch": 0.73,
      "learning_rate": 2.5022671063478978e-05,
      "loss": 4.6764,
      "step": 1311
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.501648804616653e-05,
      "loss": 4.8952,
      "step": 1312
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.5010305028854083e-05,
      "loss": 4.7997,
      "step": 1313
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.500412201154163e-05,
      "loss": 4.7273,
      "step": 1314
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4997938994229185e-05,
      "loss": 5.0474,
      "step": 1315
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4991755976916736e-05,
      "loss": 4.7927,
      "step": 1316
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4985572959604287e-05,
      "loss": 4.7037,
      "step": 1317
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.497938994229184e-05,
      "loss": 5.0804,
      "step": 1318
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.497320692497939e-05,
      "loss": 4.7662,
      "step": 1319
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4967023907666943e-05,
      "loss": 4.5565,
      "step": 1320
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4960840890354494e-05,
      "loss": 4.5382,
      "step": 1321
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4954657873042045e-05,
      "loss": 4.7858,
      "step": 1322
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.49484748557296e-05,
      "loss": 4.8455,
      "step": 1323
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4942291838417147e-05,
      "loss": 4.6919,
      "step": 1324
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4936108821104698e-05,
      "loss": 4.8476,
      "step": 1325
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4929925803792252e-05,
      "loss": 4.6162,
      "step": 1326
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4923742786479803e-05,
      "loss": 4.5293,
      "step": 1327
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4917559769167354e-05,
      "loss": 4.7376,
      "step": 1328
    },
    {
      "epoch": 0.74,
      "learning_rate": 2.4911376751854905e-05,
      "loss": 5.0416,
      "step": 1329
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4905193734542456e-05,
      "loss": 4.8876,
      "step": 1330
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.489901071723001e-05,
      "loss": 4.5845,
      "step": 1331
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4892827699917562e-05,
      "loss": 4.7264,
      "step": 1332
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.488664468260511e-05,
      "loss": 4.8058,
      "step": 1333
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4880461665292664e-05,
      "loss": 4.8424,
      "step": 1334
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4874278647980215e-05,
      "loss": 4.6142,
      "step": 1335
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.486809563066777e-05,
      "loss": 4.7372,
      "step": 1336
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4861912613355317e-05,
      "loss": 4.9721,
      "step": 1337
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4855729596042868e-05,
      "loss": 4.7447,
      "step": 1338
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4849546578730422e-05,
      "loss": 4.9666,
      "step": 1339
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4843363561417973e-05,
      "loss": 4.9163,
      "step": 1340
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4837180544105524e-05,
      "loss": 4.9066,
      "step": 1341
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4830997526793075e-05,
      "loss": 4.8551,
      "step": 1342
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.4824814509480626e-05,
      "loss": 4.9302,
      "step": 1343
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.481863149216818e-05,
      "loss": 4.7993,
      "step": 1344
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.481244847485573e-05,
      "loss": 4.8265,
      "step": 1345
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.480626545754328e-05,
      "loss": 4.8109,
      "step": 1346
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4800082440230833e-05,
      "loss": 4.9972,
      "step": 1347
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4793899422918384e-05,
      "loss": 4.4809,
      "step": 1348
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.478771640560594e-05,
      "loss": 4.5929,
      "step": 1349
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.478153338829349e-05,
      "loss": 4.2954,
      "step": 1350
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4775350370981037e-05,
      "loss": 5.1978,
      "step": 1351
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.476916735366859e-05,
      "loss": 5.1481,
      "step": 1352
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4762984336356142e-05,
      "loss": 5.0041,
      "step": 1353
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4756801319043693e-05,
      "loss": 4.9601,
      "step": 1354
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4750618301731248e-05,
      "loss": 4.7854,
      "step": 1355
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4744435284418795e-05,
      "loss": 5.053,
      "step": 1356
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.473825226710635e-05,
      "loss": 4.688,
      "step": 1357
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.47320692497939e-05,
      "loss": 4.9782,
      "step": 1358
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4725886232481452e-05,
      "loss": 4.8964,
      "step": 1359
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4719703215169003e-05,
      "loss": 5.1071,
      "step": 1360
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4713520197856554e-05,
      "loss": 4.8559,
      "step": 1361
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.4707337180544105e-05,
      "loss": 4.7969,
      "step": 1362
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.470115416323166e-05,
      "loss": 4.7487,
      "step": 1363
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.469497114591921e-05,
      "loss": 5.1274,
      "step": 1364
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.468878812860676e-05,
      "loss": 4.7195,
      "step": 1365
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4682605111294312e-05,
      "loss": 4.7278,
      "step": 1366
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4676422093981863e-05,
      "loss": 4.7715,
      "step": 1367
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4670239076669417e-05,
      "loss": 4.6523,
      "step": 1368
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4664056059356965e-05,
      "loss": 4.9966,
      "step": 1369
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4657873042044516e-05,
      "loss": 5.0167,
      "step": 1370
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.465169002473207e-05,
      "loss": 4.9866,
      "step": 1371
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.464550700741962e-05,
      "loss": 4.6505,
      "step": 1372
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4639323990107176e-05,
      "loss": 4.9471,
      "step": 1373
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4633140972794723e-05,
      "loss": 5.0542,
      "step": 1374
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4626957955482274e-05,
      "loss": 4.7042,
      "step": 1375
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.462077493816983e-05,
      "loss": 4.5117,
      "step": 1376
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.461459192085738e-05,
      "loss": 4.7971,
      "step": 1377
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.460840890354493e-05,
      "loss": 4.8509,
      "step": 1378
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.460222588623248e-05,
      "loss": 4.7724,
      "step": 1379
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4596042868920033e-05,
      "loss": 4.6841,
      "step": 1380
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4589859851607587e-05,
      "loss": 4.538,
      "step": 1381
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.4583676834295138e-05,
      "loss": 4.8565,
      "step": 1382
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4577493816982685e-05,
      "loss": 4.7113,
      "step": 1383
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.457131079967024e-05,
      "loss": 4.8392,
      "step": 1384
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.456512778235779e-05,
      "loss": 4.8789,
      "step": 1385
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4558944765045345e-05,
      "loss": 4.764,
      "step": 1386
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4552761747732896e-05,
      "loss": 4.9042,
      "step": 1387
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4546578730420444e-05,
      "loss": 4.8071,
      "step": 1388
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4540395713107998e-05,
      "loss": 4.7972,
      "step": 1389
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.453421269579555e-05,
      "loss": 4.759,
      "step": 1390
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.45280296784831e-05,
      "loss": 4.5114,
      "step": 1391
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.452184666117065e-05,
      "loss": 4.6864,
      "step": 1392
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4515663643858202e-05,
      "loss": 4.6184,
      "step": 1393
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4509480626545756e-05,
      "loss": 4.8975,
      "step": 1394
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4503297609233307e-05,
      "loss": 4.8238,
      "step": 1395
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.449711459192086e-05,
      "loss": 4.7003,
      "step": 1396
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.449093157460841e-05,
      "loss": 4.6056,
      "step": 1397
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.448474855729596e-05,
      "loss": 4.3245,
      "step": 1398
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.447856553998351e-05,
      "loss": 4.5945,
      "step": 1399
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4472382522671066e-05,
      "loss": 4.5005,
      "step": 1400
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4466199505358613e-05,
      "loss": 4.8798,
      "step": 1401
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4460016488046168e-05,
      "loss": 5.0688,
      "step": 1402
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.445383347073372e-05,
      "loss": 4.8884,
      "step": 1403
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.444765045342127e-05,
      "loss": 4.7081,
      "step": 1404
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4441467436108824e-05,
      "loss": 4.7284,
      "step": 1405
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.443528441879637e-05,
      "loss": 5.0747,
      "step": 1406
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4429101401483923e-05,
      "loss": 4.8387,
      "step": 1407
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4422918384171477e-05,
      "loss": 4.8567,
      "step": 1408
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4416735366859028e-05,
      "loss": 4.5954,
      "step": 1409
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4410552349546582e-05,
      "loss": 4.6079,
      "step": 1410
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.440436933223413e-05,
      "loss": 4.6369,
      "step": 1411
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.439818631492168e-05,
      "loss": 4.7668,
      "step": 1412
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4392003297609235e-05,
      "loss": 5.0182,
      "step": 1413
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4385820280296786e-05,
      "loss": 4.7357,
      "step": 1414
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4379637262984337e-05,
      "loss": 4.9159,
      "step": 1415
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4373454245671888e-05,
      "loss": 4.9975,
      "step": 1416
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.436727122835944e-05,
      "loss": 4.9219,
      "step": 1417
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.4361088211046994e-05,
      "loss": 4.8235,
      "step": 1418
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4354905193734544e-05,
      "loss": 4.9094,
      "step": 1419
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4348722176422092e-05,
      "loss": 4.5628,
      "step": 1420
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4342539159109646e-05,
      "loss": 4.8473,
      "step": 1421
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4336356141797197e-05,
      "loss": 4.7773,
      "step": 1422
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4330173124484752e-05,
      "loss": 4.71,
      "step": 1423
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.43239901071723e-05,
      "loss": 4.7163,
      "step": 1424
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.431780708985985e-05,
      "loss": 4.7861,
      "step": 1425
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4311624072547405e-05,
      "loss": 4.9904,
      "step": 1426
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4305441055234956e-05,
      "loss": 4.7317,
      "step": 1427
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4299258037922507e-05,
      "loss": 4.6866,
      "step": 1428
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4293075020610058e-05,
      "loss": 4.9166,
      "step": 1429
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.428689200329761e-05,
      "loss": 4.9989,
      "step": 1430
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4280708985985163e-05,
      "loss": 4.7826,
      "step": 1431
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4274525968672714e-05,
      "loss": 4.9493,
      "step": 1432
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.426834295136026e-05,
      "loss": 5.0158,
      "step": 1433
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4262159934047816e-05,
      "loss": 4.4571,
      "step": 1434
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4255976916735367e-05,
      "loss": 4.7903,
      "step": 1435
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.4249793899422918e-05,
      "loss": 4.555,
      "step": 1436
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4243610882110472e-05,
      "loss": 4.9189,
      "step": 1437
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.423742786479802e-05,
      "loss": 4.6693,
      "step": 1438
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4231244847485574e-05,
      "loss": 4.7229,
      "step": 1439
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4225061830173125e-05,
      "loss": 4.7779,
      "step": 1440
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4218878812860676e-05,
      "loss": 4.8791,
      "step": 1441
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.421269579554823e-05,
      "loss": 4.664,
      "step": 1442
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4206512778235778e-05,
      "loss": 4.5477,
      "step": 1443
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4200329760923333e-05,
      "loss": 4.6318,
      "step": 1444
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4194146743610884e-05,
      "loss": 4.6738,
      "step": 1445
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4187963726298435e-05,
      "loss": 4.9982,
      "step": 1446
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4181780708985986e-05,
      "loss": 4.5457,
      "step": 1447
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4175597691673536e-05,
      "loss": 4.7237,
      "step": 1448
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4169414674361087e-05,
      "loss": 4.533,
      "step": 1449
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4163231657048642e-05,
      "loss": 4.587,
      "step": 1450
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4157048639736193e-05,
      "loss": 5.1416,
      "step": 1451
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4150865622423744e-05,
      "loss": 5.0879,
      "step": 1452
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.4144682605111295e-05,
      "loss": 4.7527,
      "step": 1453
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4138499587798846e-05,
      "loss": 4.7484,
      "step": 1454
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.41323165704864e-05,
      "loss": 4.7214,
      "step": 1455
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4126133553173948e-05,
      "loss": 4.7162,
      "step": 1456
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.41199505358615e-05,
      "loss": 5.0603,
      "step": 1457
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4113767518549053e-05,
      "loss": 4.9569,
      "step": 1458
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4107584501236604e-05,
      "loss": 4.7624,
      "step": 1459
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.410140148392416e-05,
      "loss": 4.7816,
      "step": 1460
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4095218466611706e-05,
      "loss": 4.7166,
      "step": 1461
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4089035449299257e-05,
      "loss": 4.8808,
      "step": 1462
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.408285243198681e-05,
      "loss": 4.8097,
      "step": 1463
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4076669414674362e-05,
      "loss": 4.677,
      "step": 1464
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.407048639736191e-05,
      "loss": 4.9412,
      "step": 1465
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4064303380049464e-05,
      "loss": 4.6712,
      "step": 1466
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4058120362737015e-05,
      "loss": 4.7571,
      "step": 1467
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.405193734542457e-05,
      "loss": 4.803,
      "step": 1468
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.404575432811212e-05,
      "loss": 4.455,
      "step": 1469
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4039571310799668e-05,
      "loss": 4.6601,
      "step": 1470
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.4033388293487223e-05,
      "loss": 4.7165,
      "step": 1471
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.4027205276174774e-05,
      "loss": 4.5463,
      "step": 1472
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.4021022258862328e-05,
      "loss": 4.7685,
      "step": 1473
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.401483924154988e-05,
      "loss": 4.4762,
      "step": 1474
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.4008656224237427e-05,
      "loss": 4.8695,
      "step": 1475
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.400247320692498e-05,
      "loss": 4.8232,
      "step": 1476
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.3996290189612532e-05,
      "loss": 4.8469,
      "step": 1477
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.3990107172300083e-05,
      "loss": 4.4963,
      "step": 1478
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.3983924154987634e-05,
      "loss": 4.6917,
      "step": 1479
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.3977741137675185e-05,
      "loss": 4.5017,
      "step": 1480
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.397155812036274e-05,
      "loss": 4.7468,
      "step": 1481
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.396537510305029e-05,
      "loss": 4.8579,
      "step": 1482
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.395919208573784e-05,
      "loss": 4.7744,
      "step": 1483
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.3953009068425392e-05,
      "loss": 4.3235,
      "step": 1484
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.3946826051112943e-05,
      "loss": 4.7669,
      "step": 1485
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.3940643033800494e-05,
      "loss": 4.8762,
      "step": 1486
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.393446001648805e-05,
      "loss": 4.8029,
      "step": 1487
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.3928276999175596e-05,
      "loss": 4.8069,
      "step": 1488
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.392209398186315e-05,
      "loss": 4.8446,
      "step": 1489
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.39159109645507e-05,
      "loss": 4.5826,
      "step": 1490
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3909727947238252e-05,
      "loss": 4.9273,
      "step": 1491
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3903544929925807e-05,
      "loss": 4.8525,
      "step": 1492
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3897361912613354e-05,
      "loss": 4.3329,
      "step": 1493
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3891178895300905e-05,
      "loss": 4.658,
      "step": 1494
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.388499587798846e-05,
      "loss": 4.6897,
      "step": 1495
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.387881286067601e-05,
      "loss": 4.9453,
      "step": 1496
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.387262984336356e-05,
      "loss": 4.5403,
      "step": 1497
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3866446826051113e-05,
      "loss": 4.767,
      "step": 1498
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3860263808738664e-05,
      "loss": 4.3193,
      "step": 1499
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3854080791426218e-05,
      "loss": 4.5587,
      "step": 1500
    },
    {
      "epoch": 0.84,
      "eval_loss": 4.668270587921143,
      "eval_runtime": 863.727,
      "eval_samples_per_second": 3.059,
      "eval_steps_per_second": 0.383,
      "eval_wer": 1.0009722800960374,
      "step": 1500
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.384789777411377e-05,
      "loss": 4.8488,
      "step": 1501
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.384171475680132e-05,
      "loss": 4.9989,
      "step": 1502
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.383553173948887e-05,
      "loss": 4.8546,
      "step": 1503
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3829348722176422e-05,
      "loss": 4.7779,
      "step": 1504
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3823165704863976e-05,
      "loss": 4.9206,
      "step": 1505
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3816982687551527e-05,
      "loss": 4.8392,
      "step": 1506
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.3810799670239075e-05,
      "loss": 4.7104,
      "step": 1507
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.380461665292663e-05,
      "loss": 4.7794,
      "step": 1508
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.379843363561418e-05,
      "loss": 4.8494,
      "step": 1509
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.3792250618301735e-05,
      "loss": 4.7781,
      "step": 1510
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.3786067600989282e-05,
      "loss": 4.7345,
      "step": 1511
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.3779884583676833e-05,
      "loss": 4.7829,
      "step": 1512
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.3773701566364388e-05,
      "loss": 4.5876,
      "step": 1513
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.376751854905194e-05,
      "loss": 4.6528,
      "step": 1514
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.376133553173949e-05,
      "loss": 4.6114,
      "step": 1515
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.375515251442704e-05,
      "loss": 4.8849,
      "step": 1516
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.374896949711459e-05,
      "loss": 4.5987,
      "step": 1517
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.3742786479802146e-05,
      "loss": 4.7255,
      "step": 1518
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.3736603462489697e-05,
      "loss": 4.8888,
      "step": 1519
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.3730420445177244e-05,
      "loss": 4.6803,
      "step": 1520
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.37242374278648e-05,
      "loss": 4.7039,
      "step": 1521
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.371805441055235e-05,
      "loss": 4.6603,
      "step": 1522
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.37118713932399e-05,
      "loss": 4.6987,
      "step": 1523
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.3705688375927455e-05,
      "loss": 4.7882,
      "step": 1524
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.3699505358615003e-05,
      "loss": 4.673,
      "step": 1525
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3693322341302557e-05,
      "loss": 4.5652,
      "step": 1526
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3687139323990108e-05,
      "loss": 4.7334,
      "step": 1527
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.368095630667766e-05,
      "loss": 4.5084,
      "step": 1528
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3674773289365213e-05,
      "loss": 4.4052,
      "step": 1529
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.366859027205276e-05,
      "loss": 4.4811,
      "step": 1530
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3662407254740312e-05,
      "loss": 4.6007,
      "step": 1531
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3656224237427866e-05,
      "loss": 4.3526,
      "step": 1532
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3650041220115417e-05,
      "loss": 4.6056,
      "step": 1533
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3643858202802968e-05,
      "loss": 4.5659,
      "step": 1534
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.363767518549052e-05,
      "loss": 4.7603,
      "step": 1535
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.363149216817807e-05,
      "loss": 4.465,
      "step": 1536
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3625309150865625e-05,
      "loss": 4.6597,
      "step": 1537
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3619126133553176e-05,
      "loss": 4.872,
      "step": 1538
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3612943116240727e-05,
      "loss": 4.6756,
      "step": 1539
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3606760098928278e-05,
      "loss": 4.7388,
      "step": 1540
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.360057708161583e-05,
      "loss": 4.7221,
      "step": 1541
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.3594394064303383e-05,
      "loss": 4.4485,
      "step": 1542
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.358821104699093e-05,
      "loss": 4.6986,
      "step": 1543
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.358202802967848e-05,
      "loss": 4.494,
      "step": 1544
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3575845012366036e-05,
      "loss": 4.8649,
      "step": 1545
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3569661995053587e-05,
      "loss": 4.535,
      "step": 1546
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.356347897774114e-05,
      "loss": 4.7712,
      "step": 1547
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.355729596042869e-05,
      "loss": 4.5912,
      "step": 1548
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.355111294311624e-05,
      "loss": 4.649,
      "step": 1549
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3544929925803794e-05,
      "loss": 4.1925,
      "step": 1550
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3538746908491345e-05,
      "loss": 5.0726,
      "step": 1551
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3532563891178893e-05,
      "loss": 4.9803,
      "step": 1552
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3526380873866447e-05,
      "loss": 4.7898,
      "step": 1553
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3520197856553998e-05,
      "loss": 4.5597,
      "step": 1554
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3514014839241552e-05,
      "loss": 4.9501,
      "step": 1555
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3507831821929103e-05,
      "loss": 5.0084,
      "step": 1556
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.350164880461665e-05,
      "loss": 4.866,
      "step": 1557
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3495465787304205e-05,
      "loss": 4.9246,
      "step": 1558
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3489282769991756e-05,
      "loss": 4.5252,
      "step": 1559
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.3483099752679307e-05,
      "loss": 4.8291,
      "step": 1560
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3476916735366862e-05,
      "loss": 4.7418,
      "step": 1561
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.347073371805441e-05,
      "loss": 4.7578,
      "step": 1562
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3464550700741964e-05,
      "loss": 4.8378,
      "step": 1563
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3458367683429515e-05,
      "loss": 4.5411,
      "step": 1564
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3452184666117066e-05,
      "loss": 4.7649,
      "step": 1565
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3446001648804617e-05,
      "loss": 4.5088,
      "step": 1566
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3439818631492168e-05,
      "loss": 4.4327,
      "step": 1567
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3433635614179722e-05,
      "loss": 4.7823,
      "step": 1568
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3427452596867273e-05,
      "loss": 4.8724,
      "step": 1569
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3421269579554824e-05,
      "loss": 4.7814,
      "step": 1570
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3415086562242375e-05,
      "loss": 4.6986,
      "step": 1571
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3408903544929926e-05,
      "loss": 4.8992,
      "step": 1572
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3402720527617477e-05,
      "loss": 4.7988,
      "step": 1573
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.339653751030503e-05,
      "loss": 4.4738,
      "step": 1574
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.339035449299258e-05,
      "loss": 4.3317,
      "step": 1575
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3384171475680133e-05,
      "loss": 4.5275,
      "step": 1576
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3377988458367684e-05,
      "loss": 4.5689,
      "step": 1577
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3371805441055235e-05,
      "loss": 4.831,
      "step": 1578
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.336562242374279e-05,
      "loss": 4.7678,
      "step": 1579
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3359439406430337e-05,
      "loss": 4.721,
      "step": 1580
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3353256389117888e-05,
      "loss": 4.6628,
      "step": 1581
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3347073371805442e-05,
      "loss": 4.7193,
      "step": 1582
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3340890354492993e-05,
      "loss": 4.6996,
      "step": 1583
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3334707337180544e-05,
      "loss": 4.6246,
      "step": 1584
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3328524319868095e-05,
      "loss": 4.6055,
      "step": 1585
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3322341302555646e-05,
      "loss": 4.6672,
      "step": 1586
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.33161582852432e-05,
      "loss": 4.6408,
      "step": 1587
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3309975267930752e-05,
      "loss": 4.789,
      "step": 1588
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.33037922506183e-05,
      "loss": 4.5487,
      "step": 1589
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3297609233305854e-05,
      "loss": 4.5831,
      "step": 1590
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3291426215993405e-05,
      "loss": 4.6375,
      "step": 1591
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.328524319868096e-05,
      "loss": 4.709,
      "step": 1592
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.327906018136851e-05,
      "loss": 4.5326,
      "step": 1593
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3272877164056058e-05,
      "loss": 4.5893,
      "step": 1594
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3266694146743612e-05,
      "loss": 4.6113,
      "step": 1595
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3260511129431163e-05,
      "loss": 4.5519,
      "step": 1596
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3254328112118717e-05,
      "loss": 4.7955,
      "step": 1597
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3248145094806265e-05,
      "loss": 4.483,
      "step": 1598
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3241962077493816e-05,
      "loss": 5.2397,
      "step": 1599
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.323577906018137e-05,
      "loss": 4.3965,
      "step": 1600
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.322959604286892e-05,
      "loss": 5.0008,
      "step": 1601
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3223413025556472e-05,
      "loss": 4.6379,
      "step": 1602
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3217230008244023e-05,
      "loss": 4.9873,
      "step": 1603
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3211046990931574e-05,
      "loss": 4.881,
      "step": 1604
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.320486397361913e-05,
      "loss": 4.6124,
      "step": 1605
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.319868095630668e-05,
      "loss": 4.8099,
      "step": 1606
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3192497938994227e-05,
      "loss": 4.6437,
      "step": 1607
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.318631492168178e-05,
      "loss": 4.5221,
      "step": 1608
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3180131904369333e-05,
      "loss": 4.9206,
      "step": 1609
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3173948887056884e-05,
      "loss": 4.5835,
      "step": 1610
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3167765869744438e-05,
      "loss": 4.8983,
      "step": 1611
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3161582852431985e-05,
      "loss": 4.7797,
      "step": 1612
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.315539983511954e-05,
      "loss": 4.759,
      "step": 1613
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.314921681780709e-05,
      "loss": 4.5978,
      "step": 1614
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3143033800494642e-05,
      "loss": 4.6374,
      "step": 1615
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3136850783182196e-05,
      "loss": 4.563,
      "step": 1616
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3130667765869744e-05,
      "loss": 4.7365,
      "step": 1617
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3124484748557295e-05,
      "loss": 4.6345,
      "step": 1618
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.311830173124485e-05,
      "loss": 4.7943,
      "step": 1619
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.31121187139324e-05,
      "loss": 4.6594,
      "step": 1620
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.310593569661995e-05,
      "loss": 4.7841,
      "step": 1621
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3099752679307502e-05,
      "loss": 4.4981,
      "step": 1622
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3093569661995053e-05,
      "loss": 4.6954,
      "step": 1623
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3087386644682607e-05,
      "loss": 4.4016,
      "step": 1624
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.308120362737016e-05,
      "loss": 4.7222,
      "step": 1625
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.307502061005771e-05,
      "loss": 4.8451,
      "step": 1626
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.306883759274526e-05,
      "loss": 4.7704,
      "step": 1627
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.306265457543281e-05,
      "loss": 4.5371,
      "step": 1628
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3056471558120366e-05,
      "loss": 4.7543,
      "step": 1629
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3050288540807913e-05,
      "loss": 4.4997,
      "step": 1630
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.3044105523495464e-05,
      "loss": 4.6166,
      "step": 1631
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.303792250618302e-05,
      "loss": 4.8576,
      "step": 1632
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.303173948887057e-05,
      "loss": 4.6142,
      "step": 1633
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.3025556471558124e-05,
      "loss": 4.7469,
      "step": 1634
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.301937345424567e-05,
      "loss": 4.7987,
      "step": 1635
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.3013190436933223e-05,
      "loss": 4.5517,
      "step": 1636
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.3007007419620777e-05,
      "loss": 4.5337,
      "step": 1637
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.3000824402308328e-05,
      "loss": 4.7221,
      "step": 1638
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.2994641384995876e-05,
      "loss": 4.4326,
      "step": 1639
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.298845836768343e-05,
      "loss": 4.5634,
      "step": 1640
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.298227535037098e-05,
      "loss": 4.6779,
      "step": 1641
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.2976092333058535e-05,
      "loss": 4.5122,
      "step": 1642
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.2969909315746086e-05,
      "loss": 4.5227,
      "step": 1643
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.2963726298433634e-05,
      "loss": 4.695,
      "step": 1644
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.2957543281121188e-05,
      "loss": 4.7414,
      "step": 1645
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.295136026380874e-05,
      "loss": 4.9663,
      "step": 1646
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.294517724649629e-05,
      "loss": 4.176,
      "step": 1647
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.2938994229183844e-05,
      "loss": 4.7841,
      "step": 1648
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.2932811211871392e-05,
      "loss": 4.5365,
      "step": 1649
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.2926628194558946e-05,
      "loss": 4.1164,
      "step": 1650
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2920445177246497e-05,
      "loss": 4.9664,
      "step": 1651
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.291426215993405e-05,
      "loss": 4.9178,
      "step": 1652
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.29080791426216e-05,
      "loss": 4.9474,
      "step": 1653
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.290189612530915e-05,
      "loss": 4.7277,
      "step": 1654
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.28957131079967e-05,
      "loss": 4.5904,
      "step": 1655
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2889530090684256e-05,
      "loss": 4.7561,
      "step": 1656
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2883347073371807e-05,
      "loss": 4.7501,
      "step": 1657
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2877164056059358e-05,
      "loss": 4.7101,
      "step": 1658
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.287098103874691e-05,
      "loss": 4.5749,
      "step": 1659
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.286479802143446e-05,
      "loss": 4.5874,
      "step": 1660
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2858615004122014e-05,
      "loss": 4.6806,
      "step": 1661
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.285243198680956e-05,
      "loss": 4.7616,
      "step": 1662
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2846248969497116e-05,
      "loss": 4.4764,
      "step": 1663
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2840065952184667e-05,
      "loss": 4.8268,
      "step": 1664
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2833882934872218e-05,
      "loss": 4.5406,
      "step": 1665
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2827699917559772e-05,
      "loss": 4.6825,
      "step": 1666
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.282151690024732e-05,
      "loss": 4.6201,
      "step": 1667
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.281533388293487e-05,
      "loss": 4.499,
      "step": 1668
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2809150865622425e-05,
      "loss": 4.541,
      "step": 1669
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2802967848309976e-05,
      "loss": 4.6134,
      "step": 1670
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2796784830997527e-05,
      "loss": 4.6113,
      "step": 1671
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2790601813685078e-05,
      "loss": 4.6814,
      "step": 1672
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.278441879637263e-05,
      "loss": 4.7349,
      "step": 1673
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2778235779060184e-05,
      "loss": 4.7072,
      "step": 1674
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2772052761747735e-05,
      "loss": 5.092,
      "step": 1675
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2765869744435282e-05,
      "loss": 4.531,
      "step": 1676
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2759686727122836e-05,
      "loss": 4.4875,
      "step": 1677
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2753503709810387e-05,
      "loss": 4.7123,
      "step": 1678
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2747320692497942e-05,
      "loss": 4.8618,
      "step": 1679
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2741137675185493e-05,
      "loss": 4.7661,
      "step": 1680
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.273495465787304e-05,
      "loss": 4.3776,
      "step": 1681
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2728771640560595e-05,
      "loss": 4.3227,
      "step": 1682
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2722588623248146e-05,
      "loss": 4.7205,
      "step": 1683
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2716405605935697e-05,
      "loss": 4.4805,
      "step": 1684
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2710222588623248e-05,
      "loss": 4.7289,
      "step": 1685
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.27040395713108e-05,
      "loss": 4.6208,
      "step": 1686
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2697856553998353e-05,
      "loss": 4.5907,
      "step": 1687
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2691673536685904e-05,
      "loss": 4.6922,
      "step": 1688
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2685490519373455e-05,
      "loss": 4.7468,
      "step": 1689
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2679307502061006e-05,
      "loss": 4.457,
      "step": 1690
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2673124484748557e-05,
      "loss": 4.7973,
      "step": 1691
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.266694146743611e-05,
      "loss": 4.5601,
      "step": 1692
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2660758450123662e-05,
      "loss": 4.8681,
      "step": 1693
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.265457543281121e-05,
      "loss": 4.7352,
      "step": 1694
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2648392415498764e-05,
      "loss": 4.7535,
      "step": 1695
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2642209398186315e-05,
      "loss": 4.7696,
      "step": 1696
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2636026380873866e-05,
      "loss": 4.5193,
      "step": 1697
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.262984336356142e-05,
      "loss": 4.8257,
      "step": 1698
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2623660346248968e-05,
      "loss": 4.6848,
      "step": 1699
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2617477328936523e-05,
      "loss": 4.5816,
      "step": 1700
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2611294311624074e-05,
      "loss": 5.0481,
      "step": 1701
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2605111294311625e-05,
      "loss": 4.7169,
      "step": 1702
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.2598928276999176e-05,
      "loss": 4.8832,
      "step": 1703
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2592745259686727e-05,
      "loss": 4.7516,
      "step": 1704
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2586562242374278e-05,
      "loss": 4.6476,
      "step": 1705
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2580379225061832e-05,
      "loss": 4.4988,
      "step": 1706
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2574196207749383e-05,
      "loss": 4.5824,
      "step": 1707
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2568013190436934e-05,
      "loss": 4.8198,
      "step": 1708
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2561830173124485e-05,
      "loss": 4.8639,
      "step": 1709
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2555647155812036e-05,
      "loss": 4.8774,
      "step": 1710
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.254946413849959e-05,
      "loss": 4.4714,
      "step": 1711
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.254328112118714e-05,
      "loss": 4.8206,
      "step": 1712
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.253709810387469e-05,
      "loss": 4.9672,
      "step": 1713
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2530915086562243e-05,
      "loss": 4.7649,
      "step": 1714
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2524732069249794e-05,
      "loss": 4.6498,
      "step": 1715
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.251854905193735e-05,
      "loss": 4.852,
      "step": 1716
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2512366034624896e-05,
      "loss": 4.9669,
      "step": 1717
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2506183017312447e-05,
      "loss": 4.7852,
      "step": 1718
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.25e-05,
      "loss": 4.4806,
      "step": 1719
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2493816982687552e-05,
      "loss": 4.762,
      "step": 1720
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.2487633965375107e-05,
      "loss": 4.6568,
      "step": 1721
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2481450948062654e-05,
      "loss": 4.5291,
      "step": 1722
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2475267930750205e-05,
      "loss": 4.7789,
      "step": 1723
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.246908491343776e-05,
      "loss": 4.7528,
      "step": 1724
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.246290189612531e-05,
      "loss": 4.7447,
      "step": 1725
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2456718878812858e-05,
      "loss": 4.4643,
      "step": 1726
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2450535861500413e-05,
      "loss": 4.6454,
      "step": 1727
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2444352844187964e-05,
      "loss": 4.8605,
      "step": 1728
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2438169826875518e-05,
      "loss": 4.9224,
      "step": 1729
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.243198680956307e-05,
      "loss": 4.6734,
      "step": 1730
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2425803792250617e-05,
      "loss": 4.3627,
      "step": 1731
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.241962077493817e-05,
      "loss": 4.439,
      "step": 1732
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2413437757625722e-05,
      "loss": 4.5968,
      "step": 1733
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2407254740313273e-05,
      "loss": 4.7098,
      "step": 1734
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2401071723000827e-05,
      "loss": 4.971,
      "step": 1735
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.2394888705688375e-05,
      "loss": 5.0504,
      "step": 1736
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.238870568837593e-05,
      "loss": 4.3707,
      "step": 1737
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.238252267106348e-05,
      "loss": 4.5412,
      "step": 1738
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.237633965375103e-05,
      "loss": 4.6412,
      "step": 1739
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2370156636438582e-05,
      "loss": 4.531,
      "step": 1740
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2363973619126133e-05,
      "loss": 4.4923,
      "step": 1741
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2357790601813684e-05,
      "loss": 4.5263,
      "step": 1742
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.235160758450124e-05,
      "loss": 4.4681,
      "step": 1743
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.234542456718879e-05,
      "loss": 4.4725,
      "step": 1744
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.233924154987634e-05,
      "loss": 4.6295,
      "step": 1745
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.233305853256389e-05,
      "loss": 4.5744,
      "step": 1746
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2326875515251442e-05,
      "loss": 4.6306,
      "step": 1747
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2320692497938997e-05,
      "loss": 4.0973,
      "step": 1748
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2314509480626544e-05,
      "loss": 4.3661,
      "step": 1749
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.23083264633141e-05,
      "loss": 4.2387,
      "step": 1750
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.230214344600165e-05,
      "loss": 4.5181,
      "step": 1751
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.22959604286892e-05,
      "loss": 4.6063,
      "step": 1752
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2289777411376755e-05,
      "loss": 4.8364,
      "step": 1753
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2283594394064303e-05,
      "loss": 4.6705,
      "step": 1754
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2277411376751854e-05,
      "loss": 4.844,
      "step": 1755
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.2271228359439408e-05,
      "loss": 4.5021,
      "step": 1756
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.226504534212696e-05,
      "loss": 4.4937,
      "step": 1757
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.225886232481451e-05,
      "loss": 4.5586,
      "step": 1758
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.225267930750206e-05,
      "loss": 4.7366,
      "step": 1759
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2246496290189612e-05,
      "loss": 4.6416,
      "step": 1760
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2240313272877166e-05,
      "loss": 4.6712,
      "step": 1761
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2234130255564717e-05,
      "loss": 4.44,
      "step": 1762
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2227947238252265e-05,
      "loss": 4.6371,
      "step": 1763
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.222176422093982e-05,
      "loss": 4.7541,
      "step": 1764
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.221558120362737e-05,
      "loss": 4.5393,
      "step": 1765
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2209398186314925e-05,
      "loss": 4.649,
      "step": 1766
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2203215169002476e-05,
      "loss": 4.5222,
      "step": 1767
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2197032151690023e-05,
      "loss": 4.483,
      "step": 1768
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2190849134377578e-05,
      "loss": 4.6009,
      "step": 1769
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.218466611706513e-05,
      "loss": 4.504,
      "step": 1770
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.217848309975268e-05,
      "loss": 4.6326,
      "step": 1771
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.217230008244023e-05,
      "loss": 4.4596,
      "step": 1772
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.216611706512778e-05,
      "loss": 4.6121,
      "step": 1773
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2159934047815336e-05,
      "loss": 4.6591,
      "step": 1774
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2153751030502887e-05,
      "loss": 4.4645,
      "step": 1775
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2147568013190438e-05,
      "loss": 4.6594,
      "step": 1776
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.214138499587799e-05,
      "loss": 4.5219,
      "step": 1777
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.213520197856554e-05,
      "loss": 4.4817,
      "step": 1778
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.212901896125309e-05,
      "loss": 4.7056,
      "step": 1779
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2122835943940645e-05,
      "loss": 4.4653,
      "step": 1780
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2116652926628193e-05,
      "loss": 4.4864,
      "step": 1781
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2110469909315747e-05,
      "loss": 4.2931,
      "step": 1782
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2104286892003298e-05,
      "loss": 4.2728,
      "step": 1783
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.209810387469085e-05,
      "loss": 3.8453,
      "step": 1784
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2091920857378403e-05,
      "loss": 4.8458,
      "step": 1785
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.208573784006595e-05,
      "loss": 4.8343,
      "step": 1786
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2079554822753505e-05,
      "loss": 4.4589,
      "step": 1787
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2073371805441056e-05,
      "loss": 4.4634,
      "step": 1788
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2067188788128607e-05,
      "loss": 4.5811,
      "step": 1789
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.206100577081616e-05,
      "loss": 4.6555,
      "step": 1790
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.205482275350371e-05,
      "loss": 4.4671,
      "step": 1791
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.204863973619126e-05,
      "loss": 4.3689,
      "step": 1792
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.2042456718878815e-05,
      "loss": 4.4445,
      "step": 1793
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.2036273701566366e-05,
      "loss": 4.3695,
      "step": 1794
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.2030090684253917e-05,
      "loss": 4.5354,
      "step": 1795
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.2023907666941468e-05,
      "loss": 4.4022,
      "step": 1796
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.201772464962902e-05,
      "loss": 4.1671,
      "step": 1797
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.2011541632316573e-05,
      "loss": 4.6527,
      "step": 1798
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.2005358615004124e-05,
      "loss": 4.4746,
      "step": 1799
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.199917559769167e-05,
      "loss": 4.7566,
      "step": 1800
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.1992992580379226e-05,
      "loss": 4.5136,
      "step": 1801
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.1986809563066777e-05,
      "loss": 4.3307,
      "step": 1802
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.198062654575433e-05,
      "loss": 4.6024,
      "step": 1803
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.197444352844188e-05,
      "loss": 4.5043,
      "step": 1804
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.196826051112943e-05,
      "loss": 4.146,
      "step": 1805
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.1962077493816984e-05,
      "loss": 4.5331,
      "step": 1806
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.1955894476504535e-05,
      "loss": 4.0281,
      "step": 1807
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.1949711459192086e-05,
      "loss": 4.2893,
      "step": 1808
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.1943528441879637e-05,
      "loss": 4.4707,
      "step": 1809
    },
    {
      "epoch": 1.01,
      "learning_rate": 2.1937345424567188e-05,
      "loss": 4.4489,
      "step": 1810
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1931162407254742e-05,
      "loss": 4.4655,
      "step": 1811
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1924979389942293e-05,
      "loss": 4.6654,
      "step": 1812
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.191879637262984e-05,
      "loss": 4.4515,
      "step": 1813
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1912613355317395e-05,
      "loss": 4.0564,
      "step": 1814
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1906430338004946e-05,
      "loss": 4.4983,
      "step": 1815
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.19002473206925e-05,
      "loss": 4.6772,
      "step": 1816
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1894064303380052e-05,
      "loss": 4.4712,
      "step": 1817
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.18878812860676e-05,
      "loss": 4.2538,
      "step": 1818
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1881698268755154e-05,
      "loss": 4.4698,
      "step": 1819
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1875515251442705e-05,
      "loss": 4.5051,
      "step": 1820
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1869332234130256e-05,
      "loss": 4.5214,
      "step": 1821
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1863149216817807e-05,
      "loss": 4.4139,
      "step": 1822
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1856966199505358e-05,
      "loss": 4.3752,
      "step": 1823
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1850783182192912e-05,
      "loss": 4.4445,
      "step": 1824
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1844600164880463e-05,
      "loss": 4.3,
      "step": 1825
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1838417147568014e-05,
      "loss": 4.5263,
      "step": 1826
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1832234130255565e-05,
      "loss": 4.1775,
      "step": 1827
    },
    {
      "epoch": 1.02,
      "learning_rate": 2.1826051112943116e-05,
      "loss": 4.5575,
      "step": 1828
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1819868095630667e-05,
      "loss": 4.4384,
      "step": 1829
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.181368507831822e-05,
      "loss": 4.1125,
      "step": 1830
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1807502061005772e-05,
      "loss": 3.9479,
      "step": 1831
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1801319043693323e-05,
      "loss": 4.2675,
      "step": 1832
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1795136026380874e-05,
      "loss": 4.1815,
      "step": 1833
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1788953009068425e-05,
      "loss": 4.1436,
      "step": 1834
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.178276999175598e-05,
      "loss": 4.6732,
      "step": 1835
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1776586974443527e-05,
      "loss": 4.7953,
      "step": 1836
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1770403957131078e-05,
      "loss": 4.5053,
      "step": 1837
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1764220939818633e-05,
      "loss": 4.5942,
      "step": 1838
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1758037922506184e-05,
      "loss": 4.37,
      "step": 1839
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1751854905193738e-05,
      "loss": 4.5023,
      "step": 1840
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1745671887881285e-05,
      "loss": 4.4662,
      "step": 1841
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1739488870568836e-05,
      "loss": 4.4596,
      "step": 1842
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.173330585325639e-05,
      "loss": 4.7552,
      "step": 1843
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1727122835943942e-05,
      "loss": 4.8117,
      "step": 1844
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1720939818631493e-05,
      "loss": 4.5789,
      "step": 1845
    },
    {
      "epoch": 1.03,
      "learning_rate": 2.1714756801319044e-05,
      "loss": 4.0922,
      "step": 1846
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.1708573784006595e-05,
      "loss": 4.2885,
      "step": 1847
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.170239076669415e-05,
      "loss": 4.3748,
      "step": 1848
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.16962077493817e-05,
      "loss": 4.3391,
      "step": 1849
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.1690024732069248e-05,
      "loss": 4.403,
      "step": 1850
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.1683841714756802e-05,
      "loss": 4.3273,
      "step": 1851
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.1677658697444353e-05,
      "loss": 4.6061,
      "step": 1852
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.1671475680131907e-05,
      "loss": 4.4,
      "step": 1853
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.166529266281946e-05,
      "loss": 4.1622,
      "step": 1854
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.1659109645507006e-05,
      "loss": 4.4997,
      "step": 1855
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.165292662819456e-05,
      "loss": 4.3808,
      "step": 1856
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.164674361088211e-05,
      "loss": 4.4899,
      "step": 1857
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.1640560593569662e-05,
      "loss": 4.3855,
      "step": 1858
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.1634377576257213e-05,
      "loss": 4.2218,
      "step": 1859
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.1628194558944764e-05,
      "loss": 4.3027,
      "step": 1860
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.162201154163232e-05,
      "loss": 4.1972,
      "step": 1861
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.161582852431987e-05,
      "loss": 4.4353,
      "step": 1862
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.160964550700742e-05,
      "loss": 4.4622,
      "step": 1863
    },
    {
      "epoch": 1.04,
      "learning_rate": 2.160346248969497e-05,
      "loss": 4.3209,
      "step": 1864
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1597279472382523e-05,
      "loss": 4.4438,
      "step": 1865
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1591096455070074e-05,
      "loss": 4.1106,
      "step": 1866
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1584913437757628e-05,
      "loss": 4.4813,
      "step": 1867
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1578730420445176e-05,
      "loss": 4.4238,
      "step": 1868
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.157254740313273e-05,
      "loss": 4.3265,
      "step": 1869
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.156636438582028e-05,
      "loss": 4.3626,
      "step": 1870
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1560181368507832e-05,
      "loss": 4.4318,
      "step": 1871
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1553998351195386e-05,
      "loss": 4.5,
      "step": 1872
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1547815333882934e-05,
      "loss": 4.3244,
      "step": 1873
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1541632316570488e-05,
      "loss": 4.3205,
      "step": 1874
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.153544929925804e-05,
      "loss": 4.0583,
      "step": 1875
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.152926628194559e-05,
      "loss": 4.4819,
      "step": 1876
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.152308326463314e-05,
      "loss": 4.3454,
      "step": 1877
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1516900247320692e-05,
      "loss": 4.0743,
      "step": 1878
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1510717230008243e-05,
      "loss": 4.6224,
      "step": 1879
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.1504534212695797e-05,
      "loss": 3.8237,
      "step": 1880
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.149835119538335e-05,
      "loss": 3.985,
      "step": 1881
    },
    {
      "epoch": 1.05,
      "learning_rate": 2.14921681780709e-05,
      "loss": 4.5102,
      "step": 1882
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.148598516075845e-05,
      "loss": 3.5272,
      "step": 1883
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1479802143446e-05,
      "loss": 3.9952,
      "step": 1884
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1473619126133556e-05,
      "loss": 4.4856,
      "step": 1885
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1467436108821107e-05,
      "loss": 4.4222,
      "step": 1886
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1461253091508654e-05,
      "loss": 4.8003,
      "step": 1887
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.145507007419621e-05,
      "loss": 4.5662,
      "step": 1888
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.144888705688376e-05,
      "loss": 4.3794,
      "step": 1889
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1442704039571314e-05,
      "loss": 4.7029,
      "step": 1890
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.143652102225886e-05,
      "loss": 4.3315,
      "step": 1891
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1430338004946413e-05,
      "loss": 4.6955,
      "step": 1892
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1424154987633967e-05,
      "loss": 4.3574,
      "step": 1893
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1417971970321518e-05,
      "loss": 4.2843,
      "step": 1894
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.141178895300907e-05,
      "loss": 4.3561,
      "step": 1895
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.140560593569662e-05,
      "loss": 4.4317,
      "step": 1896
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.139942291838417e-05,
      "loss": 4.3742,
      "step": 1897
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1393239901071725e-05,
      "loss": 4.6929,
      "step": 1898
    },
    {
      "epoch": 1.06,
      "learning_rate": 2.1387056883759276e-05,
      "loss": 4.3073,
      "step": 1899
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1380873866446824e-05,
      "loss": 4.3166,
      "step": 1900
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1374690849134378e-05,
      "loss": 4.4787,
      "step": 1901
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.136850783182193e-05,
      "loss": 4.4804,
      "step": 1902
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.136232481450948e-05,
      "loss": 4.5457,
      "step": 1903
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1356141797197035e-05,
      "loss": 4.5971,
      "step": 1904
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1349958779884582e-05,
      "loss": 4.6732,
      "step": 1905
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1343775762572137e-05,
      "loss": 4.446,
      "step": 1906
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1337592745259687e-05,
      "loss": 4.3136,
      "step": 1907
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.133140972794724e-05,
      "loss": 4.6181,
      "step": 1908
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.132522671063479e-05,
      "loss": 4.3991,
      "step": 1909
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.131904369332234e-05,
      "loss": 4.2816,
      "step": 1910
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1312860676009895e-05,
      "loss": 4.65,
      "step": 1911
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1306677658697446e-05,
      "loss": 4.4371,
      "step": 1912
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1300494641384997e-05,
      "loss": 4.1686,
      "step": 1913
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1294311624072548e-05,
      "loss": 4.3151,
      "step": 1914
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.12881286067601e-05,
      "loss": 4.5103,
      "step": 1915
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.128194558944765e-05,
      "loss": 4.3034,
      "step": 1916
    },
    {
      "epoch": 1.07,
      "learning_rate": 2.1275762572135204e-05,
      "loss": 4.4243,
      "step": 1917
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1269579554822755e-05,
      "loss": 4.6143,
      "step": 1918
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1263396537510306e-05,
      "loss": 4.3572,
      "step": 1919
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1257213520197857e-05,
      "loss": 4.0835,
      "step": 1920
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1251030502885408e-05,
      "loss": 4.253,
      "step": 1921
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1244847485572962e-05,
      "loss": 4.3946,
      "step": 1922
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.123866446826051e-05,
      "loss": 4.5295,
      "step": 1923
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.123248145094806e-05,
      "loss": 4.4894,
      "step": 1924
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1226298433635615e-05,
      "loss": 4.7448,
      "step": 1925
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1220115416323166e-05,
      "loss": 4.118,
      "step": 1926
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.121393239901072e-05,
      "loss": 4.0882,
      "step": 1927
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1207749381698268e-05,
      "loss": 4.2674,
      "step": 1928
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.120156636438582e-05,
      "loss": 4.1833,
      "step": 1929
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1195383347073374e-05,
      "loss": 3.8618,
      "step": 1930
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1189200329760925e-05,
      "loss": 3.5722,
      "step": 1931
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1183017312448472e-05,
      "loss": 4.5285,
      "step": 1932
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1176834295136027e-05,
      "loss": 3.7739,
      "step": 1933
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1170651277823578e-05,
      "loss": 3.8274,
      "step": 1934
    },
    {
      "epoch": 1.08,
      "learning_rate": 2.1164468260511132e-05,
      "loss": 4.6732,
      "step": 1935
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1158285243198683e-05,
      "loss": 4.5084,
      "step": 1936
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.115210222588623e-05,
      "loss": 4.7573,
      "step": 1937
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1145919208573785e-05,
      "loss": 4.5898,
      "step": 1938
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1139736191261336e-05,
      "loss": 4.5427,
      "step": 1939
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.113355317394889e-05,
      "loss": 4.6447,
      "step": 1940
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.112737015663644e-05,
      "loss": 4.3651,
      "step": 1941
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.112118713932399e-05,
      "loss": 4.3546,
      "step": 1942
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1115004122011543e-05,
      "loss": 4.4076,
      "step": 1943
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1108821104699094e-05,
      "loss": 4.5207,
      "step": 1944
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1102638087386645e-05,
      "loss": 4.2753,
      "step": 1945
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1096455070074196e-05,
      "loss": 4.2886,
      "step": 1946
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1090272052761747e-05,
      "loss": 4.5845,
      "step": 1947
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.10840890354493e-05,
      "loss": 4.5322,
      "step": 1948
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1077906018136852e-05,
      "loss": 4.5415,
      "step": 1949
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1071723000824403e-05,
      "loss": 4.3487,
      "step": 1950
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1065539983511954e-05,
      "loss": 4.6327,
      "step": 1951
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1059356966199505e-05,
      "loss": 4.1435,
      "step": 1952
    },
    {
      "epoch": 1.09,
      "learning_rate": 2.1053173948887056e-05,
      "loss": 4.4636,
      "step": 1953
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.104699093157461e-05,
      "loss": 4.3668,
      "step": 1954
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.1040807914262158e-05,
      "loss": 4.5378,
      "step": 1955
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.1034624896949713e-05,
      "loss": 4.3728,
      "step": 1956
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.1028441879637264e-05,
      "loss": 4.3061,
      "step": 1957
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.1022258862324815e-05,
      "loss": 4.5718,
      "step": 1958
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.101607584501237e-05,
      "loss": 4.4549,
      "step": 1959
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.1009892827699917e-05,
      "loss": 4.2299,
      "step": 1960
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.1003709810387468e-05,
      "loss": 4.1922,
      "step": 1961
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.0997526793075022e-05,
      "loss": 4.1164,
      "step": 1962
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.0991343775762573e-05,
      "loss": 4.2168,
      "step": 1963
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.0985160758450124e-05,
      "loss": 4.4659,
      "step": 1964
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.0978977741137675e-05,
      "loss": 4.3387,
      "step": 1965
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.0972794723825226e-05,
      "loss": 4.253,
      "step": 1966
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.096661170651278e-05,
      "loss": 4.2611,
      "step": 1967
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.096042868920033e-05,
      "loss": 4.0644,
      "step": 1968
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.0954245671887882e-05,
      "loss": 4.5114,
      "step": 1969
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.0948062654575433e-05,
      "loss": 4.3504,
      "step": 1970
    },
    {
      "epoch": 1.1,
      "learning_rate": 2.0941879637262984e-05,
      "loss": 4.2167,
      "step": 1971
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.093569661995054e-05,
      "loss": 4.2439,
      "step": 1972
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.092951360263809e-05,
      "loss": 4.3736,
      "step": 1973
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0923330585325637e-05,
      "loss": 4.178,
      "step": 1974
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.091714756801319e-05,
      "loss": 4.6051,
      "step": 1975
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0910964550700742e-05,
      "loss": 4.2664,
      "step": 1976
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0904781533388297e-05,
      "loss": 4.3108,
      "step": 1977
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0898598516075844e-05,
      "loss": 3.9644,
      "step": 1978
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0892415498763395e-05,
      "loss": 3.8746,
      "step": 1979
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.088623248145095e-05,
      "loss": 4.1653,
      "step": 1980
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.08800494641385e-05,
      "loss": 4.3786,
      "step": 1981
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0873866446826052e-05,
      "loss": 4.1702,
      "step": 1982
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0867683429513603e-05,
      "loss": 3.9059,
      "step": 1983
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0861500412201154e-05,
      "loss": 3.6198,
      "step": 1984
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0855317394888708e-05,
      "loss": 4.3958,
      "step": 1985
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.084913437757626e-05,
      "loss": 4.4311,
      "step": 1986
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0842951360263807e-05,
      "loss": 4.5046,
      "step": 1987
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.083676834295136e-05,
      "loss": 4.491,
      "step": 1988
    },
    {
      "epoch": 1.11,
      "learning_rate": 2.0830585325638912e-05,
      "loss": 4.515,
      "step": 1989
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.0824402308326463e-05,
      "loss": 4.399,
      "step": 1990
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.0818219291014017e-05,
      "loss": 4.5491,
      "step": 1991
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.0812036273701565e-05,
      "loss": 4.1055,
      "step": 1992
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.080585325638912e-05,
      "loss": 4.3937,
      "step": 1993
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.079967023907667e-05,
      "loss": 4.3868,
      "step": 1994
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.079348722176422e-05,
      "loss": 4.4928,
      "step": 1995
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.0787304204451772e-05,
      "loss": 4.3613,
      "step": 1996
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.0781121187139323e-05,
      "loss": 4.5427,
      "step": 1997
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.0774938169826874e-05,
      "loss": 4.4401,
      "step": 1998
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.076875515251443e-05,
      "loss": 4.5417,
      "step": 1999
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.076257213520198e-05,
      "loss": 4.3428,
      "step": 2000
    },
    {
      "epoch": 1.12,
      "eval_loss": 4.474549293518066,
      "eval_runtime": 1311.2071,
      "eval_samples_per_second": 2.015,
      "eval_steps_per_second": 0.252,
      "eval_wer": 1.6443240669087447,
      "step": 2000
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.075638911788953e-05,
      "loss": 4.4516,
      "step": 2001
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.075020610057708e-05,
      "loss": 4.2322,
      "step": 2002
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.0744023083264632e-05,
      "loss": 4.5389,
      "step": 2003
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.0737840065952187e-05,
      "loss": 4.5045,
      "step": 2004
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.0731657048639738e-05,
      "loss": 4.3019,
      "step": 2005
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.072547403132729e-05,
      "loss": 4.2069,
      "step": 2006
    },
    {
      "epoch": 1.12,
      "learning_rate": 2.071929101401484e-05,
      "loss": 4.3275,
      "step": 2007
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.071310799670239e-05,
      "loss": 4.4545,
      "step": 2008
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0706924979389945e-05,
      "loss": 4.3846,
      "step": 2009
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0700741962077493e-05,
      "loss": 4.2781,
      "step": 2010
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0694558944765044e-05,
      "loss": 4.1645,
      "step": 2011
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0688375927452598e-05,
      "loss": 4.2734,
      "step": 2012
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.068219291014015e-05,
      "loss": 4.4283,
      "step": 2013
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0676009892827703e-05,
      "loss": 4.4703,
      "step": 2014
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.066982687551525e-05,
      "loss": 4.3114,
      "step": 2015
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0663643858202802e-05,
      "loss": 4.2938,
      "step": 2016
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0657460840890356e-05,
      "loss": 4.3316,
      "step": 2017
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0651277823577907e-05,
      "loss": 4.2375,
      "step": 2018
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0645094806265455e-05,
      "loss": 4.2703,
      "step": 2019
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.063891178895301e-05,
      "loss": 4.3927,
      "step": 2020
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.063272877164056e-05,
      "loss": 4.3564,
      "step": 2021
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0626545754328115e-05,
      "loss": 4.1281,
      "step": 2022
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0620362737015666e-05,
      "loss": 4.202,
      "step": 2023
    },
    {
      "epoch": 1.13,
      "learning_rate": 2.0614179719703213e-05,
      "loss": 4.3404,
      "step": 2024
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0607996702390768e-05,
      "loss": 4.2599,
      "step": 2025
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.060181368507832e-05,
      "loss": 4.2635,
      "step": 2026
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.059563066776587e-05,
      "loss": 4.185,
      "step": 2027
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.058944765045342e-05,
      "loss": 4.1584,
      "step": 2028
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.058326463314097e-05,
      "loss": 4.3667,
      "step": 2029
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0577081615828526e-05,
      "loss": 4.4673,
      "step": 2030
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0570898598516077e-05,
      "loss": 4.4837,
      "step": 2031
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0564715581203628e-05,
      "loss": 3.7614,
      "step": 2032
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.055853256389118e-05,
      "loss": 3.9464,
      "step": 2033
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.055234954657873e-05,
      "loss": 4.0917,
      "step": 2034
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0546166529266284e-05,
      "loss": 4.4338,
      "step": 2035
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0539983511953835e-05,
      "loss": 4.6446,
      "step": 2036
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0533800494641386e-05,
      "loss": 4.6644,
      "step": 2037
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0527617477328937e-05,
      "loss": 4.3749,
      "step": 2038
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0521434460016488e-05,
      "loss": 4.5948,
      "step": 2039
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.051525144270404e-05,
      "loss": 4.3888,
      "step": 2040
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.0509068425391593e-05,
      "loss": 4.2353,
      "step": 2041
    },
    {
      "epoch": 1.14,
      "learning_rate": 2.050288540807914e-05,
      "loss": 4.4745,
      "step": 2042
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0496702390766695e-05,
      "loss": 4.4788,
      "step": 2043
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0490519373454246e-05,
      "loss": 4.1575,
      "step": 2044
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0484336356141797e-05,
      "loss": 4.3755,
      "step": 2045
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0478153338829352e-05,
      "loss": 4.4879,
      "step": 2046
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.04719703215169e-05,
      "loss": 4.2257,
      "step": 2047
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.046578730420445e-05,
      "loss": 4.5838,
      "step": 2048
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0459604286892005e-05,
      "loss": 4.6056,
      "step": 2049
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0453421269579556e-05,
      "loss": 4.5973,
      "step": 2050
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0447238252267107e-05,
      "loss": 4.1346,
      "step": 2051
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0441055234954658e-05,
      "loss": 4.3612,
      "step": 2052
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.043487221764221e-05,
      "loss": 4.3383,
      "step": 2053
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0428689200329763e-05,
      "loss": 4.2103,
      "step": 2054
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0422506183017314e-05,
      "loss": 4.5121,
      "step": 2055
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.041632316570486e-05,
      "loss": 4.3207,
      "step": 2056
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0410140148392416e-05,
      "loss": 4.23,
      "step": 2057
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0403957131079967e-05,
      "loss": 4.2728,
      "step": 2058
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.039777411376752e-05,
      "loss": 4.6098,
      "step": 2059
    },
    {
      "epoch": 1.15,
      "learning_rate": 2.0391591096455072e-05,
      "loss": 4.4136,
      "step": 2060
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.038540807914262e-05,
      "loss": 4.448,
      "step": 2061
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0379225061830174e-05,
      "loss": 4.2923,
      "step": 2062
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0373042044517725e-05,
      "loss": 4.41,
      "step": 2063
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.036685902720528e-05,
      "loss": 4.2519,
      "step": 2064
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0360676009892827e-05,
      "loss": 4.4244,
      "step": 2065
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0354492992580378e-05,
      "loss": 4.4289,
      "step": 2066
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0348309975267933e-05,
      "loss": 4.19,
      "step": 2067
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0342126957955484e-05,
      "loss": 4.3552,
      "step": 2068
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0335943940643035e-05,
      "loss": 4.4873,
      "step": 2069
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0329760923330585e-05,
      "loss": 4.295,
      "step": 2070
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0323577906018136e-05,
      "loss": 4.3532,
      "step": 2071
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.031739488870569e-05,
      "loss": 4.1788,
      "step": 2072
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0311211871393242e-05,
      "loss": 4.0925,
      "step": 2073
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.030502885408079e-05,
      "loss": 3.9995,
      "step": 2074
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0298845836768344e-05,
      "loss": 4.3691,
      "step": 2075
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0292662819455895e-05,
      "loss": 4.5198,
      "step": 2076
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0286479802143446e-05,
      "loss": 4.1198,
      "step": 2077
    },
    {
      "epoch": 1.16,
      "learning_rate": 2.0280296784831e-05,
      "loss": 4.4138,
      "step": 2078
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0274113767518548e-05,
      "loss": 3.9976,
      "step": 2079
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0267930750206102e-05,
      "loss": 4.0523,
      "step": 2080
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0261747732893653e-05,
      "loss": 4.1781,
      "step": 2081
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0255564715581204e-05,
      "loss": 3.8827,
      "step": 2082
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0249381698268755e-05,
      "loss": 4.0526,
      "step": 2083
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0243198680956306e-05,
      "loss": 3.93,
      "step": 2084
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0237015663643857e-05,
      "loss": 4.5646,
      "step": 2085
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.023083264633141e-05,
      "loss": 4.1927,
      "step": 2086
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0224649629018962e-05,
      "loss": 4.5193,
      "step": 2087
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0218466611706513e-05,
      "loss": 4.5334,
      "step": 2088
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0212283594394064e-05,
      "loss": 4.4105,
      "step": 2089
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0206100577081615e-05,
      "loss": 4.453,
      "step": 2090
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.019991755976917e-05,
      "loss": 4.544,
      "step": 2091
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.019373454245672e-05,
      "loss": 4.4502,
      "step": 2092
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.018755152514427e-05,
      "loss": 4.3738,
      "step": 2093
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0181368507831823e-05,
      "loss": 4.4727,
      "step": 2094
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0175185490519374e-05,
      "loss": 4.4824,
      "step": 2095
    },
    {
      "epoch": 1.17,
      "learning_rate": 2.0169002473206928e-05,
      "loss": 4.4192,
      "step": 2096
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0162819455894476e-05,
      "loss": 4.5892,
      "step": 2097
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0156636438582027e-05,
      "loss": 4.4321,
      "step": 2098
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.015045342126958e-05,
      "loss": 4.577,
      "step": 2099
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0144270403957132e-05,
      "loss": 4.1413,
      "step": 2100
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0138087386644686e-05,
      "loss": 4.4902,
      "step": 2101
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0131904369332234e-05,
      "loss": 4.3069,
      "step": 2102
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0125721352019785e-05,
      "loss": 4.2917,
      "step": 2103
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.011953833470734e-05,
      "loss": 4.4344,
      "step": 2104
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.011335531739489e-05,
      "loss": 4.3478,
      "step": 2105
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0107172300082438e-05,
      "loss": 4.6277,
      "step": 2106
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0100989282769992e-05,
      "loss": 4.5393,
      "step": 2107
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0094806265457543e-05,
      "loss": 4.2745,
      "step": 2108
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0088623248145097e-05,
      "loss": 4.3886,
      "step": 2109
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.008244023083265e-05,
      "loss": 4.271,
      "step": 2110
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0076257213520196e-05,
      "loss": 4.4997,
      "step": 2111
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.007007419620775e-05,
      "loss": 4.4274,
      "step": 2112
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.00638911788953e-05,
      "loss": 4.3796,
      "step": 2113
    },
    {
      "epoch": 1.18,
      "learning_rate": 2.0057708161582852e-05,
      "loss": 4.3866,
      "step": 2114
    },
    {
      "epoch": 1.19,
      "learning_rate": 2.0051525144270403e-05,
      "loss": 4.3013,
      "step": 2115
    },
    {
      "epoch": 1.19,
      "learning_rate": 2.0045342126957954e-05,
      "loss": 4.0244,
      "step": 2116
    },
    {
      "epoch": 1.19,
      "learning_rate": 2.003915910964551e-05,
      "loss": 4.1405,
      "step": 2117
    },
    {
      "epoch": 1.19,
      "learning_rate": 2.003297609233306e-05,
      "loss": 4.3782,
      "step": 2118
    },
    {
      "epoch": 1.19,
      "learning_rate": 2.002679307502061e-05,
      "loss": 4.2703,
      "step": 2119
    },
    {
      "epoch": 1.19,
      "learning_rate": 2.002061005770816e-05,
      "loss": 4.0796,
      "step": 2120
    },
    {
      "epoch": 1.19,
      "learning_rate": 2.0014427040395713e-05,
      "loss": 4.3866,
      "step": 2121
    },
    {
      "epoch": 1.19,
      "learning_rate": 2.0008244023083264e-05,
      "loss": 4.5357,
      "step": 2122
    },
    {
      "epoch": 1.19,
      "learning_rate": 2.0002061005770818e-05,
      "loss": 4.394,
      "step": 2123
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.999587798845837e-05,
      "loss": 4.2282,
      "step": 2124
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.998969497114592e-05,
      "loss": 4.3078,
      "step": 2125
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.998351195383347e-05,
      "loss": 4.2107,
      "step": 2126
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.9977328936521022e-05,
      "loss": 3.9826,
      "step": 2127
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.9971145919208576e-05,
      "loss": 4.3478,
      "step": 2128
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.9964962901896124e-05,
      "loss": 4.2146,
      "step": 2129
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.9958779884583678e-05,
      "loss": 4.5415,
      "step": 2130
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.995259686727123e-05,
      "loss": 3.8677,
      "step": 2131
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.994641384995878e-05,
      "loss": 4.4043,
      "step": 2132
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9940230832646335e-05,
      "loss": 4.2371,
      "step": 2133
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9934047815333882e-05,
      "loss": 3.8848,
      "step": 2134
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9927864798021433e-05,
      "loss": 4.6289,
      "step": 2135
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9921681780708988e-05,
      "loss": 4.6668,
      "step": 2136
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.991549876339654e-05,
      "loss": 4.4422,
      "step": 2137
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.990931574608409e-05,
      "loss": 4.3394,
      "step": 2138
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.990313272877164e-05,
      "loss": 4.1751,
      "step": 2139
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.989694971145919e-05,
      "loss": 4.4324,
      "step": 2140
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9890766694146746e-05,
      "loss": 4.426,
      "step": 2141
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9884583676834297e-05,
      "loss": 4.3378,
      "step": 2142
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9878400659521844e-05,
      "loss": 4.4165,
      "step": 2143
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.98722176422094e-05,
      "loss": 4.4919,
      "step": 2144
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.986603462489695e-05,
      "loss": 4.7451,
      "step": 2145
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9859851607584504e-05,
      "loss": 4.3674,
      "step": 2146
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.985366859027205e-05,
      "loss": 4.25,
      "step": 2147
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9847485572959603e-05,
      "loss": 4.3495,
      "step": 2148
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9841302555647157e-05,
      "loss": 4.3253,
      "step": 2149
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9835119538334708e-05,
      "loss": 4.4867,
      "step": 2150
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.982893652102226e-05,
      "loss": 4.3754,
      "step": 2151
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.982275350370981e-05,
      "loss": 4.6177,
      "step": 2152
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.981657048639736e-05,
      "loss": 4.493,
      "step": 2153
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9810387469084915e-05,
      "loss": 4.3989,
      "step": 2154
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9804204451772466e-05,
      "loss": 4.4317,
      "step": 2155
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9798021434460017e-05,
      "loss": 4.274,
      "step": 2156
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9791838417147568e-05,
      "loss": 4.654,
      "step": 2157
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.978565539983512e-05,
      "loss": 4.4047,
      "step": 2158
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9779472382522674e-05,
      "loss": 4.4249,
      "step": 2159
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9773289365210225e-05,
      "loss": 4.3888,
      "step": 2160
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9767106347897772e-05,
      "loss": 4.3557,
      "step": 2161
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9760923330585327e-05,
      "loss": 4.3206,
      "step": 2162
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9754740313272878e-05,
      "loss": 4.354,
      "step": 2163
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.974855729596043e-05,
      "loss": 4.4539,
      "step": 2164
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9742374278647983e-05,
      "loss": 4.4616,
      "step": 2165
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.973619126133553e-05,
      "loss": 4.1975,
      "step": 2166
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9730008244023085e-05,
      "loss": 4.2407,
      "step": 2167
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9723825226710636e-05,
      "loss": 4.2157,
      "step": 2168
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9717642209398187e-05,
      "loss": 4.3449,
      "step": 2169
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9711459192085738e-05,
      "loss": 4.4758,
      "step": 2170
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.970527617477329e-05,
      "loss": 4.3033,
      "step": 2171
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.969909315746084e-05,
      "loss": 4.5612,
      "step": 2172
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9692910140148394e-05,
      "loss": 4.3969,
      "step": 2173
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9686727122835945e-05,
      "loss": 4.2287,
      "step": 2174
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9680544105523496e-05,
      "loss": 4.3303,
      "step": 2175
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9674361088211047e-05,
      "loss": 4.1276,
      "step": 2176
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9668178070898598e-05,
      "loss": 4.0273,
      "step": 2177
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9661995053586152e-05,
      "loss": 4.3491,
      "step": 2178
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9655812036273703e-05,
      "loss": 3.7609,
      "step": 2179
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.964962901896125e-05,
      "loss": 4.152,
      "step": 2180
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9643446001648805e-05,
      "loss": 3.6337,
      "step": 2181
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9637262984336356e-05,
      "loss": 3.4852,
      "step": 2182
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.963107996702391e-05,
      "loss": 4.1933,
      "step": 2183
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.962489694971146e-05,
      "loss": 3.4299,
      "step": 2184
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.961871393239901e-05,
      "loss": 4.6246,
      "step": 2185
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9612530915086564e-05,
      "loss": 4.505,
      "step": 2186
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9606347897774115e-05,
      "loss": 4.4134,
      "step": 2187
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.960016488046167e-05,
      "loss": 4.543,
      "step": 2188
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9593981863149217e-05,
      "loss": 4.5559,
      "step": 2189
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9587798845836768e-05,
      "loss": 4.3872,
      "step": 2190
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9581615828524322e-05,
      "loss": 4.5051,
      "step": 2191
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9575432811211873e-05,
      "loss": 4.581,
      "step": 2192
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.956924979389942e-05,
      "loss": 4.503,
      "step": 2193
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9563066776586975e-05,
      "loss": 4.2694,
      "step": 2194
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9556883759274526e-05,
      "loss": 4.2353,
      "step": 2195
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.955070074196208e-05,
      "loss": 4.3805,
      "step": 2196
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.954451772464963e-05,
      "loss": 4.3869,
      "step": 2197
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.953833470733718e-05,
      "loss": 4.5793,
      "step": 2198
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9532151690024733e-05,
      "loss": 4.4393,
      "step": 2199
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9525968672712284e-05,
      "loss": 4.5363,
      "step": 2200
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9519785655399835e-05,
      "loss": 4.507,
      "step": 2201
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9513602638087386e-05,
      "loss": 4.5218,
      "step": 2202
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9507419620774937e-05,
      "loss": 4.2717,
      "step": 2203
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.950123660346249e-05,
      "loss": 4.4879,
      "step": 2204
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9495053586150042e-05,
      "loss": 4.5186,
      "step": 2205
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9488870568837593e-05,
      "loss": 4.3545,
      "step": 2206
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9482687551525144e-05,
      "loss": 4.4373,
      "step": 2207
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9476504534212695e-05,
      "loss": 4.2843,
      "step": 2208
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9470321516900246e-05,
      "loss": 4.1208,
      "step": 2209
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.94641384995878e-05,
      "loss": 4.1567,
      "step": 2210
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9457955482275352e-05,
      "loss": 4.327,
      "step": 2211
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9451772464962903e-05,
      "loss": 4.2306,
      "step": 2212
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9445589447650454e-05,
      "loss": 4.3013,
      "step": 2213
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9439406430338005e-05,
      "loss": 4.4893,
      "step": 2214
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.943322341302556e-05,
      "loss": 4.1546,
      "step": 2215
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9427040395713107e-05,
      "loss": 4.3509,
      "step": 2216
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.942085737840066e-05,
      "loss": 4.2018,
      "step": 2217
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9414674361088212e-05,
      "loss": 4.6033,
      "step": 2218
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9408491343775763e-05,
      "loss": 4.4702,
      "step": 2219
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9402308326463317e-05,
      "loss": 4.3959,
      "step": 2220
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9396125309150865e-05,
      "loss": 4.3795,
      "step": 2221
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9389942291838416e-05,
      "loss": 4.3168,
      "step": 2222
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.938375927452597e-05,
      "loss": 4.2485,
      "step": 2223
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.937757625721352e-05,
      "loss": 4.424,
      "step": 2224
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9371393239901072e-05,
      "loss": 4.4057,
      "step": 2225
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9365210222588623e-05,
      "loss": 4.2877,
      "step": 2226
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9359027205276174e-05,
      "loss": 4.0986,
      "step": 2227
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.935284418796373e-05,
      "loss": 3.9672,
      "step": 2228
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.934666117065128e-05,
      "loss": 4.4667,
      "step": 2229
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9340478153338827e-05,
      "loss": 4.0838,
      "step": 2230
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.933429513602638e-05,
      "loss": 3.8408,
      "step": 2231
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9328112118713933e-05,
      "loss": 3.7235,
      "step": 2232
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9321929101401487e-05,
      "loss": 4.1743,
      "step": 2233
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9315746084089034e-05,
      "loss": 3.8335,
      "step": 2234
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9309563066776585e-05,
      "loss": 4.7378,
      "step": 2235
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.930338004946414e-05,
      "loss": 4.3732,
      "step": 2236
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.929719703215169e-05,
      "loss": 4.2161,
      "step": 2237
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9291014014839242e-05,
      "loss": 4.6826,
      "step": 2238
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9284830997526793e-05,
      "loss": 4.332,
      "step": 2239
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9278647980214344e-05,
      "loss": 4.3555,
      "step": 2240
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9272464962901898e-05,
      "loss": 4.4415,
      "step": 2241
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.926628194558945e-05,
      "loss": 4.716,
      "step": 2242
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9260098928277e-05,
      "loss": 4.3353,
      "step": 2243
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.925391591096455e-05,
      "loss": 4.006,
      "step": 2244
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9247732893652102e-05,
      "loss": 4.2269,
      "step": 2245
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9241549876339653e-05,
      "loss": 4.2182,
      "step": 2246
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9235366859027207e-05,
      "loss": 4.2295,
      "step": 2247
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9229183841714755e-05,
      "loss": 4.2432,
      "step": 2248
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.922300082440231e-05,
      "loss": 4.2045,
      "step": 2249
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.921681780708986e-05,
      "loss": 4.0994,
      "step": 2250
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.921063478977741e-05,
      "loss": 4.3278,
      "step": 2251
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9204451772464966e-05,
      "loss": 4.1367,
      "step": 2252
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9198268755152513e-05,
      "loss": 4.4486,
      "step": 2253
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9192085737840068e-05,
      "loss": 4.1753,
      "step": 2254
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.918590272052762e-05,
      "loss": 4.4093,
      "step": 2255
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.917971970321517e-05,
      "loss": 4.5692,
      "step": 2256
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.917353668590272e-05,
      "loss": 4.2032,
      "step": 2257
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.916735366859027e-05,
      "loss": 4.1634,
      "step": 2258
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9161170651277823e-05,
      "loss": 4.2206,
      "step": 2259
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9154987633965377e-05,
      "loss": 4.4881,
      "step": 2260
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9148804616652928e-05,
      "loss": 4.3966,
      "step": 2261
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.914262159934048e-05,
      "loss": 4.1462,
      "step": 2262
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.913643858202803e-05,
      "loss": 4.4314,
      "step": 2263
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.913025556471558e-05,
      "loss": 4.2823,
      "step": 2264
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9124072547403135e-05,
      "loss": 4.5899,
      "step": 2265
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9117889530090683e-05,
      "loss": 4.1664,
      "step": 2266
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9111706512778234e-05,
      "loss": 4.3714,
      "step": 2267
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9105523495465788e-05,
      "loss": 4.288,
      "step": 2268
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.909934047815334e-05,
      "loss": 4.223,
      "step": 2269
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9093157460840894e-05,
      "loss": 4.0231,
      "step": 2270
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.908697444352844e-05,
      "loss": 4.3629,
      "step": 2271
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9080791426215992e-05,
      "loss": 4.1251,
      "step": 2272
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9074608408903546e-05,
      "loss": 4.1471,
      "step": 2273
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9068425391591097e-05,
      "loss": 3.874,
      "step": 2274
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.906224237427865e-05,
      "loss": 4.2307,
      "step": 2275
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.90560593569662e-05,
      "loss": 4.1589,
      "step": 2276
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.904987633965375e-05,
      "loss": 4.039,
      "step": 2277
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.9043693322341305e-05,
      "loss": 3.8864,
      "step": 2278
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.9037510305028856e-05,
      "loss": 4.145,
      "step": 2279
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.9031327287716403e-05,
      "loss": 3.7966,
      "step": 2280
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.9025144270403958e-05,
      "loss": 3.954,
      "step": 2281
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.901896125309151e-05,
      "loss": 4.0318,
      "step": 2282
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.9012778235779063e-05,
      "loss": 4.0853,
      "step": 2283
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.9006595218466614e-05,
      "loss": 3.7707,
      "step": 2284
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.900041220115416e-05,
      "loss": 4.6734,
      "step": 2285
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.8994229183841716e-05,
      "loss": 4.5097,
      "step": 2286
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.8988046166529267e-05,
      "loss": 4.3823,
      "step": 2287
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.8981863149216818e-05,
      "loss": 4.3023,
      "step": 2288
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.897568013190437e-05,
      "loss": 4.5006,
      "step": 2289
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.896949711459192e-05,
      "loss": 4.3886,
      "step": 2290
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.8963314097279474e-05,
      "loss": 4.4823,
      "step": 2291
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.8957131079967025e-05,
      "loss": 4.4941,
      "step": 2292
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8950948062654576e-05,
      "loss": 4.2939,
      "step": 2293
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8944765045342127e-05,
      "loss": 4.321,
      "step": 2294
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8938582028029678e-05,
      "loss": 4.3527,
      "step": 2295
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.893239901071723e-05,
      "loss": 4.2872,
      "step": 2296
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8926215993404784e-05,
      "loss": 4.21,
      "step": 2297
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8920032976092335e-05,
      "loss": 4.562,
      "step": 2298
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8913849958779886e-05,
      "loss": 4.486,
      "step": 2299
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8907666941467436e-05,
      "loss": 4.2715,
      "step": 2300
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8901483924154987e-05,
      "loss": 4.3031,
      "step": 2301
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8895300906842542e-05,
      "loss": 4.3128,
      "step": 2302
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.888911788953009e-05,
      "loss": 4.196,
      "step": 2303
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.888293487221764e-05,
      "loss": 4.1048,
      "step": 2304
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8876751854905195e-05,
      "loss": 4.2319,
      "step": 2305
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8870568837592746e-05,
      "loss": 4.5821,
      "step": 2306
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.88643858202803e-05,
      "loss": 4.1847,
      "step": 2307
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8858202802967848e-05,
      "loss": 4.3634,
      "step": 2308
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.88520197856554e-05,
      "loss": 4.3165,
      "step": 2309
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8845836768342953e-05,
      "loss": 4.2524,
      "step": 2310
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8839653751030504e-05,
      "loss": 4.2196,
      "step": 2311
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8833470733718055e-05,
      "loss": 4.2269,
      "step": 2312
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8827287716405606e-05,
      "loss": 4.5434,
      "step": 2313
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8821104699093157e-05,
      "loss": 4.4683,
      "step": 2314
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.881492168178071e-05,
      "loss": 4.1411,
      "step": 2315
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8808738664468262e-05,
      "loss": 4.1717,
      "step": 2316
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.880255564715581e-05,
      "loss": 4.198,
      "step": 2317
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8796372629843364e-05,
      "loss": 4.4916,
      "step": 2318
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8790189612530915e-05,
      "loss": 4.4019,
      "step": 2319
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.878400659521847e-05,
      "loss": 4.5165,
      "step": 2320
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8777823577906017e-05,
      "loss": 4.2314,
      "step": 2321
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8771640560593568e-05,
      "loss": 3.9948,
      "step": 2322
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8765457543281123e-05,
      "loss": 4.3442,
      "step": 2323
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8759274525968674e-05,
      "loss": 4.0783,
      "step": 2324
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8753091508656225e-05,
      "loss": 4.2598,
      "step": 2325
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8746908491343776e-05,
      "loss": 4.0615,
      "step": 2326
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8740725474031327e-05,
      "loss": 3.941,
      "step": 2327
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.873454245671888e-05,
      "loss": 4.0286,
      "step": 2328
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8728359439406432e-05,
      "loss": 3.8476,
      "step": 2329
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8722176422093983e-05,
      "loss": 4.2197,
      "step": 2330
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8715993404781534e-05,
      "loss": 3.7276,
      "step": 2331
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8709810387469085e-05,
      "loss": 3.7128,
      "step": 2332
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8703627370156636e-05,
      "loss": 3.73,
      "step": 2333
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.869744435284419e-05,
      "loss": 3.8237,
      "step": 2334
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8691261335531738e-05,
      "loss": 4.5779,
      "step": 2335
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8685078318219292e-05,
      "loss": 4.2743,
      "step": 2336
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8678895300906843e-05,
      "loss": 4.5568,
      "step": 2337
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8672712283594394e-05,
      "loss": 4.562,
      "step": 2338
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.866652926628195e-05,
      "loss": 4.567,
      "step": 2339
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8660346248969496e-05,
      "loss": 4.479,
      "step": 2340
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.865416323165705e-05,
      "loss": 4.3397,
      "step": 2341
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.86479802143446e-05,
      "loss": 4.3375,
      "step": 2342
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8641797197032152e-05,
      "loss": 4.1428,
      "step": 2343
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8635614179719703e-05,
      "loss": 4.3554,
      "step": 2344
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8629431162407254e-05,
      "loss": 4.305,
      "step": 2345
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8623248145094805e-05,
      "loss": 4.422,
      "step": 2346
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.861706512778236e-05,
      "loss": 4.3076,
      "step": 2347
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.861088211046991e-05,
      "loss": 4.2784,
      "step": 2348
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.860469909315746e-05,
      "loss": 4.3011,
      "step": 2349
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8598516075845013e-05,
      "loss": 4.5948,
      "step": 2350
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8592333058532564e-05,
      "loss": 4.2766,
      "step": 2351
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8586150041220118e-05,
      "loss": 4.0119,
      "step": 2352
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8579967023907666e-05,
      "loss": 4.343,
      "step": 2353
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8573784006595217e-05,
      "loss": 4.1308,
      "step": 2354
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.856760098928277e-05,
      "loss": 4.2571,
      "step": 2355
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8561417971970322e-05,
      "loss": 4.0625,
      "step": 2356
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8555234954657876e-05,
      "loss": 4.4958,
      "step": 2357
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8549051937345424e-05,
      "loss": 4.0843,
      "step": 2358
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8542868920032975e-05,
      "loss": 4.4095,
      "step": 2359
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.853668590272053e-05,
      "loss": 4.6385,
      "step": 2360
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.853050288540808e-05,
      "loss": 4.3884,
      "step": 2361
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.852431986809563e-05,
      "loss": 4.3578,
      "step": 2362
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8518136850783182e-05,
      "loss": 4.2843,
      "step": 2363
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8511953833470733e-05,
      "loss": 4.3879,
      "step": 2364
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8505770816158288e-05,
      "loss": 4.3096,
      "step": 2365
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.849958779884584e-05,
      "loss": 4.1925,
      "step": 2366
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8493404781533386e-05,
      "loss": 4.2453,
      "step": 2367
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.848722176422094e-05,
      "loss": 4.3368,
      "step": 2368
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.848103874690849e-05,
      "loss": 4.4287,
      "step": 2369
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8474855729596042e-05,
      "loss": 4.3161,
      "step": 2370
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8468672712283597e-05,
      "loss": 4.2172,
      "step": 2371
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8462489694971144e-05,
      "loss": 4.1938,
      "step": 2372
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.84563066776587e-05,
      "loss": 4.2443,
      "step": 2373
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.845012366034625e-05,
      "loss": 4.2781,
      "step": 2374
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.84439406430338e-05,
      "loss": 4.4438,
      "step": 2375
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8437757625721352e-05,
      "loss": 4.2364,
      "step": 2376
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8431574608408903e-05,
      "loss": 4.0949,
      "step": 2377
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8425391591096457e-05,
      "loss": 4.0811,
      "step": 2378
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8419208573784008e-05,
      "loss": 4.2079,
      "step": 2379
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.841302555647156e-05,
      "loss": 4.0991,
      "step": 2380
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.840684253915911e-05,
      "loss": 3.822,
      "step": 2381
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.840065952184666e-05,
      "loss": 4.0566,
      "step": 2382
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8394476504534212e-05,
      "loss": 3.8419,
      "step": 2383
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8388293487221766e-05,
      "loss": 3.5919,
      "step": 2384
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8382110469909317e-05,
      "loss": 4.7195,
      "step": 2385
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8375927452596868e-05,
      "loss": 4.3668,
      "step": 2386
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.836974443528442e-05,
      "loss": 4.5737,
      "step": 2387
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.836356141797197e-05,
      "loss": 4.2144,
      "step": 2388
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8357378400659525e-05,
      "loss": 4.3556,
      "step": 2389
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8351195383347072e-05,
      "loss": 4.4636,
      "step": 2390
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8345012366034623e-05,
      "loss": 4.4662,
      "step": 2391
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8338829348722178e-05,
      "loss": 4.3624,
      "step": 2392
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.833264633140973e-05,
      "loss": 4.4144,
      "step": 2393
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8326463314097283e-05,
      "loss": 4.3875,
      "step": 2394
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.832028029678483e-05,
      "loss": 4.1073,
      "step": 2395
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.831409727947238e-05,
      "loss": 4.1739,
      "step": 2396
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8307914262159936e-05,
      "loss": 4.352,
      "step": 2397
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8301731244847487e-05,
      "loss": 4.2137,
      "step": 2398
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8295548227535034e-05,
      "loss": 4.4398,
      "step": 2399
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.828936521022259e-05,
      "loss": 4.3696,
      "step": 2400
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.828318219291014e-05,
      "loss": 4.7913,
      "step": 2401
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8276999175597694e-05,
      "loss": 4.2564,
      "step": 2402
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8270816158285245e-05,
      "loss": 4.3526,
      "step": 2403
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8264633140972793e-05,
      "loss": 4.3771,
      "step": 2404
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8258450123660347e-05,
      "loss": 4.2964,
      "step": 2405
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8252267106347898e-05,
      "loss": 4.2271,
      "step": 2406
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8246084089035452e-05,
      "loss": 4.2367,
      "step": 2407
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8239901071723e-05,
      "loss": 4.1689,
      "step": 2408
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.823371805441055e-05,
      "loss": 4.4406,
      "step": 2409
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8227535037098105e-05,
      "loss": 4.319,
      "step": 2410
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8221352019785656e-05,
      "loss": 4.3464,
      "step": 2411
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8215169002473207e-05,
      "loss": 4.2531,
      "step": 2412
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.820898598516076e-05,
      "loss": 4.3818,
      "step": 2413
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.820280296784831e-05,
      "loss": 4.2077,
      "step": 2414
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8196619950535864e-05,
      "loss": 4.1819,
      "step": 2415
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8190436933223415e-05,
      "loss": 4.4864,
      "step": 2416
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8184253915910966e-05,
      "loss": 4.3858,
      "step": 2417
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8178070898598517e-05,
      "loss": 4.326,
      "step": 2418
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8171887881286068e-05,
      "loss": 4.3395,
      "step": 2419
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.816570486397362e-05,
      "loss": 4.0963,
      "step": 2420
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8159521846661173e-05,
      "loss": 4.2899,
      "step": 2421
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.815333882934872e-05,
      "loss": 4.3945,
      "step": 2422
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8147155812036275e-05,
      "loss": 4.1482,
      "step": 2423
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8140972794723826e-05,
      "loss": 4.1416,
      "step": 2424
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8134789777411377e-05,
      "loss": 4.2687,
      "step": 2425
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.812860676009893e-05,
      "loss": 4.1619,
      "step": 2426
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.812242374278648e-05,
      "loss": 4.2813,
      "step": 2427
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.811624072547403e-05,
      "loss": 3.9782,
      "step": 2428
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8110057708161584e-05,
      "loss": 4.0144,
      "step": 2429
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8103874690849135e-05,
      "loss": 3.9539,
      "step": 2430
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8097691673536686e-05,
      "loss": 4.0618,
      "step": 2431
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8091508656224237e-05,
      "loss": 3.4768,
      "step": 2432
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8085325638911788e-05,
      "loss": 3.5415,
      "step": 2433
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8079142621599342e-05,
      "loss": 3.828,
      "step": 2434
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8072959604286893e-05,
      "loss": 4.4213,
      "step": 2435
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8066776586974444e-05,
      "loss": 4.6643,
      "step": 2436
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8060593569661995e-05,
      "loss": 4.4986,
      "step": 2437
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8054410552349546e-05,
      "loss": 4.5117,
      "step": 2438
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.80482275350371e-05,
      "loss": 4.5266,
      "step": 2439
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.804204451772465e-05,
      "loss": 4.2712,
      "step": 2440
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.80358615004122e-05,
      "loss": 4.4974,
      "step": 2441
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8029678483099754e-05,
      "loss": 4.5016,
      "step": 2442
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8023495465787305e-05,
      "loss": 4.3068,
      "step": 2443
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.801731244847486e-05,
      "loss": 4.2934,
      "step": 2444
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8011129431162407e-05,
      "loss": 4.1837,
      "step": 2445
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8004946413849958e-05,
      "loss": 4.4208,
      "step": 2446
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7998763396537512e-05,
      "loss": 4.6229,
      "step": 2447
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7992580379225063e-05,
      "loss": 4.1931,
      "step": 2448
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7986397361912614e-05,
      "loss": 4.3707,
      "step": 2449
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7980214344600165e-05,
      "loss": 4.2318,
      "step": 2450
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7974031327287716e-05,
      "loss": 4.2668,
      "step": 2451
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.796784830997527e-05,
      "loss": 4.3498,
      "step": 2452
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.796166529266282e-05,
      "loss": 4.4423,
      "step": 2453
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.795548227535037e-05,
      "loss": 4.2926,
      "step": 2454
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7949299258037923e-05,
      "loss": 4.3649,
      "step": 2455
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7943116240725474e-05,
      "loss": 4.3603,
      "step": 2456
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7936933223413025e-05,
      "loss": 4.6284,
      "step": 2457
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.793075020610058e-05,
      "loss": 4.149,
      "step": 2458
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7924567188788127e-05,
      "loss": 4.3863,
      "step": 2459
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.791838417147568e-05,
      "loss": 3.9513,
      "step": 2460
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7912201154163233e-05,
      "loss": 4.3681,
      "step": 2461
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7906018136850783e-05,
      "loss": 4.1853,
      "step": 2462
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7899835119538334e-05,
      "loss": 4.2565,
      "step": 2463
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7893652102225885e-05,
      "loss": 4.2795,
      "step": 2464
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7887469084913436e-05,
      "loss": 4.38,
      "step": 2465
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.788128606760099e-05,
      "loss": 4.1052,
      "step": 2466
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7875103050288542e-05,
      "loss": 4.2991,
      "step": 2467
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7868920032976093e-05,
      "loss": 4.3013,
      "step": 2468
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7862737015663644e-05,
      "loss": 4.3062,
      "step": 2469
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7856553998351195e-05,
      "loss": 4.3694,
      "step": 2470
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.785037098103875e-05,
      "loss": 4.1294,
      "step": 2471
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7844187963726297e-05,
      "loss": 4.1361,
      "step": 2472
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.783800494641385e-05,
      "loss": 4.1195,
      "step": 2473
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7831821929101402e-05,
      "loss": 4.0458,
      "step": 2474
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7825638911788953e-05,
      "loss": 4.3365,
      "step": 2475
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7819455894476507e-05,
      "loss": 4.2252,
      "step": 2476
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7813272877164055e-05,
      "loss": 4.1679,
      "step": 2477
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7807089859851606e-05,
      "loss": 4.3909,
      "step": 2478
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.780090684253916e-05,
      "loss": 3.8942,
      "step": 2479
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.779472382522671e-05,
      "loss": 4.334,
      "step": 2480
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7788540807914266e-05,
      "loss": 3.8052,
      "step": 2481
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7782357790601813e-05,
      "loss": 3.7321,
      "step": 2482
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7776174773289364e-05,
      "loss": 3.7147,
      "step": 2483
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.776999175597692e-05,
      "loss": 3.7862,
      "step": 2484
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.776380873866447e-05,
      "loss": 4.4665,
      "step": 2485
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7757625721352017e-05,
      "loss": 4.4924,
      "step": 2486
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.775144270403957e-05,
      "loss": 4.427,
      "step": 2487
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7745259686727123e-05,
      "loss": 4.4029,
      "step": 2488
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7739076669414677e-05,
      "loss": 4.4305,
      "step": 2489
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7732893652102228e-05,
      "loss": 4.6435,
      "step": 2490
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7726710634789776e-05,
      "loss": 4.4085,
      "step": 2491
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.772052761747733e-05,
      "loss": 4.403,
      "step": 2492
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.771434460016488e-05,
      "loss": 4.1336,
      "step": 2493
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7708161582852432e-05,
      "loss": 4.2987,
      "step": 2494
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7701978565539983e-05,
      "loss": 4.2483,
      "step": 2495
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7695795548227534e-05,
      "loss": 4.1605,
      "step": 2496
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7689612530915088e-05,
      "loss": 4.2315,
      "step": 2497
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.768342951360264e-05,
      "loss": 4.3733,
      "step": 2498
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.767724649629019e-05,
      "loss": 4.2016,
      "step": 2499
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.767106347897774e-05,
      "loss": 4.2545,
      "step": 2500
    },
    {
      "epoch": 1.4,
      "eval_loss": 4.285419940948486,
      "eval_runtime": 1298.052,
      "eval_samples_per_second": 2.035,
      "eval_steps_per_second": 0.255,
      "eval_wer": 1.3726809135464413,
      "step": 2500
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7664880461665292e-05,
      "loss": 4.1986,
      "step": 2501
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7658697444352846e-05,
      "loss": 4.4,
      "step": 2502
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7652514427040397e-05,
      "loss": 4.3718,
      "step": 2503
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.764633140972795e-05,
      "loss": 4.0193,
      "step": 2504
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.76401483924155e-05,
      "loss": 4.0458,
      "step": 2505
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.763396537510305e-05,
      "loss": 4.2418,
      "step": 2506
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.76277823577906e-05,
      "loss": 4.1722,
      "step": 2507
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7621599340478156e-05,
      "loss": 4.1772,
      "step": 2508
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7615416323165703e-05,
      "loss": 4.1955,
      "step": 2509
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7609233305853258e-05,
      "loss": 4.3231,
      "step": 2510
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.760305028854081e-05,
      "loss": 4.2997,
      "step": 2511
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.759686727122836e-05,
      "loss": 4.1671,
      "step": 2512
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7590684253915914e-05,
      "loss": 4.2173,
      "step": 2513
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.758450123660346e-05,
      "loss": 4.2508,
      "step": 2514
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7578318219291013e-05,
      "loss": 4.491,
      "step": 2515
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7572135201978567e-05,
      "loss": 4.1818,
      "step": 2516
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7565952184666118e-05,
      "loss": 4.0395,
      "step": 2517
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.755976916735367e-05,
      "loss": 4.2959,
      "step": 2518
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.755358615004122e-05,
      "loss": 4.3542,
      "step": 2519
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.754740313272877e-05,
      "loss": 4.2297,
      "step": 2520
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7541220115416325e-05,
      "loss": 4.2225,
      "step": 2521
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7535037098103876e-05,
      "loss": 4.0389,
      "step": 2522
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7528854080791424e-05,
      "loss": 3.8859,
      "step": 2523
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7522671063478978e-05,
      "loss": 4.0724,
      "step": 2524
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.751648804616653e-05,
      "loss": 3.9321,
      "step": 2525
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7510305028854084e-05,
      "loss": 4.2598,
      "step": 2526
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.750412201154163e-05,
      "loss": 4.1198,
      "step": 2527
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7497938994229182e-05,
      "loss": 3.9216,
      "step": 2528
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7491755976916736e-05,
      "loss": 3.9903,
      "step": 2529
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7485572959604287e-05,
      "loss": 3.8567,
      "step": 2530
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7479389942291842e-05,
      "loss": 4.1779,
      "step": 2531
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.747320692497939e-05,
      "loss": 4.0533,
      "step": 2532
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.746702390766694e-05,
      "loss": 4.0777,
      "step": 2533
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7460840890354495e-05,
      "loss": 3.8639,
      "step": 2534
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7454657873042046e-05,
      "loss": 4.4227,
      "step": 2535
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7448474855729597e-05,
      "loss": 4.4103,
      "step": 2536
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7442291838417148e-05,
      "loss": 4.5095,
      "step": 2537
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.74361088211047e-05,
      "loss": 4.4176,
      "step": 2538
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7429925803792253e-05,
      "loss": 4.2301,
      "step": 2539
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7423742786479804e-05,
      "loss": 4.5079,
      "step": 2540
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.741755976916735e-05,
      "loss": 4.165,
      "step": 2541
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7411376751854906e-05,
      "loss": 4.3366,
      "step": 2542
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7405193734542457e-05,
      "loss": 4.2542,
      "step": 2543
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7399010717230008e-05,
      "loss": 4.3102,
      "step": 2544
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7392827699917562e-05,
      "loss": 4.2589,
      "step": 2545
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.738664468260511e-05,
      "loss": 4.3051,
      "step": 2546
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7380461665292664e-05,
      "loss": 4.5145,
      "step": 2547
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7374278647980215e-05,
      "loss": 4.3711,
      "step": 2548
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7368095630667766e-05,
      "loss": 4.0877,
      "step": 2549
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7361912613355317e-05,
      "loss": 4.3201,
      "step": 2550
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7355729596042868e-05,
      "loss": 4.1348,
      "step": 2551
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.734954657873042e-05,
      "loss": 4.3546,
      "step": 2552
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7343363561417974e-05,
      "loss": 4.2685,
      "step": 2553
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7337180544105525e-05,
      "loss": 4.4586,
      "step": 2554
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7330997526793076e-05,
      "loss": 4.2951,
      "step": 2555
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7324814509480627e-05,
      "loss": 3.9929,
      "step": 2556
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7318631492168178e-05,
      "loss": 4.2721,
      "step": 2557
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7312448474855732e-05,
      "loss": 4.2175,
      "step": 2558
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.730626545754328e-05,
      "loss": 4.3772,
      "step": 2559
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7300082440230834e-05,
      "loss": 4.3163,
      "step": 2560
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7293899422918385e-05,
      "loss": 4.1754,
      "step": 2561
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7287716405605936e-05,
      "loss": 4.1743,
      "step": 2562
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.728153338829349e-05,
      "loss": 4.3295,
      "step": 2563
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7275350370981038e-05,
      "loss": 4.1884,
      "step": 2564
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.726916735366859e-05,
      "loss": 4.1461,
      "step": 2565
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7262984336356143e-05,
      "loss": 4.2266,
      "step": 2566
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7256801319043694e-05,
      "loss": 4.0221,
      "step": 2567
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.725061830173125e-05,
      "loss": 4.2719,
      "step": 2568
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7244435284418796e-05,
      "loss": 4.2932,
      "step": 2569
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7238252267106347e-05,
      "loss": 4.2428,
      "step": 2570
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.72320692497939e-05,
      "loss": 4.2728,
      "step": 2571
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7225886232481452e-05,
      "loss": 3.9828,
      "step": 2572
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7219703215169e-05,
      "loss": 4.3732,
      "step": 2573
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7213520197856554e-05,
      "loss": 4.214,
      "step": 2574
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7207337180544105e-05,
      "loss": 3.9273,
      "step": 2575
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.720115416323166e-05,
      "loss": 4.1424,
      "step": 2576
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.719497114591921e-05,
      "loss": 4.494,
      "step": 2577
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7188788128606758e-05,
      "loss": 3.8186,
      "step": 2578
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7182605111294313e-05,
      "loss": 3.8573,
      "step": 2579
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7176422093981864e-05,
      "loss": 4.0333,
      "step": 2580
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7170239076669415e-05,
      "loss": 4.2597,
      "step": 2581
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7164056059356966e-05,
      "loss": 4.123,
      "step": 2582
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7157873042044517e-05,
      "loss": 3.7763,
      "step": 2583
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.715169002473207e-05,
      "loss": 3.4945,
      "step": 2584
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7145507007419622e-05,
      "loss": 4.4201,
      "step": 2585
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7139323990107173e-05,
      "loss": 4.4015,
      "step": 2586
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7133140972794724e-05,
      "loss": 4.5805,
      "step": 2587
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7126957955482275e-05,
      "loss": 4.4938,
      "step": 2588
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7120774938169826e-05,
      "loss": 4.3222,
      "step": 2589
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.711459192085738e-05,
      "loss": 4.4219,
      "step": 2590
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7108408903544928e-05,
      "loss": 4.2205,
      "step": 2591
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7102225886232482e-05,
      "loss": 4.5414,
      "step": 2592
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7096042868920033e-05,
      "loss": 4.3272,
      "step": 2593
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7089859851607584e-05,
      "loss": 4.2782,
      "step": 2594
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.708367683429514e-05,
      "loss": 4.379,
      "step": 2595
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7077493816982686e-05,
      "loss": 4.1702,
      "step": 2596
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.707131079967024e-05,
      "loss": 4.407,
      "step": 2597
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.706512778235779e-05,
      "loss": 4.1123,
      "step": 2598
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7058944765045342e-05,
      "loss": 4.1668,
      "step": 2599
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7052761747732897e-05,
      "loss": 4.3475,
      "step": 2600
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7046578730420444e-05,
      "loss": 4.2732,
      "step": 2601
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7040395713107995e-05,
      "loss": 4.3251,
      "step": 2602
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.703421269579555e-05,
      "loss": 4.2796,
      "step": 2603
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.70280296784831e-05,
      "loss": 4.2008,
      "step": 2604
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7021846661170652e-05,
      "loss": 4.2654,
      "step": 2605
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7015663643858203e-05,
      "loss": 4.1169,
      "step": 2606
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7009480626545754e-05,
      "loss": 4.0355,
      "step": 2607
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7003297609233308e-05,
      "loss": 4.1266,
      "step": 2608
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.699711459192086e-05,
      "loss": 4.0125,
      "step": 2609
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.6990931574608407e-05,
      "loss": 4.2624,
      "step": 2610
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.698474855729596e-05,
      "loss": 4.3438,
      "step": 2611
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.6978565539983512e-05,
      "loss": 4.1295,
      "step": 2612
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.6972382522671066e-05,
      "loss": 4.4738,
      "step": 2613
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.6966199505358614e-05,
      "loss": 4.2452,
      "step": 2614
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.6960016488046165e-05,
      "loss": 4.2864,
      "step": 2615
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.695383347073372e-05,
      "loss": 4.1448,
      "step": 2616
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.694765045342127e-05,
      "loss": 4.1715,
      "step": 2617
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.694146743610882e-05,
      "loss": 3.9446,
      "step": 2618
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.6935284418796372e-05,
      "loss": 4.3315,
      "step": 2619
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.6929101401483923e-05,
      "loss": 4.0915,
      "step": 2620
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.6922918384171478e-05,
      "loss": 4.2424,
      "step": 2621
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.691673536685903e-05,
      "loss": 4.2598,
      "step": 2622
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.691055234954658e-05,
      "loss": 4.2149,
      "step": 2623
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.690436933223413e-05,
      "loss": 4.0598,
      "step": 2624
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.689818631492168e-05,
      "loss": 4.1193,
      "step": 2625
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.6892003297609236e-05,
      "loss": 4.1106,
      "step": 2626
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.6885820280296787e-05,
      "loss": 4.1355,
      "step": 2627
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.6879637262984334e-05,
      "loss": 4.2824,
      "step": 2628
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.687345424567189e-05,
      "loss": 3.8357,
      "step": 2629
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.686727122835944e-05,
      "loss": 4.2337,
      "step": 2630
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.686108821104699e-05,
      "loss": 3.8679,
      "step": 2631
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6854905193734545e-05,
      "loss": 3.984,
      "step": 2632
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6848722176422093e-05,
      "loss": 3.6762,
      "step": 2633
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6842539159109647e-05,
      "loss": 3.5179,
      "step": 2634
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6836356141797198e-05,
      "loss": 4.4414,
      "step": 2635
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.683017312448475e-05,
      "loss": 4.4705,
      "step": 2636
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.68239901071723e-05,
      "loss": 4.5374,
      "step": 2637
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.681780708985985e-05,
      "loss": 4.2592,
      "step": 2638
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6811624072547402e-05,
      "loss": 4.2361,
      "step": 2639
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6805441055234956e-05,
      "loss": 4.4692,
      "step": 2640
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6799258037922507e-05,
      "loss": 4.244,
      "step": 2641
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.679307502061006e-05,
      "loss": 4.2471,
      "step": 2642
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.678689200329761e-05,
      "loss": 4.2696,
      "step": 2643
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.678070898598516e-05,
      "loss": 4.2247,
      "step": 2644
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6774525968672715e-05,
      "loss": 4.3537,
      "step": 2645
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6768342951360262e-05,
      "loss": 4.2009,
      "step": 2646
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6762159934047813e-05,
      "loss": 4.1874,
      "step": 2647
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.6755976916735368e-05,
      "loss": 4.1464,
      "step": 2648
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.674979389942292e-05,
      "loss": 4.2696,
      "step": 2649
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6743610882110473e-05,
      "loss": 4.2201,
      "step": 2650
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.673742786479802e-05,
      "loss": 4.2896,
      "step": 2651
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.673124484748557e-05,
      "loss": 4.1986,
      "step": 2652
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6725061830173126e-05,
      "loss": 4.1596,
      "step": 2653
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6718878812860677e-05,
      "loss": 4.2074,
      "step": 2654
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.671269579554823e-05,
      "loss": 4.1952,
      "step": 2655
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.670651277823578e-05,
      "loss": 4.3791,
      "step": 2656
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.670032976092333e-05,
      "loss": 4.2954,
      "step": 2657
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6694146743610884e-05,
      "loss": 4.3303,
      "step": 2658
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6687963726298435e-05,
      "loss": 4.3528,
      "step": 2659
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6681780708985983e-05,
      "loss": 4.0705,
      "step": 2660
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6675597691673537e-05,
      "loss": 4.276,
      "step": 2661
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6669414674361088e-05,
      "loss": 4.2037,
      "step": 2662
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6663231657048642e-05,
      "loss": 4.268,
      "step": 2663
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6657048639736193e-05,
      "loss": 4.2835,
      "step": 2664
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.665086562242374e-05,
      "loss": 4.1794,
      "step": 2665
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6644682605111295e-05,
      "loss": 4.053,
      "step": 2666
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.6638499587798846e-05,
      "loss": 4.5928,
      "step": 2667
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6632316570486397e-05,
      "loss": 4.3238,
      "step": 2668
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.662613355317395e-05,
      "loss": 4.3227,
      "step": 2669
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.66199505358615e-05,
      "loss": 4.4503,
      "step": 2670
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6613767518549054e-05,
      "loss": 4.3852,
      "step": 2671
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6607584501236605e-05,
      "loss": 4.3503,
      "step": 2672
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6601401483924156e-05,
      "loss": 4.3371,
      "step": 2673
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6595218466611707e-05,
      "loss": 4.2294,
      "step": 2674
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6589035449299258e-05,
      "loss": 4.1902,
      "step": 2675
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.658285243198681e-05,
      "loss": 4.2176,
      "step": 2676
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6576669414674363e-05,
      "loss": 4.1135,
      "step": 2677
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.657048639736191e-05,
      "loss": 4.6214,
      "step": 2678
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6564303380049465e-05,
      "loss": 4.2115,
      "step": 2679
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6558120362737016e-05,
      "loss": 3.8039,
      "step": 2680
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6551937345424567e-05,
      "loss": 4.0036,
      "step": 2681
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.654575432811212e-05,
      "loss": 3.8901,
      "step": 2682
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.653957131079967e-05,
      "loss": 3.813,
      "step": 2683
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6533388293487223e-05,
      "loss": 2.9486,
      "step": 2684
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6527205276174774e-05,
      "loss": 4.6034,
      "step": 2685
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6521022258862325e-05,
      "loss": 4.2657,
      "step": 2686
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.651483924154988e-05,
      "loss": 4.33,
      "step": 2687
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6508656224237427e-05,
      "loss": 4.39,
      "step": 2688
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6502473206924978e-05,
      "loss": 4.2966,
      "step": 2689
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6496290189612533e-05,
      "loss": 4.0558,
      "step": 2690
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6490107172300084e-05,
      "loss": 4.3153,
      "step": 2691
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6483924154987634e-05,
      "loss": 4.4081,
      "step": 2692
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6477741137675185e-05,
      "loss": 4.2337,
      "step": 2693
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6471558120362736e-05,
      "loss": 4.2286,
      "step": 2694
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.646537510305029e-05,
      "loss": 4.4527,
      "step": 2695
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6459192085737842e-05,
      "loss": 4.2729,
      "step": 2696
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.645300906842539e-05,
      "loss": 4.0783,
      "step": 2697
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6446826051112944e-05,
      "loss": 4.039,
      "step": 2698
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6440643033800495e-05,
      "loss": 4.4649,
      "step": 2699
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.643446001648805e-05,
      "loss": 4.2963,
      "step": 2700
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6428276999175597e-05,
      "loss": 4.2189,
      "step": 2701
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6422093981863148e-05,
      "loss": 4.0137,
      "step": 2702
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6415910964550702e-05,
      "loss": 4.0932,
      "step": 2703
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6409727947238253e-05,
      "loss": 4.3175,
      "step": 2704
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6403544929925804e-05,
      "loss": 3.9888,
      "step": 2705
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6397361912613355e-05,
      "loss": 4.3169,
      "step": 2706
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6391178895300906e-05,
      "loss": 4.1197,
      "step": 2707
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.638499587798846e-05,
      "loss": 4.1855,
      "step": 2708
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.637881286067601e-05,
      "loss": 4.3958,
      "step": 2709
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.637262984336356e-05,
      "loss": 4.1741,
      "step": 2710
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6366446826051113e-05,
      "loss": 4.0879,
      "step": 2711
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6360263808738664e-05,
      "loss": 4.2647,
      "step": 2712
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6354080791426215e-05,
      "loss": 3.9999,
      "step": 2713
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.634789777411377e-05,
      "loss": 4.527,
      "step": 2714
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6341714756801317e-05,
      "loss": 4.0495,
      "step": 2715
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.633553173948887e-05,
      "loss": 4.2187,
      "step": 2716
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6329348722176423e-05,
      "loss": 4.2233,
      "step": 2717
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6323165704863974e-05,
      "loss": 4.2101,
      "step": 2718
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6316982687551528e-05,
      "loss": 3.9726,
      "step": 2719
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6310799670239076e-05,
      "loss": 4.0664,
      "step": 2720
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.630461665292663e-05,
      "loss": 4.3419,
      "step": 2721
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.629843363561418e-05,
      "loss": 4.3223,
      "step": 2722
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6292250618301732e-05,
      "loss": 4.225,
      "step": 2723
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6286067600989283e-05,
      "loss": 4.215,
      "step": 2724
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6279884583676834e-05,
      "loss": 3.9043,
      "step": 2725
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6273701566364385e-05,
      "loss": 4.3459,
      "step": 2726
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.626751854905194e-05,
      "loss": 4.1475,
      "step": 2727
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.626133553173949e-05,
      "loss": 4.3397,
      "step": 2728
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.625515251442704e-05,
      "loss": 4.3814,
      "step": 2729
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6248969497114592e-05,
      "loss": 4.2507,
      "step": 2730
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6242786479802143e-05,
      "loss": 3.7195,
      "step": 2731
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6236603462489697e-05,
      "loss": 3.7232,
      "step": 2732
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6230420445177245e-05,
      "loss": 4.4279,
      "step": 2733
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6224237427864796e-05,
      "loss": 3.2542,
      "step": 2734
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.621805441055235e-05,
      "loss": 4.7698,
      "step": 2735
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.62118713932399e-05,
      "loss": 4.4305,
      "step": 2736
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6205688375927456e-05,
      "loss": 4.5195,
      "step": 2737
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6199505358615003e-05,
      "loss": 4.2907,
      "step": 2738
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6193322341302554e-05,
      "loss": 4.1508,
      "step": 2739
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.618713932399011e-05,
      "loss": 4.2914,
      "step": 2740
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.618095630667766e-05,
      "loss": 4.1646,
      "step": 2741
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.617477328936521e-05,
      "loss": 4.3057,
      "step": 2742
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.616859027205276e-05,
      "loss": 4.3926,
      "step": 2743
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6162407254740313e-05,
      "loss": 4.2873,
      "step": 2744
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6156224237427867e-05,
      "loss": 4.4019,
      "step": 2745
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6150041220115418e-05,
      "loss": 4.2473,
      "step": 2746
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6143858202802966e-05,
      "loss": 4.2136,
      "step": 2747
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.613767518549052e-05,
      "loss": 4.5129,
      "step": 2748
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.613149216817807e-05,
      "loss": 4.2914,
      "step": 2749
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6125309150865625e-05,
      "loss": 4.2655,
      "step": 2750
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6119126133553176e-05,
      "loss": 4.1749,
      "step": 2751
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6112943116240724e-05,
      "loss": 4.1343,
      "step": 2752
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6106760098928278e-05,
      "loss": 4.0879,
      "step": 2753
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.610057708161583e-05,
      "loss": 4.4338,
      "step": 2754
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.609439406430338e-05,
      "loss": 4.2849,
      "step": 2755
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.608821104699093e-05,
      "loss": 4.048,
      "step": 2756
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.6082028029678482e-05,
      "loss": 4.4615,
      "step": 2757
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.6075845012366037e-05,
      "loss": 4.2117,
      "step": 2758
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.6069661995053587e-05,
      "loss": 4.0982,
      "step": 2759
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.606347897774114e-05,
      "loss": 4.2342,
      "step": 2760
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.605729596042869e-05,
      "loss": 4.2956,
      "step": 2761
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.605111294311624e-05,
      "loss": 4.3673,
      "step": 2762
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.604492992580379e-05,
      "loss": 4.0659,
      "step": 2763
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.6038746908491346e-05,
      "loss": 4.4019,
      "step": 2764
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.6032563891178893e-05,
      "loss": 3.9276,
      "step": 2765
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.6026380873866448e-05,
      "loss": 3.7877,
      "step": 2766
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.6020197856554e-05,
      "loss": 4.113,
      "step": 2767
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.601401483924155e-05,
      "loss": 4.3423,
      "step": 2768
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.6007831821929104e-05,
      "loss": 3.8302,
      "step": 2769
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.600164880461665e-05,
      "loss": 4.0544,
      "step": 2770
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.5995465787304203e-05,
      "loss": 4.3802,
      "step": 2771
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.5989282769991757e-05,
      "loss": 4.1181,
      "step": 2772
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.5983099752679308e-05,
      "loss": 3.9081,
      "step": 2773
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.5976916735366862e-05,
      "loss": 4.2043,
      "step": 2774
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.597073371805441e-05,
      "loss": 3.9365,
      "step": 2775
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.596455070074196e-05,
      "loss": 4.1364,
      "step": 2776
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5958367683429515e-05,
      "loss": 3.9035,
      "step": 2777
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5952184666117066e-05,
      "loss": 4.1836,
      "step": 2778
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5946001648804617e-05,
      "loss": 4.1008,
      "step": 2779
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5939818631492168e-05,
      "loss": 3.9598,
      "step": 2780
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.593363561417972e-05,
      "loss": 3.4528,
      "step": 2781
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5927452596867274e-05,
      "loss": 3.6809,
      "step": 2782
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5921269579554825e-05,
      "loss": 3.5157,
      "step": 2783
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5915086562242372e-05,
      "loss": 2.9907,
      "step": 2784
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5908903544929927e-05,
      "loss": 4.2793,
      "step": 2785
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5902720527617478e-05,
      "loss": 4.6121,
      "step": 2786
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5896537510305032e-05,
      "loss": 4.5669,
      "step": 2787
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.589035449299258e-05,
      "loss": 4.3124,
      "step": 2788
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.588417147568013e-05,
      "loss": 4.215,
      "step": 2789
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5877988458367685e-05,
      "loss": 4.2358,
      "step": 2790
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.5871805441055236e-05,
      "loss": 4.093,
      "step": 2791
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5865622423742787e-05,
      "loss": 4.1389,
      "step": 2792
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5859439406430338e-05,
      "loss": 4.2391,
      "step": 2793
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.585325638911789e-05,
      "loss": 4.2089,
      "step": 2794
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5847073371805443e-05,
      "loss": 4.3195,
      "step": 2795
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5840890354492994e-05,
      "loss": 4.2453,
      "step": 2796
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5834707337180542e-05,
      "loss": 4.2693,
      "step": 2797
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5828524319868096e-05,
      "loss": 4.2354,
      "step": 2798
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5822341302555647e-05,
      "loss": 4.3942,
      "step": 2799
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5816158285243198e-05,
      "loss": 4.1374,
      "step": 2800
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5809975267930752e-05,
      "loss": 4.2303,
      "step": 2801
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.58037922506183e-05,
      "loss": 4.008,
      "step": 2802
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5797609233305854e-05,
      "loss": 4.2131,
      "step": 2803
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5791426215993405e-05,
      "loss": 4.2347,
      "step": 2804
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5785243198680956e-05,
      "loss": 4.1587,
      "step": 2805
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.577906018136851e-05,
      "loss": 4.284,
      "step": 2806
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5772877164056058e-05,
      "loss": 4.2012,
      "step": 2807
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5766694146743613e-05,
      "loss": 4.1572,
      "step": 2808
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.5760511129431164e-05,
      "loss": 4.0167,
      "step": 2809
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5754328112118715e-05,
      "loss": 3.9715,
      "step": 2810
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5748145094806266e-05,
      "loss": 3.9684,
      "step": 2811
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5741962077493817e-05,
      "loss": 4.2325,
      "step": 2812
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5735779060181368e-05,
      "loss": 4.4054,
      "step": 2813
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5729596042868922e-05,
      "loss": 4.2864,
      "step": 2814
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5723413025556473e-05,
      "loss": 4.1916,
      "step": 2815
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5717230008244024e-05,
      "loss": 4.0466,
      "step": 2816
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5711046990931575e-05,
      "loss": 4.0889,
      "step": 2817
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5704863973619126e-05,
      "loss": 4.0504,
      "step": 2818
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.569868095630668e-05,
      "loss": 4.1955,
      "step": 2819
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5692497938994228e-05,
      "loss": 4.171,
      "step": 2820
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.568631492168178e-05,
      "loss": 3.9869,
      "step": 2821
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5680131904369333e-05,
      "loss": 4.2285,
      "step": 2822
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5673948887056884e-05,
      "loss": 4.1968,
      "step": 2823
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.566776586974444e-05,
      "loss": 4.1246,
      "step": 2824
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5661582852431986e-05,
      "loss": 4.2992,
      "step": 2825
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.5655399835119537e-05,
      "loss": 4.0777,
      "step": 2826
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.564921681780709e-05,
      "loss": 4.2286,
      "step": 2827
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5643033800494642e-05,
      "loss": 4.4091,
      "step": 2828
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5636850783182193e-05,
      "loss": 4.0752,
      "step": 2829
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5630667765869744e-05,
      "loss": 3.8249,
      "step": 2830
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5624484748557295e-05,
      "loss": 3.351,
      "step": 2831
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.561830173124485e-05,
      "loss": 3.87,
      "step": 2832
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.56121187139324e-05,
      "loss": 3.4686,
      "step": 2833
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.560593569661995e-05,
      "loss": 3.8364,
      "step": 2834
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5599752679307503e-05,
      "loss": 4.4273,
      "step": 2835
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5593569661995054e-05,
      "loss": 4.3255,
      "step": 2836
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5587386644682605e-05,
      "loss": 3.9776,
      "step": 2837
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.558120362737016e-05,
      "loss": 4.2243,
      "step": 2838
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5575020610057707e-05,
      "loss": 4.1581,
      "step": 2839
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.556883759274526e-05,
      "loss": 4.1246,
      "step": 2840
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5562654575432812e-05,
      "loss": 4.5817,
      "step": 2841
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5556471558120363e-05,
      "loss": 4.3445,
      "step": 2842
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5550288540807914e-05,
      "loss": 4.255,
      "step": 2843
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.5544105523495465e-05,
      "loss": 4.3445,
      "step": 2844
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.553792250618302e-05,
      "loss": 4.1238,
      "step": 2845
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.553173948887057e-05,
      "loss": 4.3434,
      "step": 2846
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.552555647155812e-05,
      "loss": 4.3626,
      "step": 2847
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5519373454245672e-05,
      "loss": 4.3558,
      "step": 2848
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5513190436933223e-05,
      "loss": 4.0099,
      "step": 2849
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5507007419620774e-05,
      "loss": 4.3009,
      "step": 2850
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.550082440230833e-05,
      "loss": 4.2028,
      "step": 2851
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5494641384995876e-05,
      "loss": 4.1376,
      "step": 2852
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.548845836768343e-05,
      "loss": 4.2073,
      "step": 2853
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.548227535037098e-05,
      "loss": 4.0284,
      "step": 2854
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5476092333058532e-05,
      "loss": 4.0113,
      "step": 2855
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5469909315746087e-05,
      "loss": 4.1193,
      "step": 2856
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5463726298433634e-05,
      "loss": 4.1986,
      "step": 2857
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5457543281121185e-05,
      "loss": 4.3166,
      "step": 2858
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.545136026380874e-05,
      "loss": 4.0927,
      "step": 2859
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.544517724649629e-05,
      "loss": 4.2425,
      "step": 2860
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5438994229183845e-05,
      "loss": 4.2766,
      "step": 2861
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5432811211871393e-05,
      "loss": 4.2384,
      "step": 2862
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.5426628194558944e-05,
      "loss": 4.5416,
      "step": 2863
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5420445177246498e-05,
      "loss": 4.2604,
      "step": 2864
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.541426215993405e-05,
      "loss": 4.0291,
      "step": 2865
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5408079142621597e-05,
      "loss": 4.1707,
      "step": 2866
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.540189612530915e-05,
      "loss": 4.1082,
      "step": 2867
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5395713107996702e-05,
      "loss": 4.0295,
      "step": 2868
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5389530090684256e-05,
      "loss": 4.2471,
      "step": 2869
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5383347073371807e-05,
      "loss": 3.977,
      "step": 2870
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5377164056059355e-05,
      "loss": 4.0727,
      "step": 2871
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.537098103874691e-05,
      "loss": 4.0387,
      "step": 2872
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.536479802143446e-05,
      "loss": 4.2129,
      "step": 2873
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5358615004122015e-05,
      "loss": 4.3525,
      "step": 2874
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5352431986809562e-05,
      "loss": 4.3455,
      "step": 2875
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5346248969497113e-05,
      "loss": 3.8848,
      "step": 2876
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.5340065952184668e-05,
      "loss": 4.01,
      "step": 2877
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.533388293487222e-05,
      "loss": 4.2728,
      "step": 2878
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.532769991755977e-05,
      "loss": 4.0075,
      "step": 2879
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.532151690024732e-05,
      "loss": 3.6339,
      "step": 2880
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.531533388293487e-05,
      "loss": 3.8841,
      "step": 2881
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5309150865622426e-05,
      "loss": 4.3633,
      "step": 2882
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5302967848309977e-05,
      "loss": 4.1696,
      "step": 2883
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5296784830997524e-05,
      "loss": 3.351,
      "step": 2884
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.529060181368508e-05,
      "loss": 4.5043,
      "step": 2885
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.528441879637263e-05,
      "loss": 4.3513,
      "step": 2886
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.527823577906018e-05,
      "loss": 4.3882,
      "step": 2887
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5272052761747735e-05,
      "loss": 4.1955,
      "step": 2888
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5265869744435283e-05,
      "loss": 4.27,
      "step": 2889
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5259686727122837e-05,
      "loss": 4.2587,
      "step": 2890
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5253503709810388e-05,
      "loss": 4.139,
      "step": 2891
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.524732069249794e-05,
      "loss": 4.3188,
      "step": 2892
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5241137675185492e-05,
      "loss": 4.1353,
      "step": 2893
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5234954657873041e-05,
      "loss": 4.1848,
      "step": 2894
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5228771640560594e-05,
      "loss": 4.1043,
      "step": 2895
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5222588623248146e-05,
      "loss": 4.1288,
      "step": 2896
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5216405605935697e-05,
      "loss": 4.3717,
      "step": 2897
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5210222588623247e-05,
      "loss": 4.3347,
      "step": 2898
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.52040395713108e-05,
      "loss": 4.024,
      "step": 2899
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5197856553998352e-05,
      "loss": 4.2044,
      "step": 2900
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5191673536685903e-05,
      "loss": 4.2209,
      "step": 2901
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5185490519373456e-05,
      "loss": 4.1069,
      "step": 2902
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5179307502061005e-05,
      "loss": 4.2025,
      "step": 2903
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5173124484748558e-05,
      "loss": 3.999,
      "step": 2904
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.516694146743611e-05,
      "loss": 4.2662,
      "step": 2905
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5160758450123661e-05,
      "loss": 4.3506,
      "step": 2906
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.515457543281121e-05,
      "loss": 4.2405,
      "step": 2907
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5148392415498763e-05,
      "loss": 4.3202,
      "step": 2908
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5142209398186316e-05,
      "loss": 4.1422,
      "step": 2909
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5136026380873867e-05,
      "loss": 4.0772,
      "step": 2910
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.512984336356142e-05,
      "loss": 4.2019,
      "step": 2911
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5123660346248969e-05,
      "loss": 3.9939,
      "step": 2912
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5117477328936522e-05,
      "loss": 4.2628,
      "step": 2913
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5111294311624073e-05,
      "loss": 4.2309,
      "step": 2914
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5105111294311625e-05,
      "loss": 4.1258,
      "step": 2915
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5098928276999175e-05,
      "loss": 3.8422,
      "step": 2916
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5092745259686727e-05,
      "loss": 3.9504,
      "step": 2917
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5086562242374278e-05,
      "loss": 4.0253,
      "step": 2918
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5080379225061831e-05,
      "loss": 4.0868,
      "step": 2919
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5074196207749384e-05,
      "loss": 4.3649,
      "step": 2920
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5068013190436933e-05,
      "loss": 4.1448,
      "step": 2921
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5061830173124484e-05,
      "loss": 3.965,
      "step": 2922
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5055647155812036e-05,
      "loss": 4.1985,
      "step": 2923
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5049464138499589e-05,
      "loss": 3.9565,
      "step": 2924
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5043281121187142e-05,
      "loss": 4.2067,
      "step": 2925
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.503709810387469e-05,
      "loss": 4.2176,
      "step": 2926
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5030915086562242e-05,
      "loss": 4.2448,
      "step": 2927
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5024732069249795e-05,
      "loss": 4.2834,
      "step": 2928
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5018549051937347e-05,
      "loss": 4.0741,
      "step": 2929
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5012366034624895e-05,
      "loss": 4.32,
      "step": 2930
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5006183017312448e-05,
      "loss": 3.6015,
      "step": 2931
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.5e-05,
      "loss": 3.7616,
      "step": 2932
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4993816982687553e-05,
      "loss": 3.596,
      "step": 2933
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4987633965375104e-05,
      "loss": 3.1912,
      "step": 2934
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4981450948062655e-05,
      "loss": 4.1353,
      "step": 2935
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4975267930750206e-05,
      "loss": 4.5063,
      "step": 2936
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4969084913437759e-05,
      "loss": 4.3537,
      "step": 2937
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.496290189612531e-05,
      "loss": 4.1915,
      "step": 2938
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.495671887881286e-05,
      "loss": 4.313,
      "step": 2939
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4950535861500413e-05,
      "loss": 4.2107,
      "step": 2940
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4944352844187964e-05,
      "loss": 4.1947,
      "step": 2941
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4938169826875517e-05,
      "loss": 4.1077,
      "step": 2942
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4931986809563066e-05,
      "loss": 4.3402,
      "step": 2943
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4925803792250619e-05,
      "loss": 4.2059,
      "step": 2944
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.491962077493817e-05,
      "loss": 4.2922,
      "step": 2945
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4913437757625723e-05,
      "loss": 4.285,
      "step": 2946
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4907254740313272e-05,
      "loss": 4.2971,
      "step": 2947
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4901071723000825e-05,
      "loss": 4.0198,
      "step": 2948
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4894888705688377e-05,
      "loss": 4.26,
      "step": 2949
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4888705688375928e-05,
      "loss": 4.0321,
      "step": 2950
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.488252267106348e-05,
      "loss": 4.1348,
      "step": 2951
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.487633965375103e-05,
      "loss": 4.1694,
      "step": 2952
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4870156636438583e-05,
      "loss": 4.055,
      "step": 2953
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4863973619126134e-05,
      "loss": 4.224,
      "step": 2954
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4857790601813685e-05,
      "loss": 4.0515,
      "step": 2955
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4851607584501237e-05,
      "loss": 4.5049,
      "step": 2956
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4845424567188788e-05,
      "loss": 3.9303,
      "step": 2957
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4839241549876341e-05,
      "loss": 4.1992,
      "step": 2958
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.483305853256389e-05,
      "loss": 4.0873,
      "step": 2959
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4826875515251443e-05,
      "loss": 4.1052,
      "step": 2960
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4820692497938994e-05,
      "loss": 4.1194,
      "step": 2961
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4814509480626547e-05,
      "loss": 3.9903,
      "step": 2962
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4808326463314098e-05,
      "loss": 4.0375,
      "step": 2963
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4802143446001649e-05,
      "loss": 4.3868,
      "step": 2964
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4795960428689201e-05,
      "loss": 4.0953,
      "step": 2965
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4789777411376752e-05,
      "loss": 3.788,
      "step": 2966
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4783594394064305e-05,
      "loss": 4.0399,
      "step": 2967
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4777411376751854e-05,
      "loss": 3.992,
      "step": 2968
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4771228359439407e-05,
      "loss": 4.0709,
      "step": 2969
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4765045342126958e-05,
      "loss": 4.1377,
      "step": 2970
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.475886232481451e-05,
      "loss": 4.1535,
      "step": 2971
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4752679307502062e-05,
      "loss": 4.0796,
      "step": 2972
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4746496290189613e-05,
      "loss": 4.0212,
      "step": 2973
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4740313272877165e-05,
      "loss": 4.1681,
      "step": 2974
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4734130255564716e-05,
      "loss": 4.0387,
      "step": 2975
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4727947238252267e-05,
      "loss": 4.0804,
      "step": 2976
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4721764220939818e-05,
      "loss": 4.0558,
      "step": 2977
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4715581203627371e-05,
      "loss": 3.9262,
      "step": 2978
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4709398186314922e-05,
      "loss": 4.0391,
      "step": 2979
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4703215169002473e-05,
      "loss": 3.7893,
      "step": 2980
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4697032151690026e-05,
      "loss": 3.8089,
      "step": 2981
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4690849134377577e-05,
      "loss": 3.7232,
      "step": 2982
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.468466611706513e-05,
      "loss": 3.815,
      "step": 2983
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4678483099752678e-05,
      "loss": 3.8162,
      "step": 2984
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4672300082440231e-05,
      "loss": 4.4215,
      "step": 2985
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4666117065127782e-05,
      "loss": 4.5309,
      "step": 2986
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4659934047815335e-05,
      "loss": 4.3066,
      "step": 2987
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4653751030502886e-05,
      "loss": 4.1668,
      "step": 2988
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4647568013190437e-05,
      "loss": 4.3701,
      "step": 2989
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.464138499587799e-05,
      "loss": 4.252,
      "step": 2990
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.463520197856554e-05,
      "loss": 4.4135,
      "step": 2991
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4629018961253091e-05,
      "loss": 4.2929,
      "step": 2992
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4622835943940642e-05,
      "loss": 4.4878,
      "step": 2993
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4616652926628195e-05,
      "loss": 4.2243,
      "step": 2994
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4610469909315746e-05,
      "loss": 4.3174,
      "step": 2995
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4604286892003299e-05,
      "loss": 4.2829,
      "step": 2996
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.459810387469085e-05,
      "loss": 4.4249,
      "step": 2997
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.45919208573784e-05,
      "loss": 4.1488,
      "step": 2998
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4585737840065953e-05,
      "loss": 4.0875,
      "step": 2999
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4579554822753504e-05,
      "loss": 4.0861,
      "step": 3000
    },
    {
      "epoch": 1.68,
      "eval_loss": 4.145115852355957,
      "eval_runtime": 1302.2349,
      "eval_samples_per_second": 2.029,
      "eval_steps_per_second": 0.254,
      "eval_wer": 1.3667480207155187,
      "step": 3000
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4573371805441055e-05,
      "loss": 4.1945,
      "step": 3001
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4567188788128606e-05,
      "loss": 4.2306,
      "step": 3002
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4561005770816159e-05,
      "loss": 4.1444,
      "step": 3003
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4554822753503712e-05,
      "loss": 4.092,
      "step": 3004
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4548639736191261e-05,
      "loss": 4.0804,
      "step": 3005
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.4542456718878814e-05,
      "loss": 4.3426,
      "step": 3006
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4536273701566365e-05,
      "loss": 4.4381,
      "step": 3007
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4530090684253917e-05,
      "loss": 4.026,
      "step": 3008
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4523907666941467e-05,
      "loss": 4.0121,
      "step": 3009
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.451772464962902e-05,
      "loss": 4.1906,
      "step": 3010
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.451154163231657e-05,
      "loss": 4.3087,
      "step": 3011
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4505358615004123e-05,
      "loss": 3.9579,
      "step": 3012
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4499175597691674e-05,
      "loss": 4.3683,
      "step": 3013
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4492992580379225e-05,
      "loss": 4.1974,
      "step": 3014
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4486809563066778e-05,
      "loss": 3.8764,
      "step": 3015
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4480626545754329e-05,
      "loss": 4.0954,
      "step": 3016
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.447444352844188e-05,
      "loss": 4.4107,
      "step": 3017
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.446826051112943e-05,
      "loss": 3.8527,
      "step": 3018
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4462077493816983e-05,
      "loss": 4.3511,
      "step": 3019
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4455894476504536e-05,
      "loss": 4.0624,
      "step": 3020
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4449711459192085e-05,
      "loss": 4.2219,
      "step": 3021
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4443528441879638e-05,
      "loss": 3.9617,
      "step": 3022
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.4437345424567189e-05,
      "loss": 4.4768,
      "step": 3023
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4431162407254741e-05,
      "loss": 3.6804,
      "step": 3024
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4424979389942292e-05,
      "loss": 4.1279,
      "step": 3025
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4418796372629843e-05,
      "loss": 4.1068,
      "step": 3026
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4412613355317394e-05,
      "loss": 3.9718,
      "step": 3027
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4406430338004947e-05,
      "loss": 4.1759,
      "step": 3028
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.44002473206925e-05,
      "loss": 4.0296,
      "step": 3029
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4394064303380049e-05,
      "loss": 4.022,
      "step": 3030
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4387881286067602e-05,
      "loss": 3.9234,
      "step": 3031
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4381698268755153e-05,
      "loss": 3.8164,
      "step": 3032
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4375515251442705e-05,
      "loss": 3.9073,
      "step": 3033
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4369332234130255e-05,
      "loss": 3.7572,
      "step": 3034
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4363149216817807e-05,
      "loss": 4.3506,
      "step": 3035
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.435696619950536e-05,
      "loss": 4.213,
      "step": 3036
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4350783182192911e-05,
      "loss": 4.4036,
      "step": 3037
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4344600164880462e-05,
      "loss": 4.2137,
      "step": 3038
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4338417147568013e-05,
      "loss": 4.3021,
      "step": 3039
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4332234130255566e-05,
      "loss": 4.2075,
      "step": 3040
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.4326051112943117e-05,
      "loss": 4.2829,
      "step": 3041
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4319868095630668e-05,
      "loss": 4.0218,
      "step": 3042
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.431368507831822e-05,
      "loss": 4.2819,
      "step": 3043
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4307502061005771e-05,
      "loss": 4.1061,
      "step": 3044
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4301319043693324e-05,
      "loss": 3.9341,
      "step": 3045
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4295136026380873e-05,
      "loss": 4.0812,
      "step": 3046
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4288953009068426e-05,
      "loss": 4.1321,
      "step": 3047
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4282769991755977e-05,
      "loss": 4.2509,
      "step": 3048
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.427658697444353e-05,
      "loss": 4.0738,
      "step": 3049
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4270403957131079e-05,
      "loss": 4.0188,
      "step": 3050
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4264220939818631e-05,
      "loss": 3.9902,
      "step": 3051
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4258037922506184e-05,
      "loss": 4.2119,
      "step": 3052
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4251854905193735e-05,
      "loss": 4.3547,
      "step": 3053
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4245671887881286e-05,
      "loss": 4.1492,
      "step": 3054
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4239488870568837e-05,
      "loss": 4.3357,
      "step": 3055
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.423330585325639e-05,
      "loss": 4.0821,
      "step": 3056
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.422712283594394e-05,
      "loss": 4.3212,
      "step": 3057
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4220939818631493e-05,
      "loss": 4.2405,
      "step": 3058
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.4214756801319044e-05,
      "loss": 3.9496,
      "step": 3059
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4208573784006595e-05,
      "loss": 4.2031,
      "step": 3060
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4202390766694148e-05,
      "loss": 4.0614,
      "step": 3061
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4196207749381699e-05,
      "loss": 4.1683,
      "step": 3062
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.419002473206925e-05,
      "loss": 4.1532,
      "step": 3063
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4183841714756801e-05,
      "loss": 4.1567,
      "step": 3064
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4177658697444354e-05,
      "loss": 3.8904,
      "step": 3065
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4171475680131905e-05,
      "loss": 4.1486,
      "step": 3066
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4165292662819456e-05,
      "loss": 3.9976,
      "step": 3067
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4159109645507008e-05,
      "loss": 4.2848,
      "step": 3068
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.415292662819456e-05,
      "loss": 4.1425,
      "step": 3069
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4146743610882112e-05,
      "loss": 4.3636,
      "step": 3070
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4140560593569661e-05,
      "loss": 4.1787,
      "step": 3071
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4134377576257214e-05,
      "loss": 4.1993,
      "step": 3072
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4128194558944765e-05,
      "loss": 4.208,
      "step": 3073
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4122011541632318e-05,
      "loss": 4.3413,
      "step": 3074
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4115828524319869e-05,
      "loss": 3.9559,
      "step": 3075
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.410964550700742e-05,
      "loss": 4.0265,
      "step": 3076
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.4103462489694972e-05,
      "loss": 4.0896,
      "step": 3077
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4097279472382523e-05,
      "loss": 4.0376,
      "step": 3078
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4091096455070074e-05,
      "loss": 4.1261,
      "step": 3079
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4084913437757625e-05,
      "loss": 3.8991,
      "step": 3080
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4078730420445178e-05,
      "loss": 3.7385,
      "step": 3081
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4072547403132729e-05,
      "loss": 3.6267,
      "step": 3082
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.406636438582028e-05,
      "loss": 3.6119,
      "step": 3083
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4060181368507833e-05,
      "loss": 3.4424,
      "step": 3084
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4053998351195383e-05,
      "loss": 4.4205,
      "step": 3085
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4047815333882936e-05,
      "loss": 4.2877,
      "step": 3086
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4041632316570487e-05,
      "loss": 4.29,
      "step": 3087
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4035449299258038e-05,
      "loss": 4.4428,
      "step": 3088
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4029266281945589e-05,
      "loss": 4.3212,
      "step": 3089
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4023083264633142e-05,
      "loss": 4.0473,
      "step": 3090
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4016900247320694e-05,
      "loss": 4.4141,
      "step": 3091
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4010717230008244e-05,
      "loss": 4.1388,
      "step": 3092
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.4004534212695796e-05,
      "loss": 4.1268,
      "step": 3093
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.3998351195383347e-05,
      "loss": 4.3344,
      "step": 3094
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.39921681780709e-05,
      "loss": 4.2701,
      "step": 3095
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.398598516075845e-05,
      "loss": 4.1308,
      "step": 3096
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3979802143446002e-05,
      "loss": 4.2722,
      "step": 3097
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3973619126133553e-05,
      "loss": 3.9561,
      "step": 3098
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3967436108821106e-05,
      "loss": 3.8945,
      "step": 3099
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3961253091508657e-05,
      "loss": 4.3542,
      "step": 3100
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3955070074196208e-05,
      "loss": 4.2304,
      "step": 3101
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.394888705688376e-05,
      "loss": 4.1423,
      "step": 3102
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3942704039571311e-05,
      "loss": 4.2266,
      "step": 3103
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3936521022258862e-05,
      "loss": 4.0619,
      "step": 3104
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3930338004946413e-05,
      "loss": 4.1542,
      "step": 3105
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3924154987633966e-05,
      "loss": 4.3153,
      "step": 3106
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3917971970321519e-05,
      "loss": 4.1683,
      "step": 3107
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3911788953009068e-05,
      "loss": 3.9303,
      "step": 3108
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.390560593569662e-05,
      "loss": 4.3673,
      "step": 3109
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3899422918384172e-05,
      "loss": 4.3,
      "step": 3110
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3893239901071724e-05,
      "loss": 4.1603,
      "step": 3111
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3887056883759274e-05,
      "loss": 4.3549,
      "step": 3112
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.3880873866446826e-05,
      "loss": 3.8911,
      "step": 3113
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3874690849134377e-05,
      "loss": 4.2072,
      "step": 3114
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.386850783182193e-05,
      "loss": 4.3282,
      "step": 3115
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.386232481450948e-05,
      "loss": 4.2116,
      "step": 3116
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3856141797197032e-05,
      "loss": 3.9267,
      "step": 3117
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3849958779884584e-05,
      "loss": 4.0095,
      "step": 3118
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3843775762572135e-05,
      "loss": 3.9595,
      "step": 3119
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3837592745259688e-05,
      "loss": 4.109,
      "step": 3120
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3831409727947237e-05,
      "loss": 4.3212,
      "step": 3121
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.382522671063479e-05,
      "loss": 4.134,
      "step": 3122
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3819043693322343e-05,
      "loss": 3.9496,
      "step": 3123
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3812860676009894e-05,
      "loss": 4.0897,
      "step": 3124
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3806677658697445e-05,
      "loss": 4.1412,
      "step": 3125
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3800494641384996e-05,
      "loss": 4.0131,
      "step": 3126
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3794311624072548e-05,
      "loss": 3.9323,
      "step": 3127
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.37881286067601e-05,
      "loss": 3.8201,
      "step": 3128
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.378194558944765e-05,
      "loss": 3.9265,
      "step": 3129
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.3775762572135201e-05,
      "loss": 3.6559,
      "step": 3130
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3769579554822754e-05,
      "loss": 3.8717,
      "step": 3131
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3763396537510307e-05,
      "loss": 3.9135,
      "step": 3132
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3757213520197856e-05,
      "loss": 3.5853,
      "step": 3133
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3751030502885409e-05,
      "loss": 3.1327,
      "step": 3134
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.374484748557296e-05,
      "loss": 4.5419,
      "step": 3135
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3738664468260512e-05,
      "loss": 4.5119,
      "step": 3136
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3732481450948062e-05,
      "loss": 4.2772,
      "step": 3137
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3726298433635614e-05,
      "loss": 4.2287,
      "step": 3138
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3720115416323167e-05,
      "loss": 4.1673,
      "step": 3139
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3713932399010718e-05,
      "loss": 4.2734,
      "step": 3140
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3707749381698269e-05,
      "loss": 4.3763,
      "step": 3141
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.370156636438582e-05,
      "loss": 4.3496,
      "step": 3142
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3695383347073373e-05,
      "loss": 4.2574,
      "step": 3143
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3689200329760924e-05,
      "loss": 4.3432,
      "step": 3144
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3683017312448475e-05,
      "loss": 4.0717,
      "step": 3145
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3676834295136026e-05,
      "loss": 3.958,
      "step": 3146
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3670651277823578e-05,
      "loss": 4.1502,
      "step": 3147
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.3664468260511131e-05,
      "loss": 4.202,
      "step": 3148
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.365828524319868e-05,
      "loss": 4.1706,
      "step": 3149
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3652102225886233e-05,
      "loss": 4.2435,
      "step": 3150
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3645919208573784e-05,
      "loss": 4.1787,
      "step": 3151
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3639736191261336e-05,
      "loss": 4.4726,
      "step": 3152
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3633553173948887e-05,
      "loss": 4.1557,
      "step": 3153
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3627370156636438e-05,
      "loss": 4.4571,
      "step": 3154
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3621187139323991e-05,
      "loss": 4.1549,
      "step": 3155
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3615004122011542e-05,
      "loss": 3.9417,
      "step": 3156
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3608821104699095e-05,
      "loss": 4.2305,
      "step": 3157
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3602638087386644e-05,
      "loss": 4.1789,
      "step": 3158
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3596455070074197e-05,
      "loss": 4.0515,
      "step": 3159
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3590272052761748e-05,
      "loss": 3.9816,
      "step": 3160
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.35840890354493e-05,
      "loss": 4.1659,
      "step": 3161
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3577906018136851e-05,
      "loss": 4.0308,
      "step": 3162
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3571723000824402e-05,
      "loss": 4.1174,
      "step": 3163
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3565539983511955e-05,
      "loss": 4.397,
      "step": 3164
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3559356966199506e-05,
      "loss": 4.4189,
      "step": 3165
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.3553173948887057e-05,
      "loss": 4.1194,
      "step": 3166
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3546990931574608e-05,
      "loss": 4.3686,
      "step": 3167
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.354080791426216e-05,
      "loss": 4.0899,
      "step": 3168
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3534624896949712e-05,
      "loss": 3.9566,
      "step": 3169
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3528441879637263e-05,
      "loss": 3.9763,
      "step": 3170
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3522258862324815e-05,
      "loss": 4.2514,
      "step": 3171
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3516075845012366e-05,
      "loss": 4.0037,
      "step": 3172
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3509892827699919e-05,
      "loss": 4.1145,
      "step": 3173
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3503709810387468e-05,
      "loss": 4.0614,
      "step": 3174
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3497526793075021e-05,
      "loss": 4.0303,
      "step": 3175
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3491343775762572e-05,
      "loss": 4.3113,
      "step": 3176
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3485160758450125e-05,
      "loss": 4.2339,
      "step": 3177
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3478977741137676e-05,
      "loss": 4.0166,
      "step": 3178
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3472794723825227e-05,
      "loss": 3.9666,
      "step": 3179
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.346661170651278e-05,
      "loss": 3.7303,
      "step": 3180
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.346042868920033e-05,
      "loss": 3.8114,
      "step": 3181
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3454245671887883e-05,
      "loss": 3.3849,
      "step": 3182
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3448062654575432e-05,
      "loss": 3.8452,
      "step": 3183
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.3441879637262985e-05,
      "loss": 3.5676,
      "step": 3184
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3435696619950536e-05,
      "loss": 4.2673,
      "step": 3185
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3429513602638088e-05,
      "loss": 4.2668,
      "step": 3186
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.342333058532564e-05,
      "loss": 4.4668,
      "step": 3187
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.341714756801319e-05,
      "loss": 4.2151,
      "step": 3188
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3410964550700743e-05,
      "loss": 4.4091,
      "step": 3189
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3404781533388294e-05,
      "loss": 4.4535,
      "step": 3190
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3398598516075845e-05,
      "loss": 4.2487,
      "step": 3191
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3392415498763396e-05,
      "loss": 4.4033,
      "step": 3192
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3386232481450949e-05,
      "loss": 4.1752,
      "step": 3193
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3380049464138501e-05,
      "loss": 4.1372,
      "step": 3194
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.337386644682605e-05,
      "loss": 4.4211,
      "step": 3195
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3367683429513603e-05,
      "loss": 4.0268,
      "step": 3196
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3361500412201154e-05,
      "loss": 4.1662,
      "step": 3197
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3355317394888707e-05,
      "loss": 4.3509,
      "step": 3198
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3349134377576256e-05,
      "loss": 4.2473,
      "step": 3199
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3342951360263809e-05,
      "loss": 4.2631,
      "step": 3200
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.333676834295136e-05,
      "loss": 4.0832,
      "step": 3201
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.3330585325638913e-05,
      "loss": 4.0965,
      "step": 3202
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3324402308326464e-05,
      "loss": 4.1048,
      "step": 3203
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3318219291014015e-05,
      "loss": 4.1729,
      "step": 3204
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3312036273701567e-05,
      "loss": 4.1943,
      "step": 3205
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3305853256389118e-05,
      "loss": 4.2249,
      "step": 3206
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.329967023907667e-05,
      "loss": 4.2822,
      "step": 3207
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.329348722176422e-05,
      "loss": 4.2064,
      "step": 3208
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3287304204451773e-05,
      "loss": 4.1938,
      "step": 3209
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3281121187139326e-05,
      "loss": 4.2583,
      "step": 3210
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3274938169826875e-05,
      "loss": 4.2352,
      "step": 3211
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3268755152514428e-05,
      "loss": 4.1041,
      "step": 3212
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3262572135201979e-05,
      "loss": 4.0721,
      "step": 3213
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3256389117889531e-05,
      "loss": 4.2045,
      "step": 3214
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3250206100577082e-05,
      "loss": 4.3136,
      "step": 3215
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3244023083264633e-05,
      "loss": 4.1401,
      "step": 3216
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3237840065952184e-05,
      "loss": 3.9923,
      "step": 3217
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3231657048639737e-05,
      "loss": 4.0945,
      "step": 3218
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.322547403132729e-05,
      "loss": 4.0908,
      "step": 3219
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.3219291014014839e-05,
      "loss": 4.1494,
      "step": 3220
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3213107996702391e-05,
      "loss": 4.1678,
      "step": 3221
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3206924979389942e-05,
      "loss": 4.0647,
      "step": 3222
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3200741962077495e-05,
      "loss": 3.7733,
      "step": 3223
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3194558944765044e-05,
      "loss": 3.9265,
      "step": 3224
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3188375927452597e-05,
      "loss": 4.0496,
      "step": 3225
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.318219291014015e-05,
      "loss": 4.272,
      "step": 3226
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.31760098928277e-05,
      "loss": 4.1726,
      "step": 3227
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3169826875515252e-05,
      "loss": 3.9339,
      "step": 3228
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3163643858202803e-05,
      "loss": 3.8594,
      "step": 3229
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3157460840890355e-05,
      "loss": 3.7403,
      "step": 3230
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3151277823577906e-05,
      "loss": 3.6672,
      "step": 3231
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3145094806265457e-05,
      "loss": 3.9157,
      "step": 3232
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3138911788953008e-05,
      "loss": 3.4891,
      "step": 3233
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3132728771640561e-05,
      "loss": 3.2927,
      "step": 3234
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3126545754328114e-05,
      "loss": 4.5406,
      "step": 3235
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3120362737015663e-05,
      "loss": 4.3653,
      "step": 3236
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.3114179719703216e-05,
      "loss": 4.1397,
      "step": 3237
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3107996702390767e-05,
      "loss": 4.1058,
      "step": 3238
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.310181368507832e-05,
      "loss": 4.0365,
      "step": 3239
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3095630667765869e-05,
      "loss": 4.2574,
      "step": 3240
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3089447650453421e-05,
      "loss": 4.1451,
      "step": 3241
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3083264633140974e-05,
      "loss": 4.0961,
      "step": 3242
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3077081615828525e-05,
      "loss": 4.3506,
      "step": 3243
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3070898598516078e-05,
      "loss": 4.4207,
      "step": 3244
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3064715581203627e-05,
      "loss": 4.2059,
      "step": 3245
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.305853256389118e-05,
      "loss": 4.1931,
      "step": 3246
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.305234954657873e-05,
      "loss": 4.1626,
      "step": 3247
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3046166529266283e-05,
      "loss": 4.1485,
      "step": 3248
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3039983511953832e-05,
      "loss": 3.9737,
      "step": 3249
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3033800494641385e-05,
      "loss": 4.1426,
      "step": 3250
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3027617477328938e-05,
      "loss": 4.1189,
      "step": 3251
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3021434460016489e-05,
      "loss": 4.2628,
      "step": 3252
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.301525144270404e-05,
      "loss": 4.3597,
      "step": 3253
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.300906842539159e-05,
      "loss": 4.0046,
      "step": 3254
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.3002885408079143e-05,
      "loss": 4.3708,
      "step": 3255
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2996702390766694e-05,
      "loss": 3.9641,
      "step": 3256
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2990519373454245e-05,
      "loss": 4.0974,
      "step": 3257
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2984336356141798e-05,
      "loss": 4.1568,
      "step": 3258
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2978153338829349e-05,
      "loss": 4.3123,
      "step": 3259
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2971970321516902e-05,
      "loss": 4.5175,
      "step": 3260
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2965787304204451e-05,
      "loss": 4.4574,
      "step": 3261
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2959604286892004e-05,
      "loss": 4.2115,
      "step": 3262
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2953421269579555e-05,
      "loss": 4.185,
      "step": 3263
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2947238252267107e-05,
      "loss": 3.9986,
      "step": 3264
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2941055234954658e-05,
      "loss": 4.0336,
      "step": 3265
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.293487221764221e-05,
      "loss": 3.853,
      "step": 3266
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2928689200329762e-05,
      "loss": 3.9201,
      "step": 3267
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2922506183017313e-05,
      "loss": 4.1114,
      "step": 3268
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2916323165704864e-05,
      "loss": 4.1002,
      "step": 3269
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2910140148392415e-05,
      "loss": 4.3972,
      "step": 3270
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2903957131079968e-05,
      "loss": 4.031,
      "step": 3271
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.2897774113767519e-05,
      "loss": 3.9509,
      "step": 3272
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.289159109645507e-05,
      "loss": 3.9418,
      "step": 3273
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2885408079142622e-05,
      "loss": 4.164,
      "step": 3274
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2879225061830173e-05,
      "loss": 4.1265,
      "step": 3275
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2873042044517726e-05,
      "loss": 4.0726,
      "step": 3276
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2866859027205277e-05,
      "loss": 3.6959,
      "step": 3277
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2860676009892828e-05,
      "loss": 4.0657,
      "step": 3278
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2854492992580379e-05,
      "loss": 4.0127,
      "step": 3279
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2848309975267932e-05,
      "loss": 3.655,
      "step": 3280
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2842126957955484e-05,
      "loss": 3.8832,
      "step": 3281
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2835943940643033e-05,
      "loss": 3.2824,
      "step": 3282
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2829760923330586e-05,
      "loss": 3.8595,
      "step": 3283
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2823577906018137e-05,
      "loss": 3.0577,
      "step": 3284
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.281739488870569e-05,
      "loss": 4.1369,
      "step": 3285
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2811211871393239e-05,
      "loss": 4.3956,
      "step": 3286
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2805028854080792e-05,
      "loss": 4.1238,
      "step": 3287
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2798845836768343e-05,
      "loss": 3.9904,
      "step": 3288
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2792662819455895e-05,
      "loss": 4.2932,
      "step": 3289
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2786479802143446e-05,
      "loss": 4.2258,
      "step": 3290
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.2780296784830997e-05,
      "loss": 4.2589,
      "step": 3291
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.277411376751855e-05,
      "loss": 4.1049,
      "step": 3292
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2767930750206101e-05,
      "loss": 4.1665,
      "step": 3293
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2761747732893652e-05,
      "loss": 4.0414,
      "step": 3294
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2755564715581203e-05,
      "loss": 4.2484,
      "step": 3295
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2749381698268756e-05,
      "loss": 4.1918,
      "step": 3296
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2743198680956308e-05,
      "loss": 4.2059,
      "step": 3297
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2737015663643858e-05,
      "loss": 4.214,
      "step": 3298
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.273083264633141e-05,
      "loss": 4.3154,
      "step": 3299
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2724649629018961e-05,
      "loss": 4.24,
      "step": 3300
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2718466611706514e-05,
      "loss": 4.3463,
      "step": 3301
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2712283594394063e-05,
      "loss": 4.1188,
      "step": 3302
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2706100577081616e-05,
      "loss": 4.2377,
      "step": 3303
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2699917559769167e-05,
      "loss": 4.0333,
      "step": 3304
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.269373454245672e-05,
      "loss": 4.2246,
      "step": 3305
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.268755152514427e-05,
      "loss": 3.8953,
      "step": 3306
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2681368507831822e-05,
      "loss": 4.121,
      "step": 3307
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2675185490519374e-05,
      "loss": 4.2564,
      "step": 3308
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.2669002473206925e-05,
      "loss": 4.0064,
      "step": 3309
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2662819455894478e-05,
      "loss": 4.496,
      "step": 3310
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2656636438582027e-05,
      "loss": 4.1666,
      "step": 3311
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.265045342126958e-05,
      "loss": 3.9836,
      "step": 3312
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2644270403957133e-05,
      "loss": 4.0059,
      "step": 3313
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2638087386644683e-05,
      "loss": 4.1968,
      "step": 3314
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2631904369332234e-05,
      "loss": 4.2143,
      "step": 3315
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2625721352019785e-05,
      "loss": 4.2838,
      "step": 3316
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2619538334707338e-05,
      "loss": 4.0798,
      "step": 3317
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2613355317394889e-05,
      "loss": 3.9092,
      "step": 3318
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.260717230008244e-05,
      "loss": 3.914,
      "step": 3319
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2600989282769991e-05,
      "loss": 4.0064,
      "step": 3320
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2594806265457544e-05,
      "loss": 3.8587,
      "step": 3321
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2588623248145096e-05,
      "loss": 4.1434,
      "step": 3322
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2582440230832646e-05,
      "loss": 4.1661,
      "step": 3323
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2576257213520198e-05,
      "loss": 3.8361,
      "step": 3324
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.257007419620775e-05,
      "loss": 3.975,
      "step": 3325
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2563891178895302e-05,
      "loss": 3.9816,
      "step": 3326
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.2557708161582851e-05,
      "loss": 4.0526,
      "step": 3327
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2551525144270404e-05,
      "loss": 3.7051,
      "step": 3328
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2545342126957957e-05,
      "loss": 4.0591,
      "step": 3329
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2539159109645508e-05,
      "loss": 3.9565,
      "step": 3330
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2532976092333059e-05,
      "loss": 3.9925,
      "step": 3331
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.252679307502061e-05,
      "loss": 3.8579,
      "step": 3332
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2520610057708162e-05,
      "loss": 3.6066,
      "step": 3333
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2514427040395713e-05,
      "loss": 3.2145,
      "step": 3334
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2508244023083264e-05,
      "loss": 4.2534,
      "step": 3335
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2502061005770815e-05,
      "loss": 4.489,
      "step": 3336
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2495877988458368e-05,
      "loss": 4.2962,
      "step": 3337
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.248969497114592e-05,
      "loss": 4.1319,
      "step": 3338
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2483511953833472e-05,
      "loss": 3.9872,
      "step": 3339
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2477328936521023e-05,
      "loss": 4.2424,
      "step": 3340
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2471145919208574e-05,
      "loss": 4.1677,
      "step": 3341
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2464962901896126e-05,
      "loss": 4.0903,
      "step": 3342
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2458779884583677e-05,
      "loss": 4.2539,
      "step": 3343
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.2452596867271228e-05,
      "loss": 3.9363,
      "step": 3344
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2446413849958781e-05,
      "loss": 4.4568,
      "step": 3345
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2440230832646332e-05,
      "loss": 3.9971,
      "step": 3346
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2434047815333885e-05,
      "loss": 3.8902,
      "step": 3347
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2427864798021434e-05,
      "loss": 4.2085,
      "step": 3348
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2421681780708986e-05,
      "loss": 4.0146,
      "step": 3349
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2415498763396537e-05,
      "loss": 4.2233,
      "step": 3350
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.240931574608409e-05,
      "loss": 4.1855,
      "step": 3351
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.240313272877164e-05,
      "loss": 4.109,
      "step": 3352
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2396949711459192e-05,
      "loss": 4.2391,
      "step": 3353
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2390766694146745e-05,
      "loss": 4.2613,
      "step": 3354
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2384583676834296e-05,
      "loss": 4.0771,
      "step": 3355
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2378400659521847e-05,
      "loss": 3.9732,
      "step": 3356
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2372217642209398e-05,
      "loss": 4.195,
      "step": 3357
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.236603462489695e-05,
      "loss": 3.9589,
      "step": 3358
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2359851607584501e-05,
      "loss": 4.238,
      "step": 3359
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2353668590272052e-05,
      "loss": 4.1218,
      "step": 3360
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2347485572959605e-05,
      "loss": 4.1487,
      "step": 3361
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2341302555647156e-05,
      "loss": 4.3972,
      "step": 3362
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2335119538334709e-05,
      "loss": 3.9395,
      "step": 3363
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2328936521022258e-05,
      "loss": 4.0362,
      "step": 3364
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.232275350370981e-05,
      "loss": 4.5879,
      "step": 3365
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2316570486397362e-05,
      "loss": 4.1429,
      "step": 3366
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2310387469084914e-05,
      "loss": 4.3164,
      "step": 3367
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2304204451772465e-05,
      "loss": 4.2739,
      "step": 3368
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2298021434460016e-05,
      "loss": 4.2046,
      "step": 3369
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2291838417147569e-05,
      "loss": 4.1002,
      "step": 3370
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.228565539983512e-05,
      "loss": 4.1527,
      "step": 3371
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2279472382522673e-05,
      "loss": 4.1249,
      "step": 3372
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2273289365210222e-05,
      "loss": 4.0002,
      "step": 3373
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2267106347897775e-05,
      "loss": 4.0249,
      "step": 3374
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2260923330585326e-05,
      "loss": 4.2578,
      "step": 3375
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2254740313272878e-05,
      "loss": 3.7935,
      "step": 3376
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.224855729596043e-05,
      "loss": 3.9165,
      "step": 3377
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.224237427864798e-05,
      "loss": 3.8648,
      "step": 3378
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2236191261335533e-05,
      "loss": 3.914,
      "step": 3379
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2230008244023084e-05,
      "loss": 3.637,
      "step": 3380
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2223825226710635e-05,
      "loss": 3.8682,
      "step": 3381
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2217642209398186e-05,
      "loss": 3.8085,
      "step": 3382
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2211459192085738e-05,
      "loss": 3.8315,
      "step": 3383
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2205276174773291e-05,
      "loss": 3.1408,
      "step": 3384
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.219909315746084e-05,
      "loss": 4.3078,
      "step": 3385
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2192910140148393e-05,
      "loss": 4.1892,
      "step": 3386
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2186727122835944e-05,
      "loss": 4.3329,
      "step": 3387
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2180544105523497e-05,
      "loss": 4.5227,
      "step": 3388
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2174361088211046e-05,
      "loss": 4.2544,
      "step": 3389
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2168178070898599e-05,
      "loss": 4.2,
      "step": 3390
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.216199505358615e-05,
      "loss": 4.2901,
      "step": 3391
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2155812036273702e-05,
      "loss": 3.9358,
      "step": 3392
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2149629018961253e-05,
      "loss": 4.1579,
      "step": 3393
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2143446001648804e-05,
      "loss": 4.1787,
      "step": 3394
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2137262984336357e-05,
      "loss": 4.1246,
      "step": 3395
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2131079967023908e-05,
      "loss": 4.3625,
      "step": 3396
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2124896949711459e-05,
      "loss": 4.0941,
      "step": 3397
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.211871393239901e-05,
      "loss": 4.0374,
      "step": 3398
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2112530915086563e-05,
      "loss": 4.1807,
      "step": 3399
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2106347897774115e-05,
      "loss": 4.1125,
      "step": 3400
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2100164880461666e-05,
      "loss": 3.9835,
      "step": 3401
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2093981863149217e-05,
      "loss": 4.0241,
      "step": 3402
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2087798845836768e-05,
      "loss": 4.298,
      "step": 3403
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2081615828524321e-05,
      "loss": 4.0973,
      "step": 3404
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2075432811211872e-05,
      "loss": 3.9856,
      "step": 3405
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2069249793899423e-05,
      "loss": 4.1217,
      "step": 3406
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2063066776586974e-05,
      "loss": 4.1326,
      "step": 3407
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2056883759274527e-05,
      "loss": 4.0837,
      "step": 3408
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.205070074196208e-05,
      "loss": 4.1996,
      "step": 3409
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2044517724649628e-05,
      "loss": 4.2182,
      "step": 3410
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2038334707337181e-05,
      "loss": 4.1191,
      "step": 3411
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2032151690024732e-05,
      "loss": 4.3251,
      "step": 3412
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2025968672712285e-05,
      "loss": 3.9755,
      "step": 3413
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2019785655399834e-05,
      "loss": 4.0473,
      "step": 3414
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.2013602638087387e-05,
      "loss": 4.0224,
      "step": 3415
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.200741962077494e-05,
      "loss": 3.9013,
      "step": 3416
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.200123660346249e-05,
      "loss": 4.1221,
      "step": 3417
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1995053586150041e-05,
      "loss": 3.8786,
      "step": 3418
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1988870568837592e-05,
      "loss": 4.102,
      "step": 3419
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1982687551525145e-05,
      "loss": 4.0603,
      "step": 3420
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1976504534212696e-05,
      "loss": 4.0338,
      "step": 3421
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1970321516900247e-05,
      "loss": 3.9878,
      "step": 3422
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1964138499587798e-05,
      "loss": 3.8495,
      "step": 3423
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.195795548227535e-05,
      "loss": 4.0096,
      "step": 3424
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1951772464962903e-05,
      "loss": 4.0712,
      "step": 3425
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1945589447650453e-05,
      "loss": 3.9072,
      "step": 3426
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1939406430338005e-05,
      "loss": 3.7516,
      "step": 3427
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1933223413025556e-05,
      "loss": 3.9832,
      "step": 3428
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1927040395713109e-05,
      "loss": 3.5566,
      "step": 3429
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.192085737840066e-05,
      "loss": 3.9594,
      "step": 3430
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1914674361088211e-05,
      "loss": 3.5592,
      "step": 3431
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1908491343775764e-05,
      "loss": 3.902,
      "step": 3432
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1902308326463315e-05,
      "loss": 3.4352,
      "step": 3433
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.1896125309150867e-05,
      "loss": 3.3162,
      "step": 3434
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1889942291838417e-05,
      "loss": 4.2418,
      "step": 3435
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.188375927452597e-05,
      "loss": 4.2272,
      "step": 3436
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.187757625721352e-05,
      "loss": 4.2133,
      "step": 3437
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1871393239901073e-05,
      "loss": 4.1865,
      "step": 3438
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1865210222588622e-05,
      "loss": 4.1496,
      "step": 3439
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1859027205276175e-05,
      "loss": 4.1714,
      "step": 3440
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1852844187963728e-05,
      "loss": 4.2868,
      "step": 3441
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1846661170651279e-05,
      "loss": 4.2392,
      "step": 3442
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.184047815333883e-05,
      "loss": 4.099,
      "step": 3443
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.183429513602638e-05,
      "loss": 4.1099,
      "step": 3444
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1828112118713933e-05,
      "loss": 4.1209,
      "step": 3445
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1821929101401484e-05,
      "loss": 4.1825,
      "step": 3446
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1815746084089035e-05,
      "loss": 4.4542,
      "step": 3447
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1809563066776588e-05,
      "loss": 4.0576,
      "step": 3448
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1803380049464139e-05,
      "loss": 4.0705,
      "step": 3449
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1797197032151691e-05,
      "loss": 4.1588,
      "step": 3450
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.179101401483924e-05,
      "loss": 4.0441,
      "step": 3451
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.1784830997526793e-05,
      "loss": 4.1023,
      "step": 3452
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1778647980214344e-05,
      "loss": 3.7852,
      "step": 3453
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1772464962901897e-05,
      "loss": 4.167,
      "step": 3454
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1766281945589446e-05,
      "loss": 3.9346,
      "step": 3455
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1760098928276999e-05,
      "loss": 3.9446,
      "step": 3456
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1753915910964552e-05,
      "loss": 4.2578,
      "step": 3457
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1747732893652103e-05,
      "loss": 4.2288,
      "step": 3458
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1741549876339654e-05,
      "loss": 4.1344,
      "step": 3459
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1735366859027205e-05,
      "loss": 3.9541,
      "step": 3460
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1729183841714757e-05,
      "loss": 3.9373,
      "step": 3461
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1723000824402308e-05,
      "loss": 3.837,
      "step": 3462
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1716817807089861e-05,
      "loss": 4.3243,
      "step": 3463
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1710634789777412e-05,
      "loss": 3.8258,
      "step": 3464
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1704451772464963e-05,
      "loss": 4.2064,
      "step": 3465
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1698268755152516e-05,
      "loss": 4.222,
      "step": 3466
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1692085737840067e-05,
      "loss": 4.0097,
      "step": 3467
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1685902720527618e-05,
      "loss": 3.7465,
      "step": 3468
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.1679719703215169e-05,
      "loss": 3.977,
      "step": 3469
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1673536685902721e-05,
      "loss": 4.3863,
      "step": 3470
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1667353668590272e-05,
      "loss": 3.8451,
      "step": 3471
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1661170651277823e-05,
      "loss": 3.998,
      "step": 3472
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1654987633965376e-05,
      "loss": 4.3673,
      "step": 3473
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1648804616652927e-05,
      "loss": 4.1925,
      "step": 3474
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.164262159934048e-05,
      "loss": 4.0033,
      "step": 3475
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1636438582028029e-05,
      "loss": 3.7493,
      "step": 3476
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1630255564715581e-05,
      "loss": 4.0747,
      "step": 3477
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1624072547403132e-05,
      "loss": 3.9468,
      "step": 3478
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1617889530090685e-05,
      "loss": 4.1043,
      "step": 3479
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1611706512778236e-05,
      "loss": 3.9946,
      "step": 3480
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1605523495465787e-05,
      "loss": 4.0218,
      "step": 3481
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.159934047815334e-05,
      "loss": 3.571,
      "step": 3482
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.159315746084089e-05,
      "loss": 3.2834,
      "step": 3483
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1586974443528442e-05,
      "loss": 3.3912,
      "step": 3484
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1580791426215993e-05,
      "loss": 4.4032,
      "step": 3485
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1574608408903545e-05,
      "loss": 4.1683,
      "step": 3486
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1568425391591098e-05,
      "loss": 4.2426,
      "step": 3487
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1562242374278647e-05,
      "loss": 4.255,
      "step": 3488
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.15560593569662e-05,
      "loss": 4.266,
      "step": 3489
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1549876339653751e-05,
      "loss": 4.1615,
      "step": 3490
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1543693322341304e-05,
      "loss": 4.1541,
      "step": 3491
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1537510305028855e-05,
      "loss": 4.2418,
      "step": 3492
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1531327287716406e-05,
      "loss": 4.1802,
      "step": 3493
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1525144270403957e-05,
      "loss": 4.0049,
      "step": 3494
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.151896125309151e-05,
      "loss": 4.1748,
      "step": 3495
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1512778235779062e-05,
      "loss": 4.3443,
      "step": 3496
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1506595218466611e-05,
      "loss": 4.1831,
      "step": 3497
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1500412201154164e-05,
      "loss": 4.1595,
      "step": 3498
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1494229183841715e-05,
      "loss": 4.1827,
      "step": 3499
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1488046166529268e-05,
      "loss": 4.0882,
      "step": 3500
    },
    {
      "epoch": 1.96,
      "eval_loss": 4.043225288391113,
      "eval_runtime": 1298.0587,
      "eval_samples_per_second": 2.035,
      "eval_steps_per_second": 0.255,
      "eval_wer": 1.2539635295751732,
      "step": 3500
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1481863149216817e-05,
      "loss": 4.378,
      "step": 3501
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.147568013190437e-05,
      "loss": 4.0676,
      "step": 3502
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1469497114591922e-05,
      "loss": 4.1271,
      "step": 3503
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1463314097279473e-05,
      "loss": 3.8472,
      "step": 3504
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1457131079967024e-05,
      "loss": 3.8551,
      "step": 3505
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1450948062654575e-05,
      "loss": 4.1983,
      "step": 3506
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1444765045342128e-05,
      "loss": 3.7888,
      "step": 3507
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1438582028029679e-05,
      "loss": 4.1776,
      "step": 3508
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.143239901071723e-05,
      "loss": 4.2427,
      "step": 3509
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.142621599340478e-05,
      "loss": 4.1596,
      "step": 3510
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1420032976092333e-05,
      "loss": 3.8267,
      "step": 3511
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1413849958779886e-05,
      "loss": 4.173,
      "step": 3512
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1407666941467435e-05,
      "loss": 3.9666,
      "step": 3513
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1401483924154988e-05,
      "loss": 3.8883,
      "step": 3514
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1395300906842539e-05,
      "loss": 4.0776,
      "step": 3515
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1389117889530092e-05,
      "loss": 3.8966,
      "step": 3516
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1382934872217641e-05,
      "loss": 3.85,
      "step": 3517
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1376751854905194e-05,
      "loss": 4.0982,
      "step": 3518
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1370568837592746e-05,
      "loss": 3.8796,
      "step": 3519
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1364385820280297e-05,
      "loss": 3.8407,
      "step": 3520
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1358202802967848e-05,
      "loss": 4.1038,
      "step": 3521
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.13520197856554e-05,
      "loss": 4.0328,
      "step": 3522
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1345836768342952e-05,
      "loss": 4.1955,
      "step": 3523
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1339653751030503e-05,
      "loss": 4.0986,
      "step": 3524
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1333470733718056e-05,
      "loss": 4.085,
      "step": 3525
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1327287716405605e-05,
      "loss": 4.1574,
      "step": 3526
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1321104699093158e-05,
      "loss": 4.1184,
      "step": 3527
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.131492168178071e-05,
      "loss": 4.146,
      "step": 3528
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1308738664468261e-05,
      "loss": 3.8396,
      "step": 3529
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1302555647155812e-05,
      "loss": 3.5142,
      "step": 3530
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1296372629843363e-05,
      "loss": 3.4967,
      "step": 3531
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1290189612530916e-05,
      "loss": 3.7409,
      "step": 3532
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1284006595218467e-05,
      "loss": 3.2324,
      "step": 3533
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1277823577906018e-05,
      "loss": 3.039,
      "step": 3534
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.127164056059357e-05,
      "loss": 4.212,
      "step": 3535
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1265457543281122e-05,
      "loss": 4.2122,
      "step": 3536
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1259274525968674e-05,
      "loss": 4.2536,
      "step": 3537
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1253091508656224e-05,
      "loss": 4.1218,
      "step": 3538
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1246908491343776e-05,
      "loss": 4.0656,
      "step": 3539
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1240725474031327e-05,
      "loss": 3.9913,
      "step": 3540
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.123454245671888e-05,
      "loss": 4.2125,
      "step": 3541
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1228359439406429e-05,
      "loss": 4.1184,
      "step": 3542
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1222176422093982e-05,
      "loss": 4.3635,
      "step": 3543
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1215993404781534e-05,
      "loss": 4.0008,
      "step": 3544
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1209810387469085e-05,
      "loss": 3.8248,
      "step": 3545
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1203627370156636e-05,
      "loss": 4.1672,
      "step": 3546
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1197444352844187e-05,
      "loss": 4.0797,
      "step": 3547
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.119126133553174e-05,
      "loss": 4.3071,
      "step": 3548
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1185078318219291e-05,
      "loss": 3.8248,
      "step": 3549
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1178895300906842e-05,
      "loss": 4.0751,
      "step": 3550
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1172712283594395e-05,
      "loss": 4.3459,
      "step": 3551
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1166529266281946e-05,
      "loss": 3.9807,
      "step": 3552
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1160346248969498e-05,
      "loss": 4.0475,
      "step": 3553
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.115416323165705e-05,
      "loss": 4.1018,
      "step": 3554
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.11479802143446e-05,
      "loss": 4.1579,
      "step": 3555
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1141797197032151e-05,
      "loss": 4.2835,
      "step": 3556
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1135614179719704e-05,
      "loss": 4.1515,
      "step": 3557
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1129431162407255e-05,
      "loss": 4.1121,
      "step": 3558
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1123248145094806e-05,
      "loss": 4.1126,
      "step": 3559
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1117065127782359e-05,
      "loss": 3.8472,
      "step": 3560
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.111088211046991e-05,
      "loss": 3.953,
      "step": 3561
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1104699093157462e-05,
      "loss": 4.0417,
      "step": 3562
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1098516075845012e-05,
      "loss": 3.9812,
      "step": 3563
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1092333058532564e-05,
      "loss": 4.2057,
      "step": 3564
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1086150041220115e-05,
      "loss": 3.6377,
      "step": 3565
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1079967023907668e-05,
      "loss": 4.0217,
      "step": 3566
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1073784006595219e-05,
      "loss": 3.96,
      "step": 3567
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.106760098928277e-05,
      "loss": 2.7017,
      "step": 3568
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1061417971970323e-05,
      "loss": 4.0133,
      "step": 3569
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1055234954657874e-05,
      "loss": 3.9153,
      "step": 3570
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1049051937345425e-05,
      "loss": 3.9799,
      "step": 3571
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1042868920032976e-05,
      "loss": 3.8822,
      "step": 3572
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1036685902720528e-05,
      "loss": 3.8062,
      "step": 3573
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.103050288540808e-05,
      "loss": 3.8413,
      "step": 3574
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.102431986809563e-05,
      "loss": 3.8204,
      "step": 3575
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1018136850783183e-05,
      "loss": 3.9772,
      "step": 3576
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.1011953833470734e-05,
      "loss": 3.7575,
      "step": 3577
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.1005770816158286e-05,
      "loss": 3.9021,
      "step": 3578
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0999587798845836e-05,
      "loss": 3.8922,
      "step": 3579
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0993404781533388e-05,
      "loss": 3.9187,
      "step": 3580
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.098722176422094e-05,
      "loss": 3.804,
      "step": 3581
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0981038746908492e-05,
      "loss": 3.987,
      "step": 3582
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0974855729596043e-05,
      "loss": 3.8436,
      "step": 3583
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0968672712283594e-05,
      "loss": 3.9193,
      "step": 3584
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0962489694971147e-05,
      "loss": 4.0082,
      "step": 3585
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0956306677658698e-05,
      "loss": 4.1433,
      "step": 3586
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.095012366034625e-05,
      "loss": 3.9238,
      "step": 3587
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.09439406430338e-05,
      "loss": 3.7461,
      "step": 3588
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0937757625721352e-05,
      "loss": 3.8601,
      "step": 3589
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0931574608408903e-05,
      "loss": 3.74,
      "step": 3590
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0925391591096456e-05,
      "loss": 3.8799,
      "step": 3591
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0919208573784007e-05,
      "loss": 4.1242,
      "step": 3592
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.0913025556471558e-05,
      "loss": 3.9669,
      "step": 3593
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.090684253915911e-05,
      "loss": 3.8815,
      "step": 3594
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0900659521846662e-05,
      "loss": 3.9046,
      "step": 3595
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0894476504534213e-05,
      "loss": 3.6836,
      "step": 3596
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0888293487221764e-05,
      "loss": 3.9142,
      "step": 3597
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0882110469909316e-05,
      "loss": 3.9551,
      "step": 3598
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0875927452596869e-05,
      "loss": 3.7299,
      "step": 3599
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0869744435284418e-05,
      "loss": 3.7991,
      "step": 3600
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0863561417971971e-05,
      "loss": 3.7204,
      "step": 3601
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0857378400659522e-05,
      "loss": 3.773,
      "step": 3602
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0851195383347075e-05,
      "loss": 3.4854,
      "step": 3603
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0845012366034624e-05,
      "loss": 3.7731,
      "step": 3604
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0838829348722177e-05,
      "loss": 3.946,
      "step": 3605
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.083264633140973e-05,
      "loss": 3.6986,
      "step": 3606
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.082646331409728e-05,
      "loss": 3.483,
      "step": 3607
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0820280296784831e-05,
      "loss": 3.8735,
      "step": 3608
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0814097279472382e-05,
      "loss": 3.6467,
      "step": 3609
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0807914262159935e-05,
      "loss": 3.7116,
      "step": 3610
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0801731244847486e-05,
      "loss": 3.3855,
      "step": 3611
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.0795548227535037e-05,
      "loss": 3.4257,
      "step": 3612
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0789365210222588e-05,
      "loss": 3.5334,
      "step": 3613
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.078318219291014e-05,
      "loss": 3.1703,
      "step": 3614
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0776999175597693e-05,
      "loss": 2.793,
      "step": 3615
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0770816158285244e-05,
      "loss": 3.1936,
      "step": 3616
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0764633140972795e-05,
      "loss": 2.9254,
      "step": 3617
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0758450123660346e-05,
      "loss": 2.8327,
      "step": 3618
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0752267106347899e-05,
      "loss": 4.2279,
      "step": 3619
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.074608408903545e-05,
      "loss": 3.6613,
      "step": 3620
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0739901071723e-05,
      "loss": 4.0315,
      "step": 3621
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0733718054410553e-05,
      "loss": 3.9353,
      "step": 3622
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0727535037098104e-05,
      "loss": 3.9335,
      "step": 3623
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0721352019785657e-05,
      "loss": 3.9966,
      "step": 3624
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0715169002473206e-05,
      "loss": 4.0716,
      "step": 3625
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0708985985160759e-05,
      "loss": 4.1321,
      "step": 3626
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.070280296784831e-05,
      "loss": 4.0199,
      "step": 3627
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0696619950535863e-05,
      "loss": 4.2305,
      "step": 3628
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0690436933223412e-05,
      "loss": 3.9448,
      "step": 3629
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.0684253915910965e-05,
      "loss": 3.7746,
      "step": 3630
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0678070898598517e-05,
      "loss": 3.9971,
      "step": 3631
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0671887881286068e-05,
      "loss": 3.8503,
      "step": 3632
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.066570486397362e-05,
      "loss": 4.1771,
      "step": 3633
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.065952184666117e-05,
      "loss": 3.8313,
      "step": 3634
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0653338829348723e-05,
      "loss": 3.8533,
      "step": 3635
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0647155812036274e-05,
      "loss": 4.0565,
      "step": 3636
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0640972794723825e-05,
      "loss": 4.1185,
      "step": 3637
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0634789777411378e-05,
      "loss": 3.8069,
      "step": 3638
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0628606760098929e-05,
      "loss": 3.7284,
      "step": 3639
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0622423742786481e-05,
      "loss": 3.7784,
      "step": 3640
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.061624072547403e-05,
      "loss": 3.7113,
      "step": 3641
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0610057708161583e-05,
      "loss": 3.7131,
      "step": 3642
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0603874690849134e-05,
      "loss": 3.8907,
      "step": 3643
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0597691673536687e-05,
      "loss": 3.8118,
      "step": 3644
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0591508656224236e-05,
      "loss": 3.639,
      "step": 3645
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0585325638911789e-05,
      "loss": 3.9144,
      "step": 3646
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0579142621599341e-05,
      "loss": 3.7284,
      "step": 3647
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.0572959604286892e-05,
      "loss": 3.9103,
      "step": 3648
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0566776586974445e-05,
      "loss": 3.8209,
      "step": 3649
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0560593569661994e-05,
      "loss": 3.7738,
      "step": 3650
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0554410552349547e-05,
      "loss": 4.0005,
      "step": 3651
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0548227535037098e-05,
      "loss": 3.8912,
      "step": 3652
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.054204451772465e-05,
      "loss": 3.8638,
      "step": 3653
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0535861500412202e-05,
      "loss": 3.4792,
      "step": 3654
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0529678483099753e-05,
      "loss": 3.7487,
      "step": 3655
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0523495465787305e-05,
      "loss": 3.7559,
      "step": 3656
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0517312448474856e-05,
      "loss": 3.8603,
      "step": 3657
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0511129431162407e-05,
      "loss": 3.5939,
      "step": 3658
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0504946413849958e-05,
      "loss": 3.659,
      "step": 3659
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0498763396537511e-05,
      "loss": 3.9589,
      "step": 3660
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0492580379225062e-05,
      "loss": 3.5362,
      "step": 3661
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0486397361912613e-05,
      "loss": 3.6146,
      "step": 3662
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0480214344600166e-05,
      "loss": 3.5084,
      "step": 3663
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0474031327287717e-05,
      "loss": 3.3946,
      "step": 3664
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.046784830997527e-05,
      "loss": 3.6092,
      "step": 3665
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.0461665292662819e-05,
      "loss": 3.1754,
      "step": 3666
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0455482275350371e-05,
      "loss": 3.2231,
      "step": 3667
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0449299258037922e-05,
      "loss": 2.8546,
      "step": 3668
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0443116240725475e-05,
      "loss": 3.9728,
      "step": 3669
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0436933223413026e-05,
      "loss": 4.175,
      "step": 3670
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0430750206100577e-05,
      "loss": 3.7522,
      "step": 3671
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.042456718878813e-05,
      "loss": 3.8969,
      "step": 3672
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.041838417147568e-05,
      "loss": 3.8473,
      "step": 3673
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0412201154163231e-05,
      "loss": 3.6917,
      "step": 3674
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0406018136850782e-05,
      "loss": 4.1346,
      "step": 3675
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0399835119538335e-05,
      "loss": 3.9028,
      "step": 3676
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0393652102225886e-05,
      "loss": 4.1367,
      "step": 3677
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0387469084913437e-05,
      "loss": 3.9286,
      "step": 3678
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.038128606760099e-05,
      "loss": 3.8541,
      "step": 3679
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.037510305028854e-05,
      "loss": 3.9284,
      "step": 3680
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0368920032976093e-05,
      "loss": 3.9221,
      "step": 3681
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0362737015663644e-05,
      "loss": 3.9526,
      "step": 3682
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.0356553998351195e-05,
      "loss": 3.9083,
      "step": 3683
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0350370981038746e-05,
      "loss": 3.9319,
      "step": 3684
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0344187963726299e-05,
      "loss": 3.9438,
      "step": 3685
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0338004946413852e-05,
      "loss": 4.1236,
      "step": 3686
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0331821929101401e-05,
      "loss": 3.9044,
      "step": 3687
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0325638911788954e-05,
      "loss": 4.1006,
      "step": 3688
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0319455894476505e-05,
      "loss": 3.8272,
      "step": 3689
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0313272877164057e-05,
      "loss": 3.8035,
      "step": 3690
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0307089859851607e-05,
      "loss": 3.7466,
      "step": 3691
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.030090684253916e-05,
      "loss": 3.9636,
      "step": 3692
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.029472382522671e-05,
      "loss": 3.4852,
      "step": 3693
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0288540807914263e-05,
      "loss": 3.9508,
      "step": 3694
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0282357790601814e-05,
      "loss": 4.0144,
      "step": 3695
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0276174773289365e-05,
      "loss": 3.8627,
      "step": 3696
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0269991755976918e-05,
      "loss": 3.7865,
      "step": 3697
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0263808738664469e-05,
      "loss": 3.7035,
      "step": 3698
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.025762572135202e-05,
      "loss": 3.7862,
      "step": 3699
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.025144270403957e-05,
      "loss": 3.816,
      "step": 3700
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.0245259686727123e-05,
      "loss": 3.8485,
      "step": 3701
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0239076669414676e-05,
      "loss": 3.705,
      "step": 3702
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0232893652102225e-05,
      "loss": 3.7351,
      "step": 3703
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0226710634789778e-05,
      "loss": 3.6809,
      "step": 3704
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0220527617477329e-05,
      "loss": 3.7515,
      "step": 3705
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0214344600164882e-05,
      "loss": 3.8014,
      "step": 3706
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.020816158285243e-05,
      "loss": 3.7523,
      "step": 3707
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0201978565539983e-05,
      "loss": 3.7508,
      "step": 3708
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0195795548227536e-05,
      "loss": 3.6177,
      "step": 3709
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0189612530915087e-05,
      "loss": 3.7831,
      "step": 3710
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.018342951360264e-05,
      "loss": 3.7112,
      "step": 3711
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0177246496290189e-05,
      "loss": 3.3544,
      "step": 3712
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0171063478977742e-05,
      "loss": 3.827,
      "step": 3713
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0164880461665293e-05,
      "loss": 3.8615,
      "step": 3714
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0158697444352845e-05,
      "loss": 3.4459,
      "step": 3715
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0152514427040395e-05,
      "loss": 3.0922,
      "step": 3716
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0146331409727947e-05,
      "loss": 2.9292,
      "step": 3717
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.01401483924155e-05,
      "loss": 2.9151,
      "step": 3718
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.0133965375103051e-05,
      "loss": 4.113,
      "step": 3719
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0127782357790602e-05,
      "loss": 4.1976,
      "step": 3720
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0121599340478153e-05,
      "loss": 3.8528,
      "step": 3721
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0115416323165706e-05,
      "loss": 3.9848,
      "step": 3722
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0109233305853257e-05,
      "loss": 4.2218,
      "step": 3723
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0103050288540808e-05,
      "loss": 4.0074,
      "step": 3724
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.009686727122836e-05,
      "loss": 3.687,
      "step": 3725
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0090684253915911e-05,
      "loss": 3.9434,
      "step": 3726
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0084501236603464e-05,
      "loss": 3.7921,
      "step": 3727
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0078318219291013e-05,
      "loss": 3.895,
      "step": 3728
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0072135201978566e-05,
      "loss": 3.9336,
      "step": 3729
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0065952184666117e-05,
      "loss": 3.8628,
      "step": 3730
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.005976916735367e-05,
      "loss": 3.6387,
      "step": 3731
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0053586150041219e-05,
      "loss": 3.8495,
      "step": 3732
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0047403132728772e-05,
      "loss": 3.9944,
      "step": 3733
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0041220115416324e-05,
      "loss": 4.2379,
      "step": 3734
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0035037098103875e-05,
      "loss": 3.728,
      "step": 3735
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0028854080791426e-05,
      "loss": 3.8781,
      "step": 3736
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.0022671063478977e-05,
      "loss": 3.8439,
      "step": 3737
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.001648804616653e-05,
      "loss": 3.9652,
      "step": 3738
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.001030502885408e-05,
      "loss": 3.9605,
      "step": 3739
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.0004122011541632e-05,
      "loss": 3.9188,
      "step": 3740
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.997938994229184e-06,
      "loss": 3.742,
      "step": 3741
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.991755976916735e-06,
      "loss": 3.8863,
      "step": 3742
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.985572959604288e-06,
      "loss": 3.7224,
      "step": 3743
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.979389942291839e-06,
      "loss": 3.9025,
      "step": 3744
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.97320692497939e-06,
      "loss": 3.9274,
      "step": 3745
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.967023907666941e-06,
      "loss": 3.8581,
      "step": 3746
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.960840890354494e-06,
      "loss": 3.6824,
      "step": 3747
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.954657873042045e-06,
      "loss": 3.7843,
      "step": 3748
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.948474855729596e-06,
      "loss": 3.6932,
      "step": 3749
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.942291838417148e-06,
      "loss": 3.8673,
      "step": 3750
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.9361088211047e-06,
      "loss": 3.8544,
      "step": 3751
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.929925803792252e-06,
      "loss": 3.9332,
      "step": 3752
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.923742786479801e-06,
      "loss": 3.7261,
      "step": 3753
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.917559769167354e-06,
      "loss": 3.7277,
      "step": 3754
    },
    {
      "epoch": 2.1,
      "learning_rate": 9.911376751854905e-06,
      "loss": 3.9418,
      "step": 3755
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.905193734542458e-06,
      "loss": 3.7789,
      "step": 3756
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.899010717230009e-06,
      "loss": 3.6615,
      "step": 3757
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.89282769991756e-06,
      "loss": 3.9701,
      "step": 3758
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.886644682605112e-06,
      "loss": 3.9275,
      "step": 3759
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.880461665292663e-06,
      "loss": 3.8655,
      "step": 3760
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.874278647980214e-06,
      "loss": 3.6435,
      "step": 3761
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.868095630667765e-06,
      "loss": 3.6798,
      "step": 3762
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.861912613355318e-06,
      "loss": 3.5587,
      "step": 3763
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.855729596042869e-06,
      "loss": 3.3715,
      "step": 3764
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.84954657873042e-06,
      "loss": 3.4439,
      "step": 3765
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.843363561417973e-06,
      "loss": 3.4089,
      "step": 3766
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.837180544105524e-06,
      "loss": 3.2307,
      "step": 3767
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.830997526793076e-06,
      "loss": 2.7975,
      "step": 3768
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.824814509480625e-06,
      "loss": 4.0194,
      "step": 3769
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.818631492168178e-06,
      "loss": 3.8558,
      "step": 3770
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.81244847485573e-06,
      "loss": 4.0446,
      "step": 3771
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.806265457543282e-06,
      "loss": 4.2624,
      "step": 3772
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.800082440230835e-06,
      "loss": 3.9482,
      "step": 3773
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.793899422918384e-06,
      "loss": 3.8295,
      "step": 3774
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.787716405605936e-06,
      "loss": 3.91,
      "step": 3775
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.781533388293487e-06,
      "loss": 4.0903,
      "step": 3776
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.77535037098104e-06,
      "loss": 3.9544,
      "step": 3777
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.76916735366859e-06,
      "loss": 3.9804,
      "step": 3778
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.762984336356142e-06,
      "loss": 3.8744,
      "step": 3779
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.756801319043693e-06,
      "loss": 3.8296,
      "step": 3780
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.750618301731246e-06,
      "loss": 4.01,
      "step": 3781
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.744435284418797e-06,
      "loss": 4.0605,
      "step": 3782
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.738252267106348e-06,
      "loss": 3.5081,
      "step": 3783
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.7320692497939e-06,
      "loss": 3.7632,
      "step": 3784
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.725886232481451e-06,
      "loss": 4.1628,
      "step": 3785
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.719703215169002e-06,
      "loss": 3.8815,
      "step": 3786
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.713520197856553e-06,
      "loss": 3.8367,
      "step": 3787
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.707337180544106e-06,
      "loss": 3.832,
      "step": 3788
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.701154163231659e-06,
      "loss": 3.6346,
      "step": 3789
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.694971145919208e-06,
      "loss": 3.9827,
      "step": 3790
    },
    {
      "epoch": 2.12,
      "learning_rate": 9.68878812860676e-06,
      "loss": 4.0572,
      "step": 3791
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.682605111294312e-06,
      "loss": 4.1709,
      "step": 3792
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.676422093981864e-06,
      "loss": 3.9006,
      "step": 3793
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.670239076669414e-06,
      "loss": 3.9356,
      "step": 3794
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.664056059356966e-06,
      "loss": 3.724,
      "step": 3795
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.657873042044517e-06,
      "loss": 4.021,
      "step": 3796
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.65169002473207e-06,
      "loss": 3.7885,
      "step": 3797
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.645507007419621e-06,
      "loss": 3.7697,
      "step": 3798
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.639323990107172e-06,
      "loss": 3.8083,
      "step": 3799
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.633140972794725e-06,
      "loss": 3.9628,
      "step": 3800
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.626957955482276e-06,
      "loss": 3.7269,
      "step": 3801
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.620774938169827e-06,
      "loss": 4.0163,
      "step": 3802
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.614591920857377e-06,
      "loss": 3.6338,
      "step": 3803
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.60840890354493e-06,
      "loss": 4.0475,
      "step": 3804
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.602225886232483e-06,
      "loss": 3.4443,
      "step": 3805
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.596042868920034e-06,
      "loss": 3.6645,
      "step": 3806
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.589859851607585e-06,
      "loss": 3.6387,
      "step": 3807
    },
    {
      "epoch": 2.13,
      "learning_rate": 9.583676834295136e-06,
      "loss": 3.5096,
      "step": 3808
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.577493816982688e-06,
      "loss": 3.6224,
      "step": 3809
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.57131079967024e-06,
      "loss": 3.6452,
      "step": 3810
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.56512778235779e-06,
      "loss": 3.6199,
      "step": 3811
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.558944765045341e-06,
      "loss": 3.2369,
      "step": 3812
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.552761747732894e-06,
      "loss": 3.3452,
      "step": 3813
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.546578730420447e-06,
      "loss": 3.4062,
      "step": 3814
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.540395713107996e-06,
      "loss": 3.3458,
      "step": 3815
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.534212695795549e-06,
      "loss": 2.9149,
      "step": 3816
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.5280296784831e-06,
      "loss": 2.6228,
      "step": 3817
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.521846661170652e-06,
      "loss": 2.7975,
      "step": 3818
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.515663643858202e-06,
      "loss": 4.3517,
      "step": 3819
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.509480626545754e-06,
      "loss": 4.0993,
      "step": 3820
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.503297609233307e-06,
      "loss": 4.1371,
      "step": 3821
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.497114591920858e-06,
      "loss": 3.9797,
      "step": 3822
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.490931574608409e-06,
      "loss": 4.0672,
      "step": 3823
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.48474855729596e-06,
      "loss": 3.8603,
      "step": 3824
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.478565539983513e-06,
      "loss": 4.0673,
      "step": 3825
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.472382522671064e-06,
      "loss": 3.6959,
      "step": 3826
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.466199505358615e-06,
      "loss": 4.0452,
      "step": 3827
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.460016488046167e-06,
      "loss": 3.8619,
      "step": 3828
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.453833470733718e-06,
      "loss": 3.9895,
      "step": 3829
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.447650453421271e-06,
      "loss": 3.7336,
      "step": 3830
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.44146743610882e-06,
      "loss": 4.0179,
      "step": 3831
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.435284418796373e-06,
      "loss": 3.816,
      "step": 3832
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.429101401483924e-06,
      "loss": 3.7647,
      "step": 3833
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.422918384171477e-06,
      "loss": 3.8446,
      "step": 3834
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.416735366859028e-06,
      "loss": 3.8432,
      "step": 3835
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.410552349546578e-06,
      "loss": 3.8501,
      "step": 3836
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.404369332234131e-06,
      "loss": 3.7672,
      "step": 3837
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.398186314921682e-06,
      "loss": 3.8578,
      "step": 3838
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.392003297609235e-06,
      "loss": 3.728,
      "step": 3839
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.385820280296784e-06,
      "loss": 3.889,
      "step": 3840
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.379637262984337e-06,
      "loss": 3.5919,
      "step": 3841
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.373454245671888e-06,
      "loss": 3.7642,
      "step": 3842
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.36727122835944e-06,
      "loss": 3.9535,
      "step": 3843
    },
    {
      "epoch": 2.15,
      "learning_rate": 9.361088211046991e-06,
      "loss": 3.7071,
      "step": 3844
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.354905193734542e-06,
      "loss": 3.8528,
      "step": 3845
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.348722176422095e-06,
      "loss": 3.7722,
      "step": 3846
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.342539159109646e-06,
      "loss": 3.8969,
      "step": 3847
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.336356141797197e-06,
      "loss": 4.0344,
      "step": 3848
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.330173124484748e-06,
      "loss": 3.7003,
      "step": 3849
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.3239901071723e-06,
      "loss": 3.61,
      "step": 3850
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.317807089859852e-06,
      "loss": 3.6372,
      "step": 3851
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.311624072547403e-06,
      "loss": 3.8018,
      "step": 3852
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.305441055234955e-06,
      "loss": 3.6535,
      "step": 3853
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.299258037922506e-06,
      "loss": 3.5699,
      "step": 3854
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.293075020610059e-06,
      "loss": 3.7821,
      "step": 3855
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.286892003297608e-06,
      "loss": 3.6446,
      "step": 3856
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.280708985985161e-06,
      "loss": 3.7746,
      "step": 3857
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.274525968672712e-06,
      "loss": 3.5031,
      "step": 3858
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.268342951360265e-06,
      "loss": 3.8089,
      "step": 3859
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.262159934047816e-06,
      "loss": 3.5486,
      "step": 3860
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.255976916735367e-06,
      "loss": 3.4854,
      "step": 3861
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.24979389942292e-06,
      "loss": 3.1022,
      "step": 3862
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.24361088211047e-06,
      "loss": 3.3245,
      "step": 3863
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.237427864798021e-06,
      "loss": 3.4005,
      "step": 3864
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.231244847485572e-06,
      "loss": 3.2793,
      "step": 3865
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.225061830173125e-06,
      "loss": 2.9634,
      "step": 3866
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.218878812860676e-06,
      "loss": 3.3761,
      "step": 3867
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.212695795548229e-06,
      "loss": 2.7961,
      "step": 3868
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.20651277823578e-06,
      "loss": 4.0471,
      "step": 3869
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.20032976092333e-06,
      "loss": 4.1697,
      "step": 3870
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.194146743610883e-06,
      "loss": 4.0715,
      "step": 3871
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.187963726298434e-06,
      "loss": 4.0335,
      "step": 3872
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.181780708985985e-06,
      "loss": 4.0861,
      "step": 3873
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.175597691673536e-06,
      "loss": 4.3168,
      "step": 3874
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.169414674361089e-06,
      "loss": 4.0069,
      "step": 3875
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.163231657048641e-06,
      "loss": 4.0257,
      "step": 3876
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.15704863973619e-06,
      "loss": 3.9802,
      "step": 3877
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.150865622423743e-06,
      "loss": 3.9045,
      "step": 3878
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.144682605111294e-06,
      "loss": 3.9547,
      "step": 3879
    },
    {
      "epoch": 2.17,
      "learning_rate": 9.138499587798847e-06,
      "loss": 4.063,
      "step": 3880
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.132316570486396e-06,
      "loss": 3.8057,
      "step": 3881
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.126133553173949e-06,
      "loss": 3.8129,
      "step": 3882
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.1199505358615e-06,
      "loss": 3.9923,
      "step": 3883
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.113767518549053e-06,
      "loss": 4.1676,
      "step": 3884
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.107584501236604e-06,
      "loss": 3.6005,
      "step": 3885
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.101401483924155e-06,
      "loss": 3.7594,
      "step": 3886
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.095218466611707e-06,
      "loss": 3.6411,
      "step": 3887
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.089035449299258e-06,
      "loss": 3.7443,
      "step": 3888
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.08285243198681e-06,
      "loss": 3.8614,
      "step": 3889
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.07666941467436e-06,
      "loss": 3.851,
      "step": 3890
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.070486397361913e-06,
      "loss": 3.8465,
      "step": 3891
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.064303380049466e-06,
      "loss": 3.9113,
      "step": 3892
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.058120362737015e-06,
      "loss": 3.7533,
      "step": 3893
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.051937345424568e-06,
      "loss": 3.6527,
      "step": 3894
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.045754328112119e-06,
      "loss": 3.6201,
      "step": 3895
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.039571310799671e-06,
      "loss": 3.817,
      "step": 3896
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.033388293487222e-06,
      "loss": 3.6625,
      "step": 3897
    },
    {
      "epoch": 2.18,
      "learning_rate": 9.027205276174773e-06,
      "loss": 3.6825,
      "step": 3898
    },
    {
      "epoch": 2.19,
      "learning_rate": 9.021022258862324e-06,
      "loss": 3.3623,
      "step": 3899
    },
    {
      "epoch": 2.19,
      "learning_rate": 9.014839241549877e-06,
      "loss": 3.8115,
      "step": 3900
    },
    {
      "epoch": 2.19,
      "learning_rate": 9.00865622423743e-06,
      "loss": 3.648,
      "step": 3901
    },
    {
      "epoch": 2.19,
      "learning_rate": 9.002473206924979e-06,
      "loss": 3.7283,
      "step": 3902
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.996290189612531e-06,
      "loss": 3.8784,
      "step": 3903
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.990107172300082e-06,
      "loss": 3.9678,
      "step": 3904
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.983924154987635e-06,
      "loss": 3.9822,
      "step": 3905
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.977741137675184e-06,
      "loss": 3.7248,
      "step": 3906
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.971558120362737e-06,
      "loss": 3.6844,
      "step": 3907
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.96537510305029e-06,
      "loss": 3.6701,
      "step": 3908
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.95919208573784e-06,
      "loss": 3.742,
      "step": 3909
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.953009068425392e-06,
      "loss": 3.8624,
      "step": 3910
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.946826051112943e-06,
      "loss": 3.7177,
      "step": 3911
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.940643033800495e-06,
      "loss": 3.5457,
      "step": 3912
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.934460016488046e-06,
      "loss": 3.4355,
      "step": 3913
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.928276999175597e-06,
      "loss": 3.3627,
      "step": 3914
    },
    {
      "epoch": 2.19,
      "learning_rate": 8.922093981863148e-06,
      "loss": 3.2944,
      "step": 3915
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.915910964550701e-06,
      "loss": 2.8903,
      "step": 3916
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.909727947238254e-06,
      "loss": 2.9775,
      "step": 3917
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.903544929925803e-06,
      "loss": 2.6996,
      "step": 3918
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.897361912613356e-06,
      "loss": 4.3739,
      "step": 3919
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.891178895300907e-06,
      "loss": 3.9789,
      "step": 3920
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.88499587798846e-06,
      "loss": 3.7344,
      "step": 3921
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.878812860676009e-06,
      "loss": 3.8981,
      "step": 3922
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.872629843363561e-06,
      "loss": 4.0276,
      "step": 3923
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.866446826051114e-06,
      "loss": 3.9284,
      "step": 3924
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.860263808738665e-06,
      "loss": 3.7224,
      "step": 3925
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.854080791426216e-06,
      "loss": 3.7374,
      "step": 3926
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.847897774113767e-06,
      "loss": 3.8268,
      "step": 3927
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.84171475680132e-06,
      "loss": 3.8916,
      "step": 3928
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.83553173948887e-06,
      "loss": 4.0133,
      "step": 3929
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.829348722176423e-06,
      "loss": 3.8347,
      "step": 3930
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.823165704863974e-06,
      "loss": 3.8627,
      "step": 3931
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.816982687551525e-06,
      "loss": 3.7538,
      "step": 3932
    },
    {
      "epoch": 2.2,
      "learning_rate": 8.810799670239078e-06,
      "loss": 3.9303,
      "step": 3933
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.804616652926629e-06,
      "loss": 3.916,
      "step": 3934
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.79843363561418e-06,
      "loss": 3.8366,
      "step": 3935
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.79225061830173e-06,
      "loss": 3.632,
      "step": 3936
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.786067600989283e-06,
      "loss": 3.957,
      "step": 3937
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.779884583676834e-06,
      "loss": 3.8958,
      "step": 3938
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.773701566364385e-06,
      "loss": 3.7885,
      "step": 3939
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.767518549051938e-06,
      "loss": 3.5998,
      "step": 3940
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.761335531739489e-06,
      "loss": 3.7763,
      "step": 3941
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.755152514427042e-06,
      "loss": 3.6405,
      "step": 3942
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.748969497114591e-06,
      "loss": 3.9716,
      "step": 3943
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.742786479802144e-06,
      "loss": 3.928,
      "step": 3944
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.736603462489695e-06,
      "loss": 3.7928,
      "step": 3945
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.730420445177247e-06,
      "loss": 3.672,
      "step": 3946
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.724237427864798e-06,
      "loss": 3.815,
      "step": 3947
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.71805441055235e-06,
      "loss": 4.0311,
      "step": 3948
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.711871393239902e-06,
      "loss": 3.6797,
      "step": 3949
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.705688375927453e-06,
      "loss": 3.6507,
      "step": 3950
    },
    {
      "epoch": 2.21,
      "learning_rate": 8.699505358615004e-06,
      "loss": 3.6942,
      "step": 3951
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.693322341302555e-06,
      "loss": 3.9682,
      "step": 3952
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.687139323990108e-06,
      "loss": 3.5341,
      "step": 3953
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.680956306677659e-06,
      "loss": 3.6598,
      "step": 3954
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.67477328936521e-06,
      "loss": 3.622,
      "step": 3955
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.668590272052762e-06,
      "loss": 3.5031,
      "step": 3956
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.662407254740313e-06,
      "loss": 3.5614,
      "step": 3957
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.656224237427866e-06,
      "loss": 3.531,
      "step": 3958
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.650041220115417e-06,
      "loss": 3.6304,
      "step": 3959
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.643858202802968e-06,
      "loss": 3.6141,
      "step": 3960
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.637675185490519e-06,
      "loss": 3.4961,
      "step": 3961
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.631492168178072e-06,
      "loss": 3.5619,
      "step": 3962
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.625309150865624e-06,
      "loss": 3.4955,
      "step": 3963
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.619126133553174e-06,
      "loss": 3.586,
      "step": 3964
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.612943116240726e-06,
      "loss": 3.0395,
      "step": 3965
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.606760098928277e-06,
      "loss": 3.5892,
      "step": 3966
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.60057708161583e-06,
      "loss": 3.3492,
      "step": 3967
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.594394064303379e-06,
      "loss": 2.7278,
      "step": 3968
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.588211046990932e-06,
      "loss": 4.1605,
      "step": 3969
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.582028029678483e-06,
      "loss": 4.0829,
      "step": 3970
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.575845012366035e-06,
      "loss": 4.1954,
      "step": 3971
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.569661995053586e-06,
      "loss": 3.7552,
      "step": 3972
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.563478977741137e-06,
      "loss": 4.3189,
      "step": 3973
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.55729596042869e-06,
      "loss": 3.8551,
      "step": 3974
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.551112943116241e-06,
      "loss": 4.0403,
      "step": 3975
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.544929925803792e-06,
      "loss": 3.9664,
      "step": 3976
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.538746908491343e-06,
      "loss": 3.794,
      "step": 3977
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.532563891178896e-06,
      "loss": 3.6825,
      "step": 3978
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.526380873866448e-06,
      "loss": 4.1122,
      "step": 3979
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.520197856553998e-06,
      "loss": 3.9581,
      "step": 3980
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.51401483924155e-06,
      "loss": 3.7325,
      "step": 3981
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.507831821929101e-06,
      "loss": 3.6681,
      "step": 3982
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.501648804616654e-06,
      "loss": 3.8794,
      "step": 3983
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.495465787304203e-06,
      "loss": 3.784,
      "step": 3984
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.489282769991756e-06,
      "loss": 4.0183,
      "step": 3985
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.483099752679307e-06,
      "loss": 3.7633,
      "step": 3986
    },
    {
      "epoch": 2.23,
      "learning_rate": 8.47691673536686e-06,
      "loss": 3.8726,
      "step": 3987
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.47073371805441e-06,
      "loss": 3.8154,
      "step": 3988
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.464550700741962e-06,
      "loss": 3.6482,
      "step": 3989
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.458367683429514e-06,
      "loss": 3.8959,
      "step": 3990
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.452184666117065e-06,
      "loss": 3.781,
      "step": 3991
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.446001648804618e-06,
      "loss": 3.9457,
      "step": 3992
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.439818631492167e-06,
      "loss": 3.7086,
      "step": 3993
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.43363561417972e-06,
      "loss": 4.0036,
      "step": 3994
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.427452596867273e-06,
      "loss": 3.9564,
      "step": 3995
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.421269579554824e-06,
      "loss": 3.7668,
      "step": 3996
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.415086562242375e-06,
      "loss": 3.7396,
      "step": 3997
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.408903544929926e-06,
      "loss": 3.9569,
      "step": 3998
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.402720527617478e-06,
      "loss": 3.76,
      "step": 3999
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.39653751030503e-06,
      "loss": 3.5985,
      "step": 4000
    },
    {
      "epoch": 2.24,
      "eval_loss": 4.015486240386963,
      "eval_runtime": 1251.8271,
      "eval_samples_per_second": 2.111,
      "eval_steps_per_second": 0.264,
      "eval_wer": 1.070619282893823,
      "step": 4000
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.39035449299258e-06,
      "loss": 3.6797,
      "step": 4001
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.384171475680131e-06,
      "loss": 3.7895,
      "step": 4002
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.377988458367684e-06,
      "loss": 3.8082,
      "step": 4003
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.371805441055236e-06,
      "loss": 3.6379,
      "step": 4004
    },
    {
      "epoch": 2.24,
      "learning_rate": 8.365622423742786e-06,
      "loss": 3.9338,
      "step": 4005
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.359439406430338e-06,
      "loss": 3.956,
      "step": 4006
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.35325638911789e-06,
      "loss": 3.7616,
      "step": 4007
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.347073371805442e-06,
      "loss": 3.9542,
      "step": 4008
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.340890354492991e-06,
      "loss": 3.5742,
      "step": 4009
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.334707337180544e-06,
      "loss": 3.4679,
      "step": 4010
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.328524319868097e-06,
      "loss": 3.7448,
      "step": 4011
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.322341302555648e-06,
      "loss": 3.5487,
      "step": 4012
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.316158285243199e-06,
      "loss": 3.3689,
      "step": 4013
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.30997526793075e-06,
      "loss": 3.3957,
      "step": 4014
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.303792250618302e-06,
      "loss": 2.7912,
      "step": 4015
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.297609233305853e-06,
      "loss": 3.5633,
      "step": 4016
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.291426215993404e-06,
      "loss": 3.2241,
      "step": 4017
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.285243198680955e-06,
      "loss": 2.7998,
      "step": 4018
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.279060181368508e-06,
      "loss": 4.1556,
      "step": 4019
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.27287716405606e-06,
      "loss": 4.0021,
      "step": 4020
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.266694146743612e-06,
      "loss": 3.9343,
      "step": 4021
    },
    {
      "epoch": 2.25,
      "learning_rate": 8.260511129431163e-06,
      "loss": 4.0133,
      "step": 4022
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.254328112118714e-06,
      "loss": 3.7197,
      "step": 4023
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.248145094806266e-06,
      "loss": 4.0178,
      "step": 4024
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.241962077493817e-06,
      "loss": 3.9014,
      "step": 4025
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.235779060181368e-06,
      "loss": 4.0398,
      "step": 4026
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.229596042868921e-06,
      "loss": 3.8217,
      "step": 4027
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.223413025556472e-06,
      "loss": 3.8941,
      "step": 4028
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.217230008244025e-06,
      "loss": 3.7564,
      "step": 4029
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.211046990931574e-06,
      "loss": 3.9138,
      "step": 4030
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.204863973619127e-06,
      "loss": 3.6726,
      "step": 4031
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.198680956306678e-06,
      "loss": 3.9051,
      "step": 4032
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.19249793899423e-06,
      "loss": 4.0433,
      "step": 4033
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.18631492168178e-06,
      "loss": 3.7144,
      "step": 4034
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.180131904369332e-06,
      "loss": 3.7819,
      "step": 4035
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.173948887056885e-06,
      "loss": 3.6046,
      "step": 4036
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.167765869744436e-06,
      "loss": 3.7815,
      "step": 4037
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.161582852431987e-06,
      "loss": 3.7504,
      "step": 4038
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.155399835119538e-06,
      "loss": 3.5128,
      "step": 4039
    },
    {
      "epoch": 2.26,
      "learning_rate": 8.14921681780709e-06,
      "loss": 3.9042,
      "step": 4040
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.143033800494641e-06,
      "loss": 3.7137,
      "step": 4041
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.136850783182192e-06,
      "loss": 3.6638,
      "step": 4042
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.130667765869745e-06,
      "loss": 4.0046,
      "step": 4043
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.124484748557296e-06,
      "loss": 3.8986,
      "step": 4044
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.118301731244849e-06,
      "loss": 3.7461,
      "step": 4045
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.112118713932398e-06,
      "loss": 3.8482,
      "step": 4046
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.10593569661995e-06,
      "loss": 3.5988,
      "step": 4047
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.099752679307502e-06,
      "loss": 3.8775,
      "step": 4048
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.093569661995054e-06,
      "loss": 3.6488,
      "step": 4049
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.087386644682605e-06,
      "loss": 3.9292,
      "step": 4050
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.081203627370156e-06,
      "loss": 3.704,
      "step": 4051
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.075020610057709e-06,
      "loss": 3.78,
      "step": 4052
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.06883759274526e-06,
      "loss": 3.9508,
      "step": 4053
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.062654575432813e-06,
      "loss": 3.6939,
      "step": 4054
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.056471558120362e-06,
      "loss": 3.6998,
      "step": 4055
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.050288540807915e-06,
      "loss": 3.6484,
      "step": 4056
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.044105523495466e-06,
      "loss": 3.9611,
      "step": 4057
    },
    {
      "epoch": 2.27,
      "learning_rate": 8.037922506183018e-06,
      "loss": 3.7143,
      "step": 4058
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.03173948887057e-06,
      "loss": 3.6634,
      "step": 4059
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.02555647155812e-06,
      "loss": 3.7262,
      "step": 4060
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.019373454245673e-06,
      "loss": 3.5457,
      "step": 4061
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.013190436933224e-06,
      "loss": 3.4027,
      "step": 4062
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.007007419620775e-06,
      "loss": 3.7541,
      "step": 4063
    },
    {
      "epoch": 2.28,
      "learning_rate": 8.000824402308326e-06,
      "loss": 3.8226,
      "step": 4064
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.994641384995879e-06,
      "loss": 3.1952,
      "step": 4065
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.988458367683431e-06,
      "loss": 2.9179,
      "step": 4066
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.98227535037098e-06,
      "loss": 2.9077,
      "step": 4067
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.976092333058533e-06,
      "loss": 3.2426,
      "step": 4068
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.969909315746084e-06,
      "loss": 4.074,
      "step": 4069
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.963726298433637e-06,
      "loss": 3.8695,
      "step": 4070
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.957543281121186e-06,
      "loss": 4.0667,
      "step": 4071
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.951360263808739e-06,
      "loss": 3.9425,
      "step": 4072
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.94517724649629e-06,
      "loss": 3.878,
      "step": 4073
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.938994229183842e-06,
      "loss": 4.0065,
      "step": 4074
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.932811211871393e-06,
      "loss": 3.9741,
      "step": 4075
    },
    {
      "epoch": 2.28,
      "learning_rate": 7.926628194558944e-06,
      "loss": 3.923,
      "step": 4076
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.920445177246497e-06,
      "loss": 3.9588,
      "step": 4077
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.914262159934048e-06,
      "loss": 3.7827,
      "step": 4078
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.908079142621599e-06,
      "loss": 3.9621,
      "step": 4079
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.90189612530915e-06,
      "loss": 3.7037,
      "step": 4080
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.895713107996703e-06,
      "loss": 3.8386,
      "step": 4081
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.889530090684255e-06,
      "loss": 3.6617,
      "step": 4082
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.883347073371806e-06,
      "loss": 3.9265,
      "step": 4083
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.877164056059357e-06,
      "loss": 3.8397,
      "step": 4084
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.870981038746908e-06,
      "loss": 3.8139,
      "step": 4085
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.864798021434461e-06,
      "loss": 3.9795,
      "step": 4086
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.858615004122012e-06,
      "loss": 3.8615,
      "step": 4087
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.852431986809563e-06,
      "loss": 3.86,
      "step": 4088
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.846248969497114e-06,
      "loss": 3.9004,
      "step": 4089
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.840065952184667e-06,
      "loss": 3.966,
      "step": 4090
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.83388293487222e-06,
      "loss": 3.816,
      "step": 4091
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.827699917559769e-06,
      "loss": 3.8335,
      "step": 4092
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.821516900247321e-06,
      "loss": 3.8133,
      "step": 4093
    },
    {
      "epoch": 2.29,
      "learning_rate": 7.815333882934872e-06,
      "loss": 3.9499,
      "step": 4094
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.809150865622425e-06,
      "loss": 3.9559,
      "step": 4095
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.802967848309974e-06,
      "loss": 3.8766,
      "step": 4096
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.796784830997527e-06,
      "loss": 3.9777,
      "step": 4097
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.79060181368508e-06,
      "loss": 4.0286,
      "step": 4098
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.78441879637263e-06,
      "loss": 3.7974,
      "step": 4099
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.778235779060181e-06,
      "loss": 3.9527,
      "step": 4100
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.772052761747732e-06,
      "loss": 3.7006,
      "step": 4101
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.765869744435285e-06,
      "loss": 3.6025,
      "step": 4102
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.759686727122836e-06,
      "loss": 3.4427,
      "step": 4103
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.753503709810387e-06,
      "loss": 3.7526,
      "step": 4104
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.747320692497938e-06,
      "loss": 3.5305,
      "step": 4105
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.74113767518549e-06,
      "loss": 3.5947,
      "step": 4106
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.734954657873043e-06,
      "loss": 3.923,
      "step": 4107
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.728771640560593e-06,
      "loss": 3.6112,
      "step": 4108
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.722588623248145e-06,
      "loss": 3.687,
      "step": 4109
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.716405605935696e-06,
      "loss": 3.44,
      "step": 4110
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.710222588623249e-06,
      "loss": 3.7162,
      "step": 4111
    },
    {
      "epoch": 2.3,
      "learning_rate": 7.704039571310798e-06,
      "loss": 3.4742,
      "step": 4112
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.697856553998351e-06,
      "loss": 3.1097,
      "step": 4113
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.691673536685904e-06,
      "loss": 3.7911,
      "step": 4114
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.685490519373455e-06,
      "loss": 3.0687,
      "step": 4115
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.679307502061007e-06,
      "loss": 3.1107,
      "step": 4116
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.673124484748557e-06,
      "loss": 2.7268,
      "step": 4117
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.66694146743611e-06,
      "loss": 2.7203,
      "step": 4118
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.66075845012366e-06,
      "loss": 4.0294,
      "step": 4119
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.654575432811213e-06,
      "loss": 4.0003,
      "step": 4120
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.648392415498762e-06,
      "loss": 4.0576,
      "step": 4121
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.642209398186315e-06,
      "loss": 4.0077,
      "step": 4122
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.636026380873868e-06,
      "loss": 3.8282,
      "step": 4123
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.629843363561419e-06,
      "loss": 3.9944,
      "step": 4124
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.62366034624897e-06,
      "loss": 3.6983,
      "step": 4125
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.6174773289365205e-06,
      "loss": 3.8632,
      "step": 4126
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.611294311624073e-06,
      "loss": 3.8772,
      "step": 4127
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.605111294311623e-06,
      "loss": 4.0513,
      "step": 4128
    },
    {
      "epoch": 2.31,
      "learning_rate": 7.598928276999176e-06,
      "loss": 3.9332,
      "step": 4129
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.592745259686728e-06,
      "loss": 3.7789,
      "step": 4130
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.586562242374279e-06,
      "loss": 4.3456,
      "step": 4131
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.580379225061831e-06,
      "loss": 3.8331,
      "step": 4132
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.574196207749382e-06,
      "loss": 4.223,
      "step": 4133
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.5680131904369335e-06,
      "loss": 3.8331,
      "step": 4134
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.5618301731244845e-06,
      "loss": 3.7805,
      "step": 4135
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.555647155812036e-06,
      "loss": 3.8473,
      "step": 4136
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.549464138499587e-06,
      "loss": 3.8085,
      "step": 4137
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.543281121187139e-06,
      "loss": 3.6917,
      "step": 4138
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.537098103874692e-06,
      "loss": 3.7223,
      "step": 4139
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.530915086562242e-06,
      "loss": 3.8766,
      "step": 4140
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.5247320692497946e-06,
      "loss": 3.5986,
      "step": 4141
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.518549051937345e-06,
      "loss": 3.8226,
      "step": 4142
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.512366034624897e-06,
      "loss": 3.5932,
      "step": 4143
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.5061830173124475e-06,
      "loss": 4.0789,
      "step": 4144
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.5e-06,
      "loss": 3.8743,
      "step": 4145
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.493816982687552e-06,
      "loss": 3.9016,
      "step": 4146
    },
    {
      "epoch": 2.32,
      "learning_rate": 7.487633965375103e-06,
      "loss": 3.8558,
      "step": 4147
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.481450948062655e-06,
      "loss": 4.0076,
      "step": 4148
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.475267930750207e-06,
      "loss": 3.5664,
      "step": 4149
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.4690849134377585e-06,
      "loss": 3.5708,
      "step": 4150
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.4629018961253095e-06,
      "loss": 3.5883,
      "step": 4151
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.456718878812861e-06,
      "loss": 3.6421,
      "step": 4152
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.450535861500412e-06,
      "loss": 3.8131,
      "step": 4153
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.444352844187964e-06,
      "loss": 3.736,
      "step": 4154
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.438169826875515e-06,
      "loss": 3.6011,
      "step": 4155
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.431986809563067e-06,
      "loss": 4.0112,
      "step": 4156
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.425803792250619e-06,
      "loss": 3.5546,
      "step": 4157
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.4196207749381706e-06,
      "loss": 3.7437,
      "step": 4158
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.4134377576257215e-06,
      "loss": 3.6105,
      "step": 4159
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.407254740313273e-06,
      "loss": 3.7958,
      "step": 4160
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.401071723000824e-06,
      "loss": 3.7165,
      "step": 4161
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.394888705688376e-06,
      "loss": 3.4655,
      "step": 4162
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.388705688375927e-06,
      "loss": 3.5202,
      "step": 4163
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.382522671063479e-06,
      "loss": 3.4311,
      "step": 4164
    },
    {
      "epoch": 2.33,
      "learning_rate": 7.376339653751031e-06,
      "loss": 3.2112,
      "step": 4165
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.370156636438583e-06,
      "loss": 3.1829,
      "step": 4166
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.363973619126134e-06,
      "loss": 2.823,
      "step": 4167
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.3577906018136855e-06,
      "loss": 2.8903,
      "step": 4168
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.3516075845012364e-06,
      "loss": 3.9683,
      "step": 4169
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.345424567188788e-06,
      "loss": 4.015,
      "step": 4170
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.339241549876339e-06,
      "loss": 4.0403,
      "step": 4171
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.333058532563891e-06,
      "loss": 3.7892,
      "step": 4172
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.326875515251443e-06,
      "loss": 3.927,
      "step": 4173
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.320692497938995e-06,
      "loss": 3.9556,
      "step": 4174
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.314509480626546e-06,
      "loss": 3.9537,
      "step": 4175
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.3083264633140975e-06,
      "loss": 3.9244,
      "step": 4176
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.302143446001649e-06,
      "loss": 4.0058,
      "step": 4177
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.2959604286892e-06,
      "loss": 4.198,
      "step": 4178
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.289777411376752e-06,
      "loss": 3.6892,
      "step": 4179
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.283594394064303e-06,
      "loss": 4.22,
      "step": 4180
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.277411376751856e-06,
      "loss": 4.2035,
      "step": 4181
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.271228359439407e-06,
      "loss": 4.0011,
      "step": 4182
    },
    {
      "epoch": 2.34,
      "learning_rate": 7.265045342126959e-06,
      "loss": 3.9702,
      "step": 4183
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.25886232481451e-06,
      "loss": 3.8358,
      "step": 4184
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.2526793075020615e-06,
      "loss": 3.6884,
      "step": 4185
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.2464962901896124e-06,
      "loss": 3.9895,
      "step": 4186
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.240313272877164e-06,
      "loss": 3.8284,
      "step": 4187
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.234130255564715e-06,
      "loss": 3.6764,
      "step": 4188
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.227947238252268e-06,
      "loss": 3.9165,
      "step": 4189
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.221764220939819e-06,
      "loss": 3.8903,
      "step": 4190
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.215581203627371e-06,
      "loss": 3.6454,
      "step": 4191
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.209398186314922e-06,
      "loss": 3.5804,
      "step": 4192
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.2032151690024735e-06,
      "loss": 3.788,
      "step": 4193
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.1970321516900245e-06,
      "loss": 3.9302,
      "step": 4194
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.190849134377576e-06,
      "loss": 3.9037,
      "step": 4195
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.184666117065127e-06,
      "loss": 3.7013,
      "step": 4196
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.17848309975268e-06,
      "loss": 3.7913,
      "step": 4197
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.172300082440231e-06,
      "loss": 3.8551,
      "step": 4198
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.166117065127783e-06,
      "loss": 3.7643,
      "step": 4199
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.159934047815334e-06,
      "loss": 3.6698,
      "step": 4200
    },
    {
      "epoch": 2.35,
      "learning_rate": 7.153751030502886e-06,
      "loss": 3.6774,
      "step": 4201
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.147568013190437e-06,
      "loss": 3.8607,
      "step": 4202
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.1413849958779884e-06,
      "loss": 3.8605,
      "step": 4203
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.135201978565539e-06,
      "loss": 3.7929,
      "step": 4204
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.129018961253092e-06,
      "loss": 3.868,
      "step": 4205
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.122835943940643e-06,
      "loss": 3.6749,
      "step": 4206
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.116652926628195e-06,
      "loss": 3.7128,
      "step": 4207
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.110469909315747e-06,
      "loss": 3.5396,
      "step": 4208
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.104286892003298e-06,
      "loss": 3.7542,
      "step": 4209
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.0981038746908495e-06,
      "loss": 3.5861,
      "step": 4210
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.0919208573784005e-06,
      "loss": 3.7208,
      "step": 4211
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.085737840065952e-06,
      "loss": 3.3329,
      "step": 4212
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.079554822753504e-06,
      "loss": 3.5346,
      "step": 4213
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.073371805441056e-06,
      "loss": 2.9062,
      "step": 4214
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.067188788128607e-06,
      "loss": 3.4031,
      "step": 4215
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.061005770816159e-06,
      "loss": 3.2823,
      "step": 4216
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.05482275350371e-06,
      "loss": 3.2881,
      "step": 4217
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.048639736191262e-06,
      "loss": 3.0743,
      "step": 4218
    },
    {
      "epoch": 2.36,
      "learning_rate": 7.042456718878813e-06,
      "loss": 3.9466,
      "step": 4219
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.0362737015663644e-06,
      "loss": 3.9496,
      "step": 4220
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.030090684253916e-06,
      "loss": 4.0599,
      "step": 4221
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.023907666941468e-06,
      "loss": 4.1197,
      "step": 4222
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.017724649629019e-06,
      "loss": 3.8464,
      "step": 4223
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.011541632316571e-06,
      "loss": 3.8473,
      "step": 4224
    },
    {
      "epoch": 2.37,
      "learning_rate": 7.005358615004122e-06,
      "loss": 3.7828,
      "step": 4225
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.999175597691674e-06,
      "loss": 3.8204,
      "step": 4226
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.992992580379225e-06,
      "loss": 3.8284,
      "step": 4227
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.9868095630667765e-06,
      "loss": 4.2685,
      "step": 4228
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.980626545754328e-06,
      "loss": 3.8651,
      "step": 4229
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.97444352844188e-06,
      "loss": 3.7613,
      "step": 4230
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.968260511129431e-06,
      "loss": 3.9223,
      "step": 4231
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.962077493816983e-06,
      "loss": 3.8023,
      "step": 4232
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.955894476504534e-06,
      "loss": 3.7112,
      "step": 4233
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.949711459192086e-06,
      "loss": 3.8675,
      "step": 4234
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.943528441879637e-06,
      "loss": 3.7733,
      "step": 4235
    },
    {
      "epoch": 2.37,
      "learning_rate": 6.937345424567189e-06,
      "loss": 3.993,
      "step": 4236
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.93116240725474e-06,
      "loss": 3.7979,
      "step": 4237
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.924979389942292e-06,
      "loss": 3.6774,
      "step": 4238
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.918796372629844e-06,
      "loss": 3.7935,
      "step": 4239
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.912613355317395e-06,
      "loss": 3.7438,
      "step": 4240
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.906430338004947e-06,
      "loss": 3.7579,
      "step": 4241
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.900247320692498e-06,
      "loss": 4.0738,
      "step": 4242
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.89406430338005e-06,
      "loss": 3.729,
      "step": 4243
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.887881286067601e-06,
      "loss": 3.8812,
      "step": 4244
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.881698268755153e-06,
      "loss": 3.8116,
      "step": 4245
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.875515251442704e-06,
      "loss": 3.7733,
      "step": 4246
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.869332234130256e-06,
      "loss": 3.7585,
      "step": 4247
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.863149216817807e-06,
      "loss": 3.7072,
      "step": 4248
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.856966199505359e-06,
      "loss": 3.8151,
      "step": 4249
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.85078318219291e-06,
      "loss": 3.5626,
      "step": 4250
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.844600164880462e-06,
      "loss": 3.8276,
      "step": 4251
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.838417147568013e-06,
      "loss": 3.5494,
      "step": 4252
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.8322341302555654e-06,
      "loss": 3.6756,
      "step": 4253
    },
    {
      "epoch": 2.38,
      "learning_rate": 6.826051112943116e-06,
      "loss": 3.6994,
      "step": 4254
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.819868095630668e-06,
      "loss": 3.7247,
      "step": 4255
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.813685078318219e-06,
      "loss": 3.5821,
      "step": 4256
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.807502061005771e-06,
      "loss": 3.4826,
      "step": 4257
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.801319043693322e-06,
      "loss": 3.7665,
      "step": 4258
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.795136026380874e-06,
      "loss": 3.5718,
      "step": 4259
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.788953009068426e-06,
      "loss": 3.3629,
      "step": 4260
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.7827699917559775e-06,
      "loss": 3.3324,
      "step": 4261
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.7765869744435285e-06,
      "loss": 3.7438,
      "step": 4262
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.77040395713108e-06,
      "loss": 3.5707,
      "step": 4263
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.764220939818631e-06,
      "loss": 3.4394,
      "step": 4264
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.758037922506183e-06,
      "loss": 3.0667,
      "step": 4265
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.751854905193734e-06,
      "loss": 2.8844,
      "step": 4266
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.745671887881286e-06,
      "loss": 2.9936,
      "step": 4267
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.739488870568838e-06,
      "loss": 2.9085,
      "step": 4268
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.73330585325639e-06,
      "loss": 3.9634,
      "step": 4269
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.7271228359439414e-06,
      "loss": 4.1027,
      "step": 4270
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.720939818631492e-06,
      "loss": 3.9689,
      "step": 4271
    },
    {
      "epoch": 2.39,
      "learning_rate": 6.714756801319044e-06,
      "loss": 3.6509,
      "step": 4272
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.708573784006595e-06,
      "loss": 4.0709,
      "step": 4273
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.702390766694147e-06,
      "loss": 3.798,
      "step": 4274
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.696207749381698e-06,
      "loss": 3.7985,
      "step": 4275
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.690024732069251e-06,
      "loss": 4.0065,
      "step": 4276
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.683841714756802e-06,
      "loss": 3.7958,
      "step": 4277
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.6776586974443535e-06,
      "loss": 3.7822,
      "step": 4278
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.6714756801319045e-06,
      "loss": 3.7904,
      "step": 4279
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.665292662819456e-06,
      "loss": 3.8489,
      "step": 4280
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.659109645507007e-06,
      "loss": 3.8359,
      "step": 4281
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.652926628194559e-06,
      "loss": 3.7588,
      "step": 4282
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.64674361088211e-06,
      "loss": 4.0774,
      "step": 4283
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.640560593569663e-06,
      "loss": 3.8772,
      "step": 4284
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.634377576257214e-06,
      "loss": 4.0723,
      "step": 4285
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.628194558944766e-06,
      "loss": 3.7728,
      "step": 4286
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.6220115416323166e-06,
      "loss": 3.7584,
      "step": 4287
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.615828524319868e-06,
      "loss": 3.6662,
      "step": 4288
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.609645507007419e-06,
      "loss": 3.995,
      "step": 4289
    },
    {
      "epoch": 2.4,
      "learning_rate": 6.603462489694971e-06,
      "loss": 3.6765,
      "step": 4290
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.597279472382522e-06,
      "loss": 3.7941,
      "step": 4291
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.591096455070075e-06,
      "loss": 3.7338,
      "step": 4292
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.584913437757626e-06,
      "loss": 3.8527,
      "step": 4293
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.578730420445178e-06,
      "loss": 3.8893,
      "step": 4294
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.572547403132729e-06,
      "loss": 3.8213,
      "step": 4295
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.5663643858202805e-06,
      "loss": 3.717,
      "step": 4296
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.5601813685078315e-06,
      "loss": 3.6198,
      "step": 4297
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.553998351195383e-06,
      "loss": 3.8413,
      "step": 4298
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.547815333882934e-06,
      "loss": 3.6843,
      "step": 4299
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.541632316570487e-06,
      "loss": 3.6567,
      "step": 4300
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.535449299258039e-06,
      "loss": 3.8148,
      "step": 4301
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.52926628194559e-06,
      "loss": 3.672,
      "step": 4302
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.523083264633142e-06,
      "loss": 3.9674,
      "step": 4303
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.5169002473206926e-06,
      "loss": 3.7529,
      "step": 4304
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.510717230008244e-06,
      "loss": 3.6821,
      "step": 4305
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.504534212695795e-06,
      "loss": 3.7039,
      "step": 4306
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.498351195383347e-06,
      "loss": 3.6725,
      "step": 4307
    },
    {
      "epoch": 2.41,
      "learning_rate": 6.492168178070899e-06,
      "loss": 3.6109,
      "step": 4308
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.485985160758451e-06,
      "loss": 3.7409,
      "step": 4309
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.479802143446002e-06,
      "loss": 3.7397,
      "step": 4310
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.473619126133554e-06,
      "loss": 3.6118,
      "step": 4311
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.467436108821105e-06,
      "loss": 3.605,
      "step": 4312
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.4612530915086565e-06,
      "loss": 3.4621,
      "step": 4313
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.4550700741962075e-06,
      "loss": 3.3176,
      "step": 4314
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.448887056883759e-06,
      "loss": 3.446,
      "step": 4315
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.442704039571311e-06,
      "loss": 3.0602,
      "step": 4316
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.436521022258863e-06,
      "loss": 2.6684,
      "step": 4317
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.430338004946414e-06,
      "loss": 2.6803,
      "step": 4318
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.424154987633966e-06,
      "loss": 3.9052,
      "step": 4319
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.417971970321517e-06,
      "loss": 4.306,
      "step": 4320
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.4117889530090686e-06,
      "loss": 3.7972,
      "step": 4321
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.4056059356966195e-06,
      "loss": 3.9216,
      "step": 4322
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.399422918384171e-06,
      "loss": 3.9421,
      "step": 4323
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.393239901071723e-06,
      "loss": 4.0259,
      "step": 4324
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.387056883759275e-06,
      "loss": 3.7384,
      "step": 4325
    },
    {
      "epoch": 2.42,
      "learning_rate": 6.380873866446826e-06,
      "loss": 3.9943,
      "step": 4326
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.374690849134378e-06,
      "loss": 4.0813,
      "step": 4327
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.368507831821929e-06,
      "loss": 3.6718,
      "step": 4328
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.362324814509481e-06,
      "loss": 3.7471,
      "step": 4329
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.356141797197032e-06,
      "loss": 3.8185,
      "step": 4330
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.3499587798845835e-06,
      "loss": 3.993,
      "step": 4331
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.343775762572135e-06,
      "loss": 4.0046,
      "step": 4332
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.337592745259687e-06,
      "loss": 3.8027,
      "step": 4333
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.331409727947239e-06,
      "loss": 3.9202,
      "step": 4334
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.32522671063479e-06,
      "loss": 3.8384,
      "step": 4335
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.319043693322342e-06,
      "loss": 4.1991,
      "step": 4336
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.312860676009893e-06,
      "loss": 3.7465,
      "step": 4337
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.3066776586974446e-06,
      "loss": 4.0554,
      "step": 4338
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.3004946413849955e-06,
      "loss": 3.9169,
      "step": 4339
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.294311624072548e-06,
      "loss": 3.8176,
      "step": 4340
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.288128606760099e-06,
      "loss": 3.9283,
      "step": 4341
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.281945589447651e-06,
      "loss": 3.6129,
      "step": 4342
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.275762572135202e-06,
      "loss": 3.7223,
      "step": 4343
    },
    {
      "epoch": 2.43,
      "learning_rate": 6.269579554822754e-06,
      "loss": 3.9977,
      "step": 4344
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.263396537510305e-06,
      "loss": 3.9695,
      "step": 4345
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.257213520197857e-06,
      "loss": 3.5599,
      "step": 4346
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.251030502885408e-06,
      "loss": 3.7765,
      "step": 4347
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.24484748557296e-06,
      "loss": 3.5848,
      "step": 4348
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.238664468260511e-06,
      "loss": 3.9012,
      "step": 4349
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.232481450948063e-06,
      "loss": 3.8178,
      "step": 4350
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.226298433635614e-06,
      "loss": 3.4966,
      "step": 4351
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.220115416323166e-06,
      "loss": 3.466,
      "step": 4352
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.213932399010717e-06,
      "loss": 3.768,
      "step": 4353
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.207749381698269e-06,
      "loss": 3.6794,
      "step": 4354
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.20156636438582e-06,
      "loss": 3.7562,
      "step": 4355
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.195383347073372e-06,
      "loss": 3.9473,
      "step": 4356
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.189200329760923e-06,
      "loss": 3.7983,
      "step": 4357
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.183017312448475e-06,
      "loss": 3.5557,
      "step": 4358
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.176834295136026e-06,
      "loss": 3.6733,
      "step": 4359
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.170651277823578e-06,
      "loss": 3.7965,
      "step": 4360
    },
    {
      "epoch": 2.44,
      "learning_rate": 6.164468260511129e-06,
      "loss": 3.7083,
      "step": 4361
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.158285243198681e-06,
      "loss": 3.7485,
      "step": 4362
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.152102225886233e-06,
      "loss": 3.7135,
      "step": 4363
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.1459192085737845e-06,
      "loss": 3.5547,
      "step": 4364
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.139736191261336e-06,
      "loss": 3.3604,
      "step": 4365
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.133553173948887e-06,
      "loss": 3.1014,
      "step": 4366
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.127370156636439e-06,
      "loss": 3.1218,
      "step": 4367
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.12118713932399e-06,
      "loss": 2.707,
      "step": 4368
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.115004122011542e-06,
      "loss": 4.0284,
      "step": 4369
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.108821104699093e-06,
      "loss": 4.0411,
      "step": 4370
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.1026380873866456e-06,
      "loss": 3.8214,
      "step": 4371
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.0964550700741965e-06,
      "loss": 4.0168,
      "step": 4372
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.090272052761748e-06,
      "loss": 3.789,
      "step": 4373
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.084089035449299e-06,
      "loss": 3.8266,
      "step": 4374
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.077906018136851e-06,
      "loss": 3.9624,
      "step": 4375
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.071723000824402e-06,
      "loss": 3.9463,
      "step": 4376
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.065539983511954e-06,
      "loss": 3.9982,
      "step": 4377
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.059356966199505e-06,
      "loss": 4.1394,
      "step": 4378
    },
    {
      "epoch": 2.45,
      "learning_rate": 6.053173948887058e-06,
      "loss": 3.9229,
      "step": 4379
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.046990931574609e-06,
      "loss": 3.8871,
      "step": 4380
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.0408079142621605e-06,
      "loss": 3.6756,
      "step": 4381
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.0346248969497114e-06,
      "loss": 3.7937,
      "step": 4382
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.028441879637263e-06,
      "loss": 3.8905,
      "step": 4383
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.022258862324814e-06,
      "loss": 3.9301,
      "step": 4384
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.016075845012366e-06,
      "loss": 4.0017,
      "step": 4385
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.009892827699917e-06,
      "loss": 3.5704,
      "step": 4386
    },
    {
      "epoch": 2.46,
      "learning_rate": 6.00370981038747e-06,
      "loss": 3.7911,
      "step": 4387
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.997526793075021e-06,
      "loss": 3.7345,
      "step": 4388
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.9913437757625725e-06,
      "loss": 3.9058,
      "step": 4389
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.9851607584501235e-06,
      "loss": 3.8101,
      "step": 4390
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.978977741137675e-06,
      "loss": 3.7099,
      "step": 4391
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.972794723825226e-06,
      "loss": 3.7816,
      "step": 4392
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.966611706512778e-06,
      "loss": 3.7955,
      "step": 4393
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.96042868920033e-06,
      "loss": 3.5463,
      "step": 4394
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.954245671887882e-06,
      "loss": 3.6459,
      "step": 4395
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.948062654575434e-06,
      "loss": 3.6754,
      "step": 4396
    },
    {
      "epoch": 2.46,
      "learning_rate": 5.941879637262985e-06,
      "loss": 4.0264,
      "step": 4397
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.9356966199505365e-06,
      "loss": 3.8745,
      "step": 4398
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.9295136026380874e-06,
      "loss": 3.8639,
      "step": 4399
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.923330585325639e-06,
      "loss": 3.8053,
      "step": 4400
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.91714756801319e-06,
      "loss": 3.5358,
      "step": 4401
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.910964550700742e-06,
      "loss": 3.7287,
      "step": 4402
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.904781533388294e-06,
      "loss": 3.6887,
      "step": 4403
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.898598516075846e-06,
      "loss": 4.0044,
      "step": 4404
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.892415498763397e-06,
      "loss": 3.6777,
      "step": 4405
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.8862324814509485e-06,
      "loss": 3.5982,
      "step": 4406
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.8800494641384995e-06,
      "loss": 3.8517,
      "step": 4407
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.873866446826051e-06,
      "loss": 3.5806,
      "step": 4408
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.867683429513602e-06,
      "loss": 3.5763,
      "step": 4409
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.861500412201154e-06,
      "loss": 3.6678,
      "step": 4410
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.855317394888706e-06,
      "loss": 3.6259,
      "step": 4411
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.849134377576258e-06,
      "loss": 3.464,
      "step": 4412
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.842951360263809e-06,
      "loss": 3.551,
      "step": 4413
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.836768342951361e-06,
      "loss": 3.3858,
      "step": 4414
    },
    {
      "epoch": 2.47,
      "learning_rate": 5.830585325638912e-06,
      "loss": 3.1389,
      "step": 4415
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.8244023083264634e-06,
      "loss": 3.0361,
      "step": 4416
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.818219291014014e-06,
      "loss": 2.9703,
      "step": 4417
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.812036273701566e-06,
      "loss": 3.0996,
      "step": 4418
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.805853256389118e-06,
      "loss": 3.9136,
      "step": 4419
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.79967023907667e-06,
      "loss": 3.8505,
      "step": 4420
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.793487221764221e-06,
      "loss": 4.0267,
      "step": 4421
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.787304204451773e-06,
      "loss": 3.8767,
      "step": 4422
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.781121187139324e-06,
      "loss": 3.9693,
      "step": 4423
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.7749381698268755e-06,
      "loss": 3.6839,
      "step": 4424
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.768755152514427e-06,
      "loss": 3.8857,
      "step": 4425
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.762572135201978e-06,
      "loss": 3.892,
      "step": 4426
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.756389117889531e-06,
      "loss": 3.9997,
      "step": 4427
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.750206100577082e-06,
      "loss": 3.9567,
      "step": 4428
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.744023083264634e-06,
      "loss": 3.9996,
      "step": 4429
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.737840065952185e-06,
      "loss": 3.7286,
      "step": 4430
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.731657048639737e-06,
      "loss": 3.8224,
      "step": 4431
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.725474031327288e-06,
      "loss": 3.7028,
      "step": 4432
    },
    {
      "epoch": 2.48,
      "learning_rate": 5.719291014014839e-06,
      "loss": 3.8966,
      "step": 4433
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.71310799670239e-06,
      "loss": 3.6805,
      "step": 4434
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.706924979389943e-06,
      "loss": 3.859,
      "step": 4435
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.700741962077494e-06,
      "loss": 3.8103,
      "step": 4436
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.694558944765046e-06,
      "loss": 3.6384,
      "step": 4437
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.688375927452597e-06,
      "loss": 3.7625,
      "step": 4438
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.682192910140149e-06,
      "loss": 3.7991,
      "step": 4439
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.6760098928277e-06,
      "loss": 3.9519,
      "step": 4440
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.6698268755152515e-06,
      "loss": 3.7827,
      "step": 4441
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.6636438582028025e-06,
      "loss": 3.7575,
      "step": 4442
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.657460840890355e-06,
      "loss": 4.0248,
      "step": 4443
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.651277823577906e-06,
      "loss": 3.7011,
      "step": 4444
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.645094806265458e-06,
      "loss": 4.1126,
      "step": 4445
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.638911788953009e-06,
      "loss": 3.8996,
      "step": 4446
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.632728771640561e-06,
      "loss": 3.7837,
      "step": 4447
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.626545754328112e-06,
      "loss": 3.9642,
      "step": 4448
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.620362737015664e-06,
      "loss": 3.9439,
      "step": 4449
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.6141797197032146e-06,
      "loss": 3.6933,
      "step": 4450
    },
    {
      "epoch": 2.49,
      "learning_rate": 5.607996702390767e-06,
      "loss": 3.7545,
      "step": 4451
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.601813685078318e-06,
      "loss": 3.478,
      "step": 4452
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.59563066776587e-06,
      "loss": 3.6474,
      "step": 4453
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.589447650453421e-06,
      "loss": 3.7669,
      "step": 4454
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.583264633140973e-06,
      "loss": 3.8493,
      "step": 4455
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.577081615828525e-06,
      "loss": 3.6259,
      "step": 4456
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.570898598516076e-06,
      "loss": 3.9199,
      "step": 4457
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.5647155812036275e-06,
      "loss": 3.5779,
      "step": 4458
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.558532563891179e-06,
      "loss": 3.7107,
      "step": 4459
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.552349546578731e-06,
      "loss": 3.838,
      "step": 4460
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.546166529266282e-06,
      "loss": 3.5209,
      "step": 4461
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.539983511953834e-06,
      "loss": 3.3353,
      "step": 4462
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.533800494641385e-06,
      "loss": 3.4626,
      "step": 4463
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.527617477328937e-06,
      "loss": 3.4304,
      "step": 4464
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.521434460016488e-06,
      "loss": 3.3765,
      "step": 4465
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.51525144270404e-06,
      "loss": 3.4009,
      "step": 4466
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.509068425391591e-06,
      "loss": 3.2513,
      "step": 4467
    },
    {
      "epoch": 2.5,
      "learning_rate": 5.502885408079143e-06,
      "loss": 2.7912,
      "step": 4468
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.496702390766694e-06,
      "loss": 3.9577,
      "step": 4469
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.490519373454246e-06,
      "loss": 3.9522,
      "step": 4470
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.484336356141797e-06,
      "loss": 3.9163,
      "step": 4471
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.478153338829349e-06,
      "loss": 4.2045,
      "step": 4472
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.4719703215169e-06,
      "loss": 3.8539,
      "step": 4473
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.465787304204452e-06,
      "loss": 3.9052,
      "step": 4474
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.4596042868920035e-06,
      "loss": 3.6529,
      "step": 4475
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.453421269579555e-06,
      "loss": 3.7271,
      "step": 4476
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.447238252267106e-06,
      "loss": 3.8434,
      "step": 4477
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.441055234954658e-06,
      "loss": 3.9788,
      "step": 4478
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.434872217642209e-06,
      "loss": 3.736,
      "step": 4479
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.428689200329761e-06,
      "loss": 3.8091,
      "step": 4480
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.422506183017312e-06,
      "loss": 3.8505,
      "step": 4481
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.416323165704865e-06,
      "loss": 3.8649,
      "step": 4482
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.410140148392416e-06,
      "loss": 3.5908,
      "step": 4483
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.403957131079967e-06,
      "loss": 3.7392,
      "step": 4484
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.397774113767518e-06,
      "loss": 3.8259,
      "step": 4485
    },
    {
      "epoch": 2.51,
      "learning_rate": 5.39159109645507e-06,
      "loss": 3.819,
      "step": 4486
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.385408079142622e-06,
      "loss": 3.7435,
      "step": 4487
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.379225061830173e-06,
      "loss": 3.8183,
      "step": 4488
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.373042044517725e-06,
      "loss": 3.7548,
      "step": 4489
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.366859027205277e-06,
      "loss": 3.7173,
      "step": 4490
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.3606760098928285e-06,
      "loss": 3.8391,
      "step": 4491
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.3544929925803795e-06,
      "loss": 3.8496,
      "step": 4492
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.348309975267931e-06,
      "loss": 3.7519,
      "step": 4493
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.342126957955482e-06,
      "loss": 4.0674,
      "step": 4494
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.335943940643034e-06,
      "loss": 3.7315,
      "step": 4495
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.329760923330585e-06,
      "loss": 4.0416,
      "step": 4496
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.323577906018137e-06,
      "loss": 3.7946,
      "step": 4497
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.317394888705689e-06,
      "loss": 3.7308,
      "step": 4498
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.311211871393241e-06,
      "loss": 3.6317,
      "step": 4499
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.3050288540807916e-06,
      "loss": 3.7731,
      "step": 4500
    },
    {
      "epoch": 2.52,
      "eval_loss": 3.963338613510132,
      "eval_runtime": 1254.7061,
      "eval_samples_per_second": 2.106,
      "eval_steps_per_second": 0.264,
      "eval_wer": 1.0627021449689467,
      "step": 4500
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.298845836768343e-06,
      "loss": 3.7946,
      "step": 4501
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.292662819455894e-06,
      "loss": 3.8881,
      "step": 4502
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.286479802143446e-06,
      "loss": 3.9974,
      "step": 4503
    },
    {
      "epoch": 2.52,
      "learning_rate": 5.280296784830997e-06,
      "loss": 3.8196,
      "step": 4504
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.274113767518549e-06,
      "loss": 3.9778,
      "step": 4505
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.267930750206101e-06,
      "loss": 3.5675,
      "step": 4506
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.261747732893653e-06,
      "loss": 3.6256,
      "step": 4507
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.255564715581204e-06,
      "loss": 3.7976,
      "step": 4508
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.2493816982687555e-06,
      "loss": 3.6743,
      "step": 4509
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.2431986809563065e-06,
      "loss": 3.702,
      "step": 4510
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.237015663643858e-06,
      "loss": 3.5734,
      "step": 4511
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.230832646331409e-06,
      "loss": 3.8031,
      "step": 4512
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.224649629018961e-06,
      "loss": 3.6306,
      "step": 4513
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.218466611706513e-06,
      "loss": 3.6789,
      "step": 4514
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.212283594394065e-06,
      "loss": 3.2407,
      "step": 4515
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.206100577081616e-06,
      "loss": 3.2101,
      "step": 4516
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.1999175597691676e-06,
      "loss": 2.8256,
      "step": 4517
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.1937345424567185e-06,
      "loss": 2.5461,
      "step": 4518
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.18755152514427e-06,
      "loss": 4.2911,
      "step": 4519
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.181368507831822e-06,
      "loss": 3.9909,
      "step": 4520
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.175185490519373e-06,
      "loss": 4.1638,
      "step": 4521
    },
    {
      "epoch": 2.53,
      "learning_rate": 5.169002473206926e-06,
      "loss": 3.9759,
      "step": 4522
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.162819455894477e-06,
      "loss": 3.7529,
      "step": 4523
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.156636438582029e-06,
      "loss": 4.0058,
      "step": 4524
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.15045342126958e-06,
      "loss": 3.7657,
      "step": 4525
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.1442704039571315e-06,
      "loss": 4.0259,
      "step": 4526
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.1380873866446825e-06,
      "loss": 4.0475,
      "step": 4527
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.131904369332234e-06,
      "loss": 4.0633,
      "step": 4528
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.125721352019785e-06,
      "loss": 3.8851,
      "step": 4529
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.119538334707338e-06,
      "loss": 3.9656,
      "step": 4530
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.113355317394889e-06,
      "loss": 3.9845,
      "step": 4531
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.107172300082441e-06,
      "loss": 3.9516,
      "step": 4532
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.100989282769992e-06,
      "loss": 3.7905,
      "step": 4533
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.0948062654575436e-06,
      "loss": 3.8753,
      "step": 4534
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.0886232481450945e-06,
      "loss": 3.6278,
      "step": 4535
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.082440230832646e-06,
      "loss": 4.0098,
      "step": 4536
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.076257213520197e-06,
      "loss": 3.968,
      "step": 4537
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.07007419620775e-06,
      "loss": 3.7417,
      "step": 4538
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.063891178895301e-06,
      "loss": 3.7017,
      "step": 4539
    },
    {
      "epoch": 2.54,
      "learning_rate": 5.057708161582853e-06,
      "loss": 3.7224,
      "step": 4540
    },
    {
      "epoch": 2.55,
      "learning_rate": 5.051525144270404e-06,
      "loss": 3.7998,
      "step": 4541
    },
    {
      "epoch": 2.55,
      "learning_rate": 5.045342126957956e-06,
      "loss": 3.6593,
      "step": 4542
    },
    {
      "epoch": 2.55,
      "learning_rate": 5.039159109645507e-06,
      "loss": 3.6942,
      "step": 4543
    },
    {
      "epoch": 2.55,
      "learning_rate": 5.0329760923330585e-06,
      "loss": 3.8094,
      "step": 4544
    },
    {
      "epoch": 2.55,
      "learning_rate": 5.0267930750206094e-06,
      "loss": 3.8117,
      "step": 4545
    },
    {
      "epoch": 2.55,
      "learning_rate": 5.020610057708162e-06,
      "loss": 3.6456,
      "step": 4546
    },
    {
      "epoch": 2.55,
      "learning_rate": 5.014427040395713e-06,
      "loss": 3.9972,
      "step": 4547
    },
    {
      "epoch": 2.55,
      "learning_rate": 5.008244023083265e-06,
      "loss": 4.0688,
      "step": 4548
    },
    {
      "epoch": 2.55,
      "learning_rate": 5.002061005770816e-06,
      "loss": 3.719,
      "step": 4549
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.995877988458368e-06,
      "loss": 3.894,
      "step": 4550
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.9896949711459196e-06,
      "loss": 3.7714,
      "step": 4551
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.9835119538334705e-06,
      "loss": 3.7588,
      "step": 4552
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.977328936521022e-06,
      "loss": 3.6285,
      "step": 4553
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.971145919208574e-06,
      "loss": 3.9666,
      "step": 4554
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.964962901896126e-06,
      "loss": 3.6765,
      "step": 4555
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.958779884583677e-06,
      "loss": 3.7983,
      "step": 4556
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.952596867271229e-06,
      "loss": 3.8315,
      "step": 4557
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.94641384995878e-06,
      "loss": 3.6089,
      "step": 4558
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.940230832646332e-06,
      "loss": 3.6158,
      "step": 4559
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.934047815333883e-06,
      "loss": 3.4232,
      "step": 4560
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.9278647980214345e-06,
      "loss": 3.7013,
      "step": 4561
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.921681780708986e-06,
      "loss": 3.6202,
      "step": 4562
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.915498763396538e-06,
      "loss": 3.3511,
      "step": 4563
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.909315746084089e-06,
      "loss": 3.442,
      "step": 4564
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.903132728771641e-06,
      "loss": 3.4376,
      "step": 4565
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.896949711459192e-06,
      "loss": 3.1641,
      "step": 4566
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.890766694146744e-06,
      "loss": 2.8593,
      "step": 4567
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.884583676834295e-06,
      "loss": 3.0553,
      "step": 4568
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.8784006595218465e-06,
      "loss": 3.7617,
      "step": 4569
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.872217642209398e-06,
      "loss": 3.7648,
      "step": 4570
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.86603462489695e-06,
      "loss": 3.9893,
      "step": 4571
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.859851607584501e-06,
      "loss": 3.8461,
      "step": 4572
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.853668590272053e-06,
      "loss": 3.9737,
      "step": 4573
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.847485572959604e-06,
      "loss": 3.7891,
      "step": 4574
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.841302555647156e-06,
      "loss": 3.784,
      "step": 4575
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.835119538334707e-06,
      "loss": 3.792,
      "step": 4576
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.828936521022259e-06,
      "loss": 3.7586,
      "step": 4577
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.8227535037098104e-06,
      "loss": 3.7329,
      "step": 4578
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.816570486397362e-06,
      "loss": 3.8837,
      "step": 4579
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.810387469084913e-06,
      "loss": 3.5845,
      "step": 4580
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.804204451772465e-06,
      "loss": 4.0274,
      "step": 4581
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.798021434460017e-06,
      "loss": 3.8449,
      "step": 4582
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.791838417147568e-06,
      "loss": 3.8703,
      "step": 4583
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.78565539983512e-06,
      "loss": 3.8063,
      "step": 4584
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.779472382522671e-06,
      "loss": 4.1964,
      "step": 4585
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.773289365210223e-06,
      "loss": 3.6743,
      "step": 4586
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.767106347897774e-06,
      "loss": 4.1151,
      "step": 4587
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.760923330585326e-06,
      "loss": 3.5179,
      "step": 4588
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.754740313272877e-06,
      "loss": 3.813,
      "step": 4589
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.748557295960429e-06,
      "loss": 3.8576,
      "step": 4590
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.74237427864798e-06,
      "loss": 3.9086,
      "step": 4591
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.736191261335532e-06,
      "loss": 3.9179,
      "step": 4592
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.730008244023084e-06,
      "loss": 3.9147,
      "step": 4593
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.7238252267106355e-06,
      "loss": 3.8401,
      "step": 4594
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.7176422093981864e-06,
      "loss": 3.7463,
      "step": 4595
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.711459192085738e-06,
      "loss": 3.8067,
      "step": 4596
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.705276174773289e-06,
      "loss": 3.6151,
      "step": 4597
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.699093157460841e-06,
      "loss": 4.0017,
      "step": 4598
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.692910140148392e-06,
      "loss": 3.6579,
      "step": 4599
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.686727122835944e-06,
      "loss": 3.8601,
      "step": 4600
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.680544105523496e-06,
      "loss": 3.7814,
      "step": 4601
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.6743610882110475e-06,
      "loss": 3.615,
      "step": 4602
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.6681780708985985e-06,
      "loss": 3.9147,
      "step": 4603
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.66199505358615e-06,
      "loss": 3.6874,
      "step": 4604
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.655812036273701e-06,
      "loss": 3.8544,
      "step": 4605
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.649629018961253e-06,
      "loss": 3.9608,
      "step": 4606
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.643446001648804e-06,
      "loss": 3.7459,
      "step": 4607
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.637262984336356e-06,
      "loss": 3.6346,
      "step": 4608
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.631079967023908e-06,
      "loss": 3.7202,
      "step": 4609
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.62489694971146e-06,
      "loss": 3.797,
      "step": 4610
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.618713932399011e-06,
      "loss": 3.6136,
      "step": 4611
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.6125309150865624e-06,
      "loss": 3.4288,
      "step": 4612
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.606347897774114e-06,
      "loss": 3.8531,
      "step": 4613
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.600164880461665e-06,
      "loss": 3.1832,
      "step": 4614
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.593981863149217e-06,
      "loss": 3.0291,
      "step": 4615
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.587798845836768e-06,
      "loss": 2.9777,
      "step": 4616
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.581615828524321e-06,
      "loss": 3.0248,
      "step": 4617
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.575432811211872e-06,
      "loss": 2.3455,
      "step": 4618
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.5692497938994235e-06,
      "loss": 4.3673,
      "step": 4619
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.5630667765869745e-06,
      "loss": 3.9453,
      "step": 4620
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.556883759274526e-06,
      "loss": 3.8634,
      "step": 4621
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.550700741962077e-06,
      "loss": 3.9287,
      "step": 4622
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.544517724649629e-06,
      "loss": 3.8987,
      "step": 4623
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.53833470733718e-06,
      "loss": 4.1309,
      "step": 4624
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.532151690024733e-06,
      "loss": 4.1106,
      "step": 4625
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.525968672712284e-06,
      "loss": 3.9561,
      "step": 4626
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.519785655399836e-06,
      "loss": 3.9098,
      "step": 4627
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.513602638087387e-06,
      "loss": 3.8422,
      "step": 4628
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.5074196207749384e-06,
      "loss": 4.019,
      "step": 4629
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.501236603462489e-06,
      "loss": 3.965,
      "step": 4630
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.495053586150041e-06,
      "loss": 3.7668,
      "step": 4631
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.488870568837592e-06,
      "loss": 3.9179,
      "step": 4632
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.482687551525145e-06,
      "loss": 3.8316,
      "step": 4633
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.476504534212696e-06,
      "loss": 3.9113,
      "step": 4634
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.470321516900248e-06,
      "loss": 3.6639,
      "step": 4635
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.464138499587799e-06,
      "loss": 3.8876,
      "step": 4636
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.4579554822753505e-06,
      "loss": 3.7916,
      "step": 4637
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.4517724649629015e-06,
      "loss": 3.7801,
      "step": 4638
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.445589447650453e-06,
      "loss": 3.9744,
      "step": 4639
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.439406430338004e-06,
      "loss": 3.7543,
      "step": 4640
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.433223413025557e-06,
      "loss": 3.6879,
      "step": 4641
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.427040395713108e-06,
      "loss": 3.8769,
      "step": 4642
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.42085737840066e-06,
      "loss": 3.8233,
      "step": 4643
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.414674361088212e-06,
      "loss": 3.7875,
      "step": 4644
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.408491343775763e-06,
      "loss": 3.6643,
      "step": 4645
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.402308326463314e-06,
      "loss": 3.9415,
      "step": 4646
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.396125309150865e-06,
      "loss": 3.8753,
      "step": 4647
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.389942291838417e-06,
      "loss": 3.8698,
      "step": 4648
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.383759274525969e-06,
      "loss": 3.8728,
      "step": 4649
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.377576257213521e-06,
      "loss": 3.5115,
      "step": 4650
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.371393239901072e-06,
      "loss": 3.7474,
      "step": 4651
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.365210222588624e-06,
      "loss": 3.5554,
      "step": 4652
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.359027205276175e-06,
      "loss": 4.0653,
      "step": 4653
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.3528441879637265e-06,
      "loss": 3.8945,
      "step": 4654
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.3466611706512775e-06,
      "loss": 3.6825,
      "step": 4655
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.340478153338829e-06,
      "loss": 4.0393,
      "step": 4656
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.334295136026381e-06,
      "loss": 3.756,
      "step": 4657
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.328112118713933e-06,
      "loss": 3.6127,
      "step": 4658
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.321929101401484e-06,
      "loss": 3.5492,
      "step": 4659
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.315746084089036e-06,
      "loss": 3.5592,
      "step": 4660
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.309563066776587e-06,
      "loss": 3.7317,
      "step": 4661
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.303380049464139e-06,
      "loss": 3.5244,
      "step": 4662
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.2971970321516896e-06,
      "loss": 3.3188,
      "step": 4663
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.291014014839241e-06,
      "loss": 3.3131,
      "step": 4664
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.284830997526793e-06,
      "loss": 3.2921,
      "step": 4665
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.278647980214345e-06,
      "loss": 3.002,
      "step": 4666
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.272464962901896e-06,
      "loss": 3.2091,
      "step": 4667
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.266281945589448e-06,
      "loss": 2.5511,
      "step": 4668
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.260098928276999e-06,
      "loss": 3.9188,
      "step": 4669
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.253915910964551e-06,
      "loss": 4.0579,
      "step": 4670
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.247732893652102e-06,
      "loss": 4.0869,
      "step": 4671
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.2415498763396535e-06,
      "loss": 3.9254,
      "step": 4672
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.235366859027205e-06,
      "loss": 3.7126,
      "step": 4673
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.229183841714757e-06,
      "loss": 3.9907,
      "step": 4674
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.223000824402309e-06,
      "loss": 4.0093,
      "step": 4675
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.21681780708986e-06,
      "loss": 3.9707,
      "step": 4676
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.210634789777412e-06,
      "loss": 3.7078,
      "step": 4677
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.204451772464963e-06,
      "loss": 3.8661,
      "step": 4678
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.198268755152515e-06,
      "loss": 4.0016,
      "step": 4679
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.1920857378400656e-06,
      "loss": 4.0056,
      "step": 4680
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.185902720527618e-06,
      "loss": 3.7915,
      "step": 4681
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.179719703215169e-06,
      "loss": 3.6205,
      "step": 4682
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.173536685902721e-06,
      "loss": 3.9813,
      "step": 4683
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.167353668590272e-06,
      "loss": 3.7238,
      "step": 4684
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.161170651277824e-06,
      "loss": 3.9849,
      "step": 4685
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.154987633965375e-06,
      "loss": 3.9025,
      "step": 4686
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.148804616652927e-06,
      "loss": 3.8612,
      "step": 4687
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.142621599340478e-06,
      "loss": 3.6942,
      "step": 4688
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.13643858202803e-06,
      "loss": 3.8457,
      "step": 4689
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.130255564715581e-06,
      "loss": 3.7266,
      "step": 4690
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.124072547403133e-06,
      "loss": 3.9244,
      "step": 4691
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.117889530090684e-06,
      "loss": 3.9183,
      "step": 4692
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.111706512778236e-06,
      "loss": 3.73,
      "step": 4693
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.105523495465787e-06,
      "loss": 3.6834,
      "step": 4694
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.099340478153339e-06,
      "loss": 3.7531,
      "step": 4695
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.09315746084089e-06,
      "loss": 3.9024,
      "step": 4696
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.086974443528442e-06,
      "loss": 3.6562,
      "step": 4697
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.080791426215993e-06,
      "loss": 3.8962,
      "step": 4698
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.074608408903545e-06,
      "loss": 3.7805,
      "step": 4699
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.068425391591096e-06,
      "loss": 3.7325,
      "step": 4700
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.062242374278648e-06,
      "loss": 3.8449,
      "step": 4701
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.056059356966199e-06,
      "loss": 3.6593,
      "step": 4702
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.049876339653751e-06,
      "loss": 3.8851,
      "step": 4703
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.043693322341303e-06,
      "loss": 3.8891,
      "step": 4704
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.0375103050288545e-06,
      "loss": 3.8363,
      "step": 4705
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.031327287716406e-06,
      "loss": 3.9192,
      "step": 4706
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.025144270403957e-06,
      "loss": 3.7228,
      "step": 4707
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.018961253091509e-06,
      "loss": 3.7766,
      "step": 4708
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.01277823577906e-06,
      "loss": 3.419,
      "step": 4709
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.006595218466612e-06,
      "loss": 3.7928,
      "step": 4710
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.000412201154163e-06,
      "loss": 3.4733,
      "step": 4711
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.994229183841716e-06,
      "loss": 3.4333,
      "step": 4712
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.9880461665292666e-06,
      "loss": 3.3686,
      "step": 4713
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.981863149216818e-06,
      "loss": 3.3474,
      "step": 4714
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.975680131904369e-06,
      "loss": 2.9665,
      "step": 4715
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.969497114591921e-06,
      "loss": 3.4523,
      "step": 4716
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.963314097279472e-06,
      "loss": 3.1559,
      "step": 4717
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.957131079967024e-06,
      "loss": 2.9838,
      "step": 4718
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.950948062654575e-06,
      "loss": 4.0729,
      "step": 4719
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.944765045342128e-06,
      "loss": 3.979,
      "step": 4720
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.938582028029679e-06,
      "loss": 3.967,
      "step": 4721
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.9323990107172305e-06,
      "loss": 3.8282,
      "step": 4722
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.9262159934047815e-06,
      "loss": 3.7758,
      "step": 4723
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.920032976092333e-06,
      "loss": 4.1319,
      "step": 4724
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.913849958779884e-06,
      "loss": 3.9354,
      "step": 4725
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.907666941467436e-06,
      "loss": 3.9751,
      "step": 4726
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.901483924154987e-06,
      "loss": 3.8022,
      "step": 4727
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.89530090684254e-06,
      "loss": 3.6797,
      "step": 4728
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.889117889530091e-06,
      "loss": 3.9457,
      "step": 4729
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.8829348722176426e-06,
      "loss": 4.1442,
      "step": 4730
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.8767518549051935e-06,
      "loss": 3.7542,
      "step": 4731
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.870568837592745e-06,
      "loss": 4.0951,
      "step": 4732
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.864385820280296e-06,
      "loss": 3.6932,
      "step": 4733
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.858202802967848e-06,
      "loss": 3.939,
      "step": 4734
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.852019785655399e-06,
      "loss": 3.857,
      "step": 4735
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.845836768342952e-06,
      "loss": 4.0835,
      "step": 4736
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.839653751030504e-06,
      "loss": 3.7499,
      "step": 4737
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.833470733718055e-06,
      "loss": 3.6524,
      "step": 4738
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.8272877164056065e-06,
      "loss": 3.7582,
      "step": 4739
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.8211046990931575e-06,
      "loss": 3.6861,
      "step": 4740
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.814921681780709e-06,
      "loss": 3.7435,
      "step": 4741
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.8087386644682603e-06,
      "loss": 3.6263,
      "step": 4742
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.8025556471558117e-06,
      "loss": 3.8614,
      "step": 4743
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.796372629843364e-06,
      "loss": 3.8188,
      "step": 4744
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.7901896125309153e-06,
      "loss": 3.7487,
      "step": 4745
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.7840065952184667e-06,
      "loss": 3.7966,
      "step": 4746
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.777823577906018e-06,
      "loss": 3.8924,
      "step": 4747
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.7716405605935695e-06,
      "loss": 3.6449,
      "step": 4748
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.765457543281121e-06,
      "loss": 3.766,
      "step": 4749
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.7592745259686724e-06,
      "loss": 4.002,
      "step": 4750
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.7530915086562238e-06,
      "loss": 3.8822,
      "step": 4751
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.746908491343776e-06,
      "loss": 3.5962,
      "step": 4752
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.7407254740313274e-06,
      "loss": 3.6991,
      "step": 4753
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.7345424567188792e-06,
      "loss": 3.6666,
      "step": 4754
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.7283594394064306e-06,
      "loss": 3.7807,
      "step": 4755
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.722176422093982e-06,
      "loss": 3.9212,
      "step": 4756
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.7159934047815335e-06,
      "loss": 3.764,
      "step": 4757
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.7098103874690853e-06,
      "loss": 3.5566,
      "step": 4758
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.7036273701566367e-06,
      "loss": 3.6345,
      "step": 4759
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.697444352844188e-06,
      "loss": 4.0083,
      "step": 4760
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6912613355317395e-06,
      "loss": 3.6725,
      "step": 4761
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6850783182192913e-06,
      "loss": 3.5114,
      "step": 4762
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6788953009068427e-06,
      "loss": 3.5092,
      "step": 4763
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.672712283594394e-06,
      "loss": 3.3726,
      "step": 4764
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6665292662819455e-06,
      "loss": 3.3765,
      "step": 4765
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6603462489694974e-06,
      "loss": 3.0411,
      "step": 4766
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6541632316570488e-06,
      "loss": 3.2353,
      "step": 4767
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6479802143446e-06,
      "loss": 2.4941,
      "step": 4768
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6417971970321516e-06,
      "loss": 4.0059,
      "step": 4769
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6356141797197034e-06,
      "loss": 4.0135,
      "step": 4770
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.629431162407255e-06,
      "loss": 3.9318,
      "step": 4771
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.6232481450948062e-06,
      "loss": 3.9439,
      "step": 4772
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.6170651277823576e-06,
      "loss": 3.8541,
      "step": 4773
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.6108821104699094e-06,
      "loss": 3.9035,
      "step": 4774
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.604699093157461e-06,
      "loss": 3.88,
      "step": 4775
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5985160758450123e-06,
      "loss": 3.8752,
      "step": 4776
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5923330585325637e-06,
      "loss": 3.9311,
      "step": 4777
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5861500412201155e-06,
      "loss": 3.4588,
      "step": 4778
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.579967023907667e-06,
      "loss": 3.7393,
      "step": 4779
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5737840065952183e-06,
      "loss": 3.8342,
      "step": 4780
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5676009892827697e-06,
      "loss": 3.7617,
      "step": 4781
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5614179719703215e-06,
      "loss": 3.9893,
      "step": 4782
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5552349546578734e-06,
      "loss": 3.5152,
      "step": 4783
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5490519373454248e-06,
      "loss": 3.6553,
      "step": 4784
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.542868920032976e-06,
      "loss": 3.7422,
      "step": 4785
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.536685902720528e-06,
      "loss": 3.7303,
      "step": 4786
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5305028854080794e-06,
      "loss": 3.9061,
      "step": 4787
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.524319868095631e-06,
      "loss": 3.798,
      "step": 4788
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.5181368507831822e-06,
      "loss": 3.7967,
      "step": 4789
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.511953833470734e-06,
      "loss": 4.1444,
      "step": 4790
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.5057708161582854e-06,
      "loss": 3.7521,
      "step": 4791
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.499587798845837e-06,
      "loss": 3.9263,
      "step": 4792
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.4934047815333883e-06,
      "loss": 3.729,
      "step": 4793
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.48722176422094e-06,
      "loss": 3.7595,
      "step": 4794
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.4810387469084915e-06,
      "loss": 3.8673,
      "step": 4795
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.474855729596043e-06,
      "loss": 3.7466,
      "step": 4796
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.4686727122835943e-06,
      "loss": 3.8273,
      "step": 4797
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.462489694971146e-06,
      "loss": 3.9152,
      "step": 4798
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.4563066776586975e-06,
      "loss": 3.6065,
      "step": 4799
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.450123660346249e-06,
      "loss": 3.7252,
      "step": 4800
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.4439406430338003e-06,
      "loss": 3.7231,
      "step": 4801
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.437757625721352e-06,
      "loss": 3.9628,
      "step": 4802
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.4315746084089036e-06,
      "loss": 3.6315,
      "step": 4803
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.425391591096455e-06,
      "loss": 3.4909,
      "step": 4804
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.4192085737840064e-06,
      "loss": 3.9501,
      "step": 4805
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.413025556471558e-06,
      "loss": 3.6114,
      "step": 4806
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.4068425391591096e-06,
      "loss": 3.8401,
      "step": 4807
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.400659521846661e-06,
      "loss": 3.5931,
      "step": 4808
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.394476504534213e-06,
      "loss": 3.4675,
      "step": 4809
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3882934872217642e-06,
      "loss": 3.7776,
      "step": 4810
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3821104699093157e-06,
      "loss": 3.6035,
      "step": 4811
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.375927452596867e-06,
      "loss": 3.7422,
      "step": 4812
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.369744435284419e-06,
      "loss": 3.4169,
      "step": 4813
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3635614179719707e-06,
      "loss": 3.5634,
      "step": 4814
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.357378400659522e-06,
      "loss": 3.0019,
      "step": 4815
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3511953833470735e-06,
      "loss": 3.2856,
      "step": 4816
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3450123660346254e-06,
      "loss": 3.0656,
      "step": 4817
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3388293487221768e-06,
      "loss": 3.039,
      "step": 4818
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.332646331409728e-06,
      "loss": 3.988,
      "step": 4819
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3264633140972796e-06,
      "loss": 4.0897,
      "step": 4820
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3202802967848314e-06,
      "loss": 3.9508,
      "step": 4821
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.314097279472383e-06,
      "loss": 3.8714,
      "step": 4822
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.307914262159934e-06,
      "loss": 3.7647,
      "step": 4823
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3017312448474856e-06,
      "loss": 3.7796,
      "step": 4824
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.2955482275350374e-06,
      "loss": 3.9359,
      "step": 4825
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.289365210222589e-06,
      "loss": 3.7497,
      "step": 4826
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2831821929101402e-06,
      "loss": 4.015,
      "step": 4827
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2769991755976916e-06,
      "loss": 3.6184,
      "step": 4828
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2708161582852435e-06,
      "loss": 3.8439,
      "step": 4829
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.264633140972795e-06,
      "loss": 3.8356,
      "step": 4830
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2584501236603463e-06,
      "loss": 3.5604,
      "step": 4831
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2522671063478977e-06,
      "loss": 3.7283,
      "step": 4832
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2460840890354495e-06,
      "loss": 4.0685,
      "step": 4833
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.239901071723001e-06,
      "loss": 3.8503,
      "step": 4834
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2337180544105523e-06,
      "loss": 3.9037,
      "step": 4835
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2275350370981037e-06,
      "loss": 3.7351,
      "step": 4836
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2213520197856556e-06,
      "loss": 3.8985,
      "step": 4837
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.215169002473207e-06,
      "loss": 3.729,
      "step": 4838
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2089859851607584e-06,
      "loss": 3.7014,
      "step": 4839
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.2028029678483098e-06,
      "loss": 3.7267,
      "step": 4840
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.1966199505358616e-06,
      "loss": 4.0147,
      "step": 4841
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.190436933223413e-06,
      "loss": 3.8244,
      "step": 4842
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.1842539159109644e-06,
      "loss": 3.6844,
      "step": 4843
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.178070898598516e-06,
      "loss": 3.7092,
      "step": 4844
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.1718878812860676e-06,
      "loss": 3.7479,
      "step": 4845
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.1657048639736195e-06,
      "loss": 3.8789,
      "step": 4846
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.159521846661171e-06,
      "loss": 3.4155,
      "step": 4847
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.1533388293487223e-06,
      "loss": 3.6324,
      "step": 4848
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.147155812036274e-06,
      "loss": 3.8015,
      "step": 4849
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.1409727947238255e-06,
      "loss": 3.9367,
      "step": 4850
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.134789777411377e-06,
      "loss": 3.6392,
      "step": 4851
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.1286067600989283e-06,
      "loss": 3.6896,
      "step": 4852
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.12242374278648e-06,
      "loss": 3.6616,
      "step": 4853
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.1162407254740316e-06,
      "loss": 4.0873,
      "step": 4854
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.110057708161583e-06,
      "loss": 3.8618,
      "step": 4855
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.1038746908491344e-06,
      "loss": 3.7201,
      "step": 4856
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.097691673536686e-06,
      "loss": 3.7649,
      "step": 4857
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.0915086562242376e-06,
      "loss": 3.6531,
      "step": 4858
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.085325638911789e-06,
      "loss": 3.5815,
      "step": 4859
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.0791426215993404e-06,
      "loss": 3.4923,
      "step": 4860
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.0729596042868922e-06,
      "loss": 3.4985,
      "step": 4861
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.0667765869744436e-06,
      "loss": 3.9417,
      "step": 4862
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.060593569661995e-06,
      "loss": 3.4746,
      "step": 4863
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.0544105523495464e-06,
      "loss": 3.4973,
      "step": 4864
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.0482275350370983e-06,
      "loss": 3.1624,
      "step": 4865
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.0420445177246497e-06,
      "loss": 3.0464,
      "step": 4866
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.035861500412201e-06,
      "loss": 2.4788,
      "step": 4867
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.0296784830997525e-06,
      "loss": 2.6501,
      "step": 4868
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.0234954657873043e-06,
      "loss": 3.9303,
      "step": 4869
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.0173124484748557e-06,
      "loss": 3.858,
      "step": 4870
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.011129431162407e-06,
      "loss": 3.9781,
      "step": 4871
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.0049464138499585e-06,
      "loss": 4.0131,
      "step": 4872
    },
    {
      "epoch": 2.73,
      "learning_rate": 2.9987633965375104e-06,
      "loss": 3.991,
      "step": 4873
    },
    {
      "epoch": 2.73,
      "learning_rate": 2.9925803792250618e-06,
      "loss": 3.9049,
      "step": 4874
    },
    {
      "epoch": 2.73,
      "learning_rate": 2.986397361912613e-06,
      "loss": 3.908,
      "step": 4875
    },
    {
      "epoch": 2.73,
      "learning_rate": 2.980214344600165e-06,
      "loss": 3.8902,
      "step": 4876
    },
    {
      "epoch": 2.73,
      "learning_rate": 2.974031327287717e-06,
      "loss": 3.7654,
      "step": 4877
    },
    {
      "epoch": 2.73,
      "learning_rate": 2.9678483099752682e-06,
      "loss": 3.6402,
      "step": 4878
    },
    {
      "epoch": 2.73,
      "learning_rate": 2.9616652926628196e-06,
      "loss": 3.6467,
      "step": 4879
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.955482275350371e-06,
      "loss": 4.0762,
      "step": 4880
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.949299258037923e-06,
      "loss": 3.7885,
      "step": 4881
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.9431162407254743e-06,
      "loss": 3.9098,
      "step": 4882
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.9369332234130257e-06,
      "loss": 3.7926,
      "step": 4883
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.930750206100577e-06,
      "loss": 3.7176,
      "step": 4884
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.924567188788129e-06,
      "loss": 4.1639,
      "step": 4885
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.9183841714756803e-06,
      "loss": 3.6547,
      "step": 4886
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.9122011541632317e-06,
      "loss": 3.9668,
      "step": 4887
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.906018136850783e-06,
      "loss": 3.6653,
      "step": 4888
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.899835119538335e-06,
      "loss": 4.166,
      "step": 4889
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.8936521022258864e-06,
      "loss": 3.6809,
      "step": 4890
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.8874690849134378e-06,
      "loss": 3.7447,
      "step": 4891
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.881286067600989e-06,
      "loss": 3.8584,
      "step": 4892
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.875103050288541e-06,
      "loss": 3.9242,
      "step": 4893
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.8689200329760924e-06,
      "loss": 3.7339,
      "step": 4894
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.862737015663644e-06,
      "loss": 3.6486,
      "step": 4895
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.856553998351195e-06,
      "loss": 3.8556,
      "step": 4896
    },
    {
      "epoch": 2.74,
      "learning_rate": 2.850370981038747e-06,
      "loss": 3.7267,
      "step": 4897
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.8441879637262984e-06,
      "loss": 3.8765,
      "step": 4898
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.83800494641385e-06,
      "loss": 3.7752,
      "step": 4899
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.8318219291014012e-06,
      "loss": 3.7787,
      "step": 4900
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.825638911788953e-06,
      "loss": 3.6425,
      "step": 4901
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.8194558944765045e-06,
      "loss": 3.9642,
      "step": 4902
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.813272877164056e-06,
      "loss": 3.5663,
      "step": 4903
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.8070898598516073e-06,
      "loss": 3.5651,
      "step": 4904
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.800906842539159e-06,
      "loss": 3.7924,
      "step": 4905
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.7947238252267105e-06,
      "loss": 3.6582,
      "step": 4906
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.7885408079142623e-06,
      "loss": 3.3774,
      "step": 4907
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.7823577906018138e-06,
      "loss": 3.6791,
      "step": 4908
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.7761747732893656e-06,
      "loss": 3.4624,
      "step": 4909
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.769991755976917e-06,
      "loss": 3.793,
      "step": 4910
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.7638087386644684e-06,
      "loss": 3.6108,
      "step": 4911
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.75762572135202e-06,
      "loss": 3.6356,
      "step": 4912
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.7514427040395716e-06,
      "loss": 3.4061,
      "step": 4913
    },
    {
      "epoch": 2.75,
      "learning_rate": 2.745259686727123e-06,
      "loss": 3.422,
      "step": 4914
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.7390766694146744e-06,
      "loss": 3.1048,
      "step": 4915
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.732893652102226e-06,
      "loss": 3.0825,
      "step": 4916
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.7267106347897777e-06,
      "loss": 3.0617,
      "step": 4917
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.720527617477329e-06,
      "loss": 2.842,
      "step": 4918
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.7143446001648805e-06,
      "loss": 3.8968,
      "step": 4919
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.7081615828524323e-06,
      "loss": 3.9707,
      "step": 4920
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.7019785655399837e-06,
      "loss": 4.0885,
      "step": 4921
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.695795548227535e-06,
      "loss": 4.113,
      "step": 4922
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.6896125309150865e-06,
      "loss": 3.8827,
      "step": 4923
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.6834295136026383e-06,
      "loss": 4.0734,
      "step": 4924
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.6772464962901897e-06,
      "loss": 3.9176,
      "step": 4925
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.671063478977741e-06,
      "loss": 3.7914,
      "step": 4926
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.6648804616652926e-06,
      "loss": 3.8567,
      "step": 4927
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.6586974443528444e-06,
      "loss": 4.1076,
      "step": 4928
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.6525144270403958e-06,
      "loss": 3.7881,
      "step": 4929
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.646331409727947e-06,
      "loss": 3.7811,
      "step": 4930
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.6401483924154986e-06,
      "loss": 3.9769,
      "step": 4931
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.6339653751030504e-06,
      "loss": 4.0463,
      "step": 4932
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.627782357790602e-06,
      "loss": 3.8973,
      "step": 4933
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.6215993404781532e-06,
      "loss": 3.8335,
      "step": 4934
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.6154163231657046e-06,
      "loss": 3.7573,
      "step": 4935
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.6092333058532565e-06,
      "loss": 4.0395,
      "step": 4936
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.603050288540808e-06,
      "loss": 3.8969,
      "step": 4937
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.5968672712283593e-06,
      "loss": 3.8499,
      "step": 4938
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.590684253915911e-06,
      "loss": 3.9732,
      "step": 4939
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.584501236603463e-06,
      "loss": 3.64,
      "step": 4940
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.5783182192910143e-06,
      "loss": 3.7841,
      "step": 4941
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.5721352019785657e-06,
      "loss": 3.7246,
      "step": 4942
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.565952184666117e-06,
      "loss": 3.6615,
      "step": 4943
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.559769167353669e-06,
      "loss": 3.931,
      "step": 4944
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.5535861500412204e-06,
      "loss": 3.7489,
      "step": 4945
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.5474031327287718e-06,
      "loss": 3.885,
      "step": 4946
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.541220115416323e-06,
      "loss": 3.694,
      "step": 4947
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.535037098103875e-06,
      "loss": 4.0883,
      "step": 4948
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.5288540807914264e-06,
      "loss": 3.9706,
      "step": 4949
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.522671063478978e-06,
      "loss": 3.7489,
      "step": 4950
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.5164880461665292e-06,
      "loss": 3.7757,
      "step": 4951
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.510305028854081e-06,
      "loss": 3.7325,
      "step": 4952
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.5041220115416325e-06,
      "loss": 3.7513,
      "step": 4953
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.497938994229184e-06,
      "loss": 3.6335,
      "step": 4954
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.4917559769167353e-06,
      "loss": 3.664,
      "step": 4955
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.485572959604287e-06,
      "loss": 3.7747,
      "step": 4956
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.4793899422918385e-06,
      "loss": 3.9183,
      "step": 4957
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.47320692497939e-06,
      "loss": 3.8372,
      "step": 4958
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.4670239076669413e-06,
      "loss": 3.6021,
      "step": 4959
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.460840890354493e-06,
      "loss": 3.8294,
      "step": 4960
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.4546578730420445e-06,
      "loss": 3.3607,
      "step": 4961
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.448474855729596e-06,
      "loss": 3.5014,
      "step": 4962
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.4422918384171474e-06,
      "loss": 3.5976,
      "step": 4963
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.436108821104699e-06,
      "loss": 3.6248,
      "step": 4964
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.4299258037922506e-06,
      "loss": 2.8797,
      "step": 4965
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.423742786479802e-06,
      "loss": 3.1199,
      "step": 4966
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.4175597691673534e-06,
      "loss": 2.8917,
      "step": 4967
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.4113767518549052e-06,
      "loss": 2.6861,
      "step": 4968
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.4051937345424566e-06,
      "loss": 3.8569,
      "step": 4969
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.3990107172300085e-06,
      "loss": 3.8381,
      "step": 4970
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.39282769991756e-06,
      "loss": 4.0642,
      "step": 4971
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.3866446826051117e-06,
      "loss": 3.8657,
      "step": 4972
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.380461665292663e-06,
      "loss": 3.9632,
      "step": 4973
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.3742786479802145e-06,
      "loss": 3.6956,
      "step": 4974
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.368095630667766e-06,
      "loss": 3.7776,
      "step": 4975
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.3619126133553177e-06,
      "loss": 4.1026,
      "step": 4976
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.355729596042869e-06,
      "loss": 3.7899,
      "step": 4977
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.3495465787304205e-06,
      "loss": 4.0158,
      "step": 4978
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.343363561417972e-06,
      "loss": 3.9184,
      "step": 4979
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.3371805441055238e-06,
      "loss": 3.8224,
      "step": 4980
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.330997526793075e-06,
      "loss": 3.8538,
      "step": 4981
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.3248145094806266e-06,
      "loss": 3.7059,
      "step": 4982
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.318631492168178e-06,
      "loss": 3.8667,
      "step": 4983
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.31244847485573e-06,
      "loss": 3.5736,
      "step": 4984
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.3062654575432812e-06,
      "loss": 3.819,
      "step": 4985
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.3000824402308326e-06,
      "loss": 3.8982,
      "step": 4986
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.293899422918384e-06,
      "loss": 3.7269,
      "step": 4987
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.287716405605936e-06,
      "loss": 3.9318,
      "step": 4988
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2815333882934873e-06,
      "loss": 3.7877,
      "step": 4989
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2753503709810387e-06,
      "loss": 4.064,
      "step": 4990
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.26916735366859e-06,
      "loss": 3.8589,
      "step": 4991
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.262984336356142e-06,
      "loss": 3.759,
      "step": 4992
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2568013190436933e-06,
      "loss": 3.8188,
      "step": 4993
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2506183017312447e-06,
      "loss": 3.6922,
      "step": 4994
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.244435284418796e-06,
      "loss": 3.9492,
      "step": 4995
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.238252267106348e-06,
      "loss": 3.618,
      "step": 4996
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2320692497938993e-06,
      "loss": 4.0176,
      "step": 4997
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2258862324814507e-06,
      "loss": 3.8433,
      "step": 4998
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.219703215169002e-06,
      "loss": 3.7851,
      "step": 4999
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.213520197856554e-06,
      "loss": 3.6984,
      "step": 5000
    },
    {
      "epoch": 2.8,
      "eval_loss": 3.9280078411102295,
      "eval_runtime": 1200.8405,
      "eval_samples_per_second": 2.2,
      "eval_steps_per_second": 0.276,
      "eval_wer": 0.9561084985217374,
      "step": 5000
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.207337180544106e-06,
      "loss": 3.7693,
      "step": 5001
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.201154163231657e-06,
      "loss": 3.8946,
      "step": 5002
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.1949711459192086e-06,
      "loss": 3.6278,
      "step": 5003
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.1887881286067604e-06,
      "loss": 3.6871,
      "step": 5004
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.182605111294312e-06,
      "loss": 3.7711,
      "step": 5005
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1764220939818633e-06,
      "loss": 3.8787,
      "step": 5006
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1702390766694147e-06,
      "loss": 3.827,
      "step": 5007
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1640560593569665e-06,
      "loss": 3.8409,
      "step": 5008
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.157873042044518e-06,
      "loss": 3.4782,
      "step": 5009
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1516900247320693e-06,
      "loss": 3.6091,
      "step": 5010
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1455070074196207e-06,
      "loss": 3.852,
      "step": 5011
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1393239901071725e-06,
      "loss": 3.6878,
      "step": 5012
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.133140972794724e-06,
      "loss": 3.6957,
      "step": 5013
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1269579554822753e-06,
      "loss": 3.6893,
      "step": 5014
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1207749381698267e-06,
      "loss": 3.2957,
      "step": 5015
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1145919208573786e-06,
      "loss": 3.261,
      "step": 5016
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.10840890354493e-06,
      "loss": 3.0586,
      "step": 5017
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1022258862324814e-06,
      "loss": 2.4403,
      "step": 5018
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0960428689200328e-06,
      "loss": 3.7924,
      "step": 5019
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0898598516075846e-06,
      "loss": 3.8857,
      "step": 5020
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.083676834295136e-06,
      "loss": 3.8952,
      "step": 5021
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.0774938169826874e-06,
      "loss": 4.0976,
      "step": 5022
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.071310799670239e-06,
      "loss": 3.9408,
      "step": 5023
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.0651277823577907e-06,
      "loss": 3.8453,
      "step": 5024
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.058944765045342e-06,
      "loss": 3.8891,
      "step": 5025
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.0527617477328935e-06,
      "loss": 4.0792,
      "step": 5026
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.046578730420445e-06,
      "loss": 3.7159,
      "step": 5027
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.0403957131079967e-06,
      "loss": 4.0082,
      "step": 5028
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.034212695795548e-06,
      "loss": 4.0482,
      "step": 5029
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.0280296784830995e-06,
      "loss": 3.8716,
      "step": 5030
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.0218466611706513e-06,
      "loss": 3.9387,
      "step": 5031
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.015663643858203e-06,
      "loss": 3.6971,
      "step": 5032
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.0094806265457546e-06,
      "loss": 3.706,
      "step": 5033
    },
    {
      "epoch": 2.82,
      "learning_rate": 2.003297609233306e-06,
      "loss": 3.4614,
      "step": 5034
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.997114591920858e-06,
      "loss": 3.614,
      "step": 5035
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.990931574608409e-06,
      "loss": 3.7498,
      "step": 5036
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9847485572959606e-06,
      "loss": 3.7194,
      "step": 5037
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.978565539983512e-06,
      "loss": 3.724,
      "step": 5038
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.972382522671064e-06,
      "loss": 3.8163,
      "step": 5039
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.9661995053586152e-06,
      "loss": 3.7071,
      "step": 5040
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.9600164880461666e-06,
      "loss": 3.6644,
      "step": 5041
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.953833470733718e-06,
      "loss": 3.8573,
      "step": 5042
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.94765045342127e-06,
      "loss": 3.9119,
      "step": 5043
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.9414674361088213e-06,
      "loss": 4.0105,
      "step": 5044
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.9352844187963727e-06,
      "loss": 3.7054,
      "step": 5045
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.929101401483924e-06,
      "loss": 3.9304,
      "step": 5046
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.922918384171476e-06,
      "loss": 3.8166,
      "step": 5047
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.9167353668590273e-06,
      "loss": 3.8046,
      "step": 5048
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.9105523495465787e-06,
      "loss": 3.6326,
      "step": 5049
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.9043693322341301e-06,
      "loss": 3.9054,
      "step": 5050
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.898186314921682e-06,
      "loss": 3.7617,
      "step": 5051
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.8920032976092334e-06,
      "loss": 3.7357,
      "step": 5052
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.8858202802967848e-06,
      "loss": 3.8461,
      "step": 5053
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.8796372629843362e-06,
      "loss": 3.5226,
      "step": 5054
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.873454245671888e-06,
      "loss": 3.7547,
      "step": 5055
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.8672712283594396e-06,
      "loss": 3.6716,
      "step": 5056
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.861088211046991e-06,
      "loss": 3.9163,
      "step": 5057
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.8549051937345426e-06,
      "loss": 3.6308,
      "step": 5058
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.848722176422094e-06,
      "loss": 3.7793,
      "step": 5059
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.8425391591096457e-06,
      "loss": 3.8278,
      "step": 5060
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.836356141797197e-06,
      "loss": 3.67,
      "step": 5061
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.8301731244847487e-06,
      "loss": 3.5739,
      "step": 5062
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.8239901071723e-06,
      "loss": 3.3377,
      "step": 5063
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.8178070898598517e-06,
      "loss": 3.2736,
      "step": 5064
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.8116240725474031e-06,
      "loss": 3.3019,
      "step": 5065
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.8054410552349547e-06,
      "loss": 3.1114,
      "step": 5066
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.7992580379225061e-06,
      "loss": 2.8948,
      "step": 5067
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.7930750206100577e-06,
      "loss": 3.0246,
      "step": 5068
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.7868920032976092e-06,
      "loss": 3.8401,
      "step": 5069
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.7807089859851608e-06,
      "loss": 4.0266,
      "step": 5070
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.7745259686727124e-06,
      "loss": 3.9041,
      "step": 5071
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.768342951360264e-06,
      "loss": 3.9128,
      "step": 5072
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.7621599340478154e-06,
      "loss": 3.8826,
      "step": 5073
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.755976916735367e-06,
      "loss": 3.8776,
      "step": 5074
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.7497938994229184e-06,
      "loss": 3.8399,
      "step": 5075
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.74361088211047e-06,
      "loss": 3.7863,
      "step": 5076
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.7374278647980214e-06,
      "loss": 3.9241,
      "step": 5077
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.731244847485573e-06,
      "loss": 3.9853,
      "step": 5078
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.7250618301731245e-06,
      "loss": 3.8134,
      "step": 5079
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.718878812860676e-06,
      "loss": 3.8345,
      "step": 5080
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.7126957955482275e-06,
      "loss": 3.9021,
      "step": 5081
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.706512778235779e-06,
      "loss": 3.5662,
      "step": 5082
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.7003297609233305e-06,
      "loss": 3.8167,
      "step": 5083
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6941467436108821e-06,
      "loss": 3.8092,
      "step": 5084
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6879637262984335e-06,
      "loss": 3.784,
      "step": 5085
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6817807089859854e-06,
      "loss": 3.7562,
      "step": 5086
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6755976916735368e-06,
      "loss": 3.6973,
      "step": 5087
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6694146743610884e-06,
      "loss": 3.8856,
      "step": 5088
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6632316570486398e-06,
      "loss": 3.5434,
      "step": 5089
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6570486397361914e-06,
      "loss": 3.7906,
      "step": 5090
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6508656224237428e-06,
      "loss": 3.545,
      "step": 5091
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6446826051112944e-06,
      "loss": 3.8865,
      "step": 5092
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.6384995877988458e-06,
      "loss": 3.6961,
      "step": 5093
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.6323165704863974e-06,
      "loss": 3.8476,
      "step": 5094
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.6261335531739488e-06,
      "loss": 3.764,
      "step": 5095
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.6199505358615005e-06,
      "loss": 3.6554,
      "step": 5096
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.6137675185490519e-06,
      "loss": 3.6585,
      "step": 5097
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.6075845012366035e-06,
      "loss": 3.7136,
      "step": 5098
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.6014014839241549e-06,
      "loss": 3.6401,
      "step": 5099
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5952184666117065e-06,
      "loss": 3.7146,
      "step": 5100
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.589035449299258e-06,
      "loss": 3.4933,
      "step": 5101
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5828524319868097e-06,
      "loss": 3.6848,
      "step": 5102
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5766694146743611e-06,
      "loss": 3.7351,
      "step": 5103
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5704863973619128e-06,
      "loss": 3.5432,
      "step": 5104
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5643033800494642e-06,
      "loss": 3.7333,
      "step": 5105
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5581203627370158e-06,
      "loss": 3.7319,
      "step": 5106
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5519373454245672e-06,
      "loss": 3.721,
      "step": 5107
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5457543281121188e-06,
      "loss": 3.5305,
      "step": 5108
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5395713107996702e-06,
      "loss": 3.4845,
      "step": 5109
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5333882934872218e-06,
      "loss": 3.3642,
      "step": 5110
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5272052761747732e-06,
      "loss": 3.5277,
      "step": 5111
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.5210222588623248e-06,
      "loss": 3.5048,
      "step": 5112
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.5148392415498762e-06,
      "loss": 3.2933,
      "step": 5113
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.5086562242374279e-06,
      "loss": 3.3201,
      "step": 5114
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.5024732069249793e-06,
      "loss": 3.4168,
      "step": 5115
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4962901896125309e-06,
      "loss": 3.183,
      "step": 5116
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4901071723000825e-06,
      "loss": 3.1068,
      "step": 5117
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4839241549876341e-06,
      "loss": 2.2874,
      "step": 5118
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4777411376751855e-06,
      "loss": 4.034,
      "step": 5119
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4715581203627371e-06,
      "loss": 4.0301,
      "step": 5120
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4653751030502885e-06,
      "loss": 3.8092,
      "step": 5121
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4591920857378402e-06,
      "loss": 3.961,
      "step": 5122
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4530090684253916e-06,
      "loss": 3.9141,
      "step": 5123
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4468260511129432e-06,
      "loss": 3.9095,
      "step": 5124
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4406430338004946e-06,
      "loss": 4.0469,
      "step": 5125
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4344600164880462e-06,
      "loss": 3.9742,
      "step": 5126
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4282769991755976e-06,
      "loss": 3.866,
      "step": 5127
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4220939818631492e-06,
      "loss": 3.8813,
      "step": 5128
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.4159109645507006e-06,
      "loss": 3.9091,
      "step": 5129
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.4097279472382522e-06,
      "loss": 3.8532,
      "step": 5130
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.4035449299258036e-06,
      "loss": 4.072,
      "step": 5131
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3973619126133553e-06,
      "loss": 3.7636,
      "step": 5132
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3911788953009069e-06,
      "loss": 3.8808,
      "step": 5133
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3849958779884585e-06,
      "loss": 3.7205,
      "step": 5134
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.37881286067601e-06,
      "loss": 3.9011,
      "step": 5135
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3726298433635615e-06,
      "loss": 3.7594,
      "step": 5136
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.366446826051113e-06,
      "loss": 3.7803,
      "step": 5137
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3602638087386645e-06,
      "loss": 3.6081,
      "step": 5138
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3540807914262161e-06,
      "loss": 3.6966,
      "step": 5139
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3478977741137676e-06,
      "loss": 3.6593,
      "step": 5140
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3417147568013192e-06,
      "loss": 3.9662,
      "step": 5141
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3355317394888706e-06,
      "loss": 4.0859,
      "step": 5142
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3293487221764222e-06,
      "loss": 4.1014,
      "step": 5143
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3231657048639736e-06,
      "loss": 3.7424,
      "step": 5144
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3169826875515252e-06,
      "loss": 3.8064,
      "step": 5145
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.3107996702390766e-06,
      "loss": 3.676,
      "step": 5146
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.3046166529266282e-06,
      "loss": 3.8026,
      "step": 5147
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2984336356141796e-06,
      "loss": 3.5681,
      "step": 5148
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2922506183017315e-06,
      "loss": 3.7417,
      "step": 5149
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2860676009892829e-06,
      "loss": 3.8418,
      "step": 5150
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2798845836768345e-06,
      "loss": 3.49,
      "step": 5151
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2737015663643859e-06,
      "loss": 3.7868,
      "step": 5152
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2675185490519375e-06,
      "loss": 4.1761,
      "step": 5153
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.261335531739489e-06,
      "loss": 3.7778,
      "step": 5154
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2551525144270405e-06,
      "loss": 3.5485,
      "step": 5155
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.248969497114592e-06,
      "loss": 3.4745,
      "step": 5156
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2427864798021435e-06,
      "loss": 3.6412,
      "step": 5157
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.236603462489695e-06,
      "loss": 4.0024,
      "step": 5158
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2304204451772466e-06,
      "loss": 3.6197,
      "step": 5159
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.224237427864798e-06,
      "loss": 3.8233,
      "step": 5160
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2180544105523496e-06,
      "loss": 3.7316,
      "step": 5161
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.211871393239901e-06,
      "loss": 3.8575,
      "step": 5162
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.2056883759274526e-06,
      "loss": 3.3915,
      "step": 5163
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.1995053586150042e-06,
      "loss": 3.3378,
      "step": 5164
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1933223413025558e-06,
      "loss": 3.399,
      "step": 5165
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1871393239901072e-06,
      "loss": 3.1239,
      "step": 5166
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1809563066776589e-06,
      "loss": 2.9467,
      "step": 5167
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1747732893652103e-06,
      "loss": 2.7099,
      "step": 5168
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1685902720527619e-06,
      "loss": 4.2205,
      "step": 5169
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1624072547403133e-06,
      "loss": 3.755,
      "step": 5170
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.156224237427865e-06,
      "loss": 3.8608,
      "step": 5171
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1500412201154163e-06,
      "loss": 4.0479,
      "step": 5172
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.143858202802968e-06,
      "loss": 3.6431,
      "step": 5173
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1376751854905193e-06,
      "loss": 3.9522,
      "step": 5174
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.131492168178071e-06,
      "loss": 3.754,
      "step": 5175
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1253091508656224e-06,
      "loss": 4.0635,
      "step": 5176
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.119126133553174e-06,
      "loss": 3.7784,
      "step": 5177
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1129431162407254e-06,
      "loss": 3.9004,
      "step": 5178
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.106760098928277e-06,
      "loss": 3.7213,
      "step": 5179
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1005770816158286e-06,
      "loss": 3.9298,
      "step": 5180
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.0943940643033802e-06,
      "loss": 3.8604,
      "step": 5181
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.0882110469909316e-06,
      "loss": 4.0188,
      "step": 5182
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0820280296784832e-06,
      "loss": 3.8444,
      "step": 5183
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0758450123660346e-06,
      "loss": 3.6503,
      "step": 5184
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0696619950535863e-06,
      "loss": 3.9279,
      "step": 5185
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0634789777411377e-06,
      "loss": 3.9353,
      "step": 5186
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0572959604286893e-06,
      "loss": 4.0058,
      "step": 5187
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0511129431162407e-06,
      "loss": 3.757,
      "step": 5188
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0449299258037923e-06,
      "loss": 3.6825,
      "step": 5189
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0387469084913437e-06,
      "loss": 3.6873,
      "step": 5190
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0325638911788953e-06,
      "loss": 3.7711,
      "step": 5191
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0263808738664467e-06,
      "loss": 3.6403,
      "step": 5192
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0201978565539983e-06,
      "loss": 3.6403,
      "step": 5193
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0140148392415498e-06,
      "loss": 3.894,
      "step": 5194
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.0078318219291016e-06,
      "loss": 3.8591,
      "step": 5195
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.001648804616653e-06,
      "loss": 3.7207,
      "step": 5196
    },
    {
      "epoch": 2.91,
      "learning_rate": 9.954657873042046e-07,
      "loss": 3.9057,
      "step": 5197
    },
    {
      "epoch": 2.91,
      "learning_rate": 9.89282769991756e-07,
      "loss": 3.5262,
      "step": 5198
    },
    {
      "epoch": 2.91,
      "learning_rate": 9.830997526793076e-07,
      "loss": 3.8396,
      "step": 5199
    },
    {
      "epoch": 2.91,
      "learning_rate": 9.76916735366859e-07,
      "loss": 3.9257,
      "step": 5200
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.707337180544106e-07,
      "loss": 3.96,
      "step": 5201
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.64550700741962e-07,
      "loss": 3.6903,
      "step": 5202
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.583676834295137e-07,
      "loss": 3.8853,
      "step": 5203
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.521846661170651e-07,
      "loss": 3.6986,
      "step": 5204
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.460016488046167e-07,
      "loss": 3.689,
      "step": 5205
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.398186314921681e-07,
      "loss": 3.6156,
      "step": 5206
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.336356141797198e-07,
      "loss": 3.6976,
      "step": 5207
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.274525968672713e-07,
      "loss": 3.7251,
      "step": 5208
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.212695795548228e-07,
      "loss": 3.8929,
      "step": 5209
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.150865622423743e-07,
      "loss": 3.5454,
      "step": 5210
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.089035449299259e-07,
      "loss": 3.4427,
      "step": 5211
    },
    {
      "epoch": 2.92,
      "learning_rate": 9.027205276174774e-07,
      "loss": 3.4656,
      "step": 5212
    },
    {
      "epoch": 2.92,
      "learning_rate": 8.965375103050289e-07,
      "loss": 3.1614,
      "step": 5213
    },
    {
      "epoch": 2.92,
      "learning_rate": 8.903544929925804e-07,
      "loss": 3.4306,
      "step": 5214
    },
    {
      "epoch": 2.92,
      "learning_rate": 8.84171475680132e-07,
      "loss": 3.6911,
      "step": 5215
    },
    {
      "epoch": 2.92,
      "learning_rate": 8.779884583676835e-07,
      "loss": 3.0489,
      "step": 5216
    },
    {
      "epoch": 2.92,
      "learning_rate": 8.71805441055235e-07,
      "loss": 3.0479,
      "step": 5217
    },
    {
      "epoch": 2.92,
      "learning_rate": 8.656224237427865e-07,
      "loss": 2.6066,
      "step": 5218
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.59439406430338e-07,
      "loss": 3.9508,
      "step": 5219
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.532563891178896e-07,
      "loss": 3.9067,
      "step": 5220
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.470733718054411e-07,
      "loss": 4.0079,
      "step": 5221
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.408903544929927e-07,
      "loss": 3.8598,
      "step": 5222
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.347073371805442e-07,
      "loss": 4.016,
      "step": 5223
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.285243198680957e-07,
      "loss": 3.8212,
      "step": 5224
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.223413025556472e-07,
      "loss": 4.1915,
      "step": 5225
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.161582852431987e-07,
      "loss": 3.7775,
      "step": 5226
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.099752679307502e-07,
      "loss": 4.0045,
      "step": 5227
    },
    {
      "epoch": 2.93,
      "learning_rate": 8.037922506183017e-07,
      "loss": 3.712,
      "step": 5228
    },
    {
      "epoch": 2.93,
      "learning_rate": 7.976092333058533e-07,
      "loss": 3.7169,
      "step": 5229
    },
    {
      "epoch": 2.93,
      "learning_rate": 7.914262159934049e-07,
      "loss": 3.8312,
      "step": 5230
    },
    {
      "epoch": 2.93,
      "learning_rate": 7.852431986809564e-07,
      "loss": 3.5479,
      "step": 5231
    },
    {
      "epoch": 2.93,
      "learning_rate": 7.790601813685079e-07,
      "loss": 3.9733,
      "step": 5232
    },
    {
      "epoch": 2.93,
      "learning_rate": 7.728771640560594e-07,
      "loss": 3.5249,
      "step": 5233
    },
    {
      "epoch": 2.93,
      "learning_rate": 7.666941467436109e-07,
      "loss": 3.6962,
      "step": 5234
    },
    {
      "epoch": 2.93,
      "learning_rate": 7.605111294311624e-07,
      "loss": 3.7486,
      "step": 5235
    },
    {
      "epoch": 2.93,
      "learning_rate": 7.543281121187139e-07,
      "loss": 4.0608,
      "step": 5236
    },
    {
      "epoch": 2.94,
      "learning_rate": 7.481450948062654e-07,
      "loss": 3.5296,
      "step": 5237
    },
    {
      "epoch": 2.94,
      "learning_rate": 7.419620774938171e-07,
      "loss": 3.9227,
      "step": 5238
    },
    {
      "epoch": 2.94,
      "learning_rate": 7.357790601813686e-07,
      "loss": 3.5781,
      "step": 5239
    },
    {
      "epoch": 2.94,
      "learning_rate": 7.295960428689201e-07,
      "loss": 3.6811,
      "step": 5240
    },
    {
      "epoch": 2.94,
      "learning_rate": 7.234130255564716e-07,
      "loss": 3.964,
      "step": 5241
    },
    {
      "epoch": 2.94,
      "learning_rate": 7.172300082440231e-07,
      "loss": 3.6384,
      "step": 5242
    },
    {
      "epoch": 2.94,
      "learning_rate": 7.110469909315746e-07,
      "loss": 3.855,
      "step": 5243
    },
    {
      "epoch": 2.94,
      "learning_rate": 7.048639736191261e-07,
      "loss": 3.7323,
      "step": 5244
    },
    {
      "epoch": 2.94,
      "learning_rate": 6.986809563066776e-07,
      "loss": 3.5715,
      "step": 5245
    },
    {
      "epoch": 2.94,
      "learning_rate": 6.924979389942292e-07,
      "loss": 3.6127,
      "step": 5246
    },
    {
      "epoch": 2.94,
      "learning_rate": 6.863149216817808e-07,
      "loss": 3.9287,
      "step": 5247
    },
    {
      "epoch": 2.94,
      "learning_rate": 6.801319043693323e-07,
      "loss": 3.8215,
      "step": 5248
    },
    {
      "epoch": 2.94,
      "learning_rate": 6.739488870568838e-07,
      "loss": 3.7167,
      "step": 5249
    },
    {
      "epoch": 2.94,
      "learning_rate": 6.677658697444353e-07,
      "loss": 3.6421,
      "step": 5250
    },
    {
      "epoch": 2.94,
      "learning_rate": 6.615828524319868e-07,
      "loss": 3.3736,
      "step": 5251
    },
    {
      "epoch": 2.94,
      "learning_rate": 6.553998351195383e-07,
      "loss": 3.6181,
      "step": 5252
    },
    {
      "epoch": 2.94,
      "learning_rate": 6.492168178070898e-07,
      "loss": 3.8497,
      "step": 5253
    },
    {
      "epoch": 2.95,
      "learning_rate": 6.430338004946414e-07,
      "loss": 3.721,
      "step": 5254
    },
    {
      "epoch": 2.95,
      "learning_rate": 6.368507831821929e-07,
      "loss": 3.663,
      "step": 5255
    },
    {
      "epoch": 2.95,
      "learning_rate": 6.306677658697445e-07,
      "loss": 3.7621,
      "step": 5256
    },
    {
      "epoch": 2.95,
      "learning_rate": 6.24484748557296e-07,
      "loss": 3.8191,
      "step": 5257
    },
    {
      "epoch": 2.95,
      "learning_rate": 6.183017312448475e-07,
      "loss": 3.5469,
      "step": 5258
    },
    {
      "epoch": 2.95,
      "learning_rate": 6.12118713932399e-07,
      "loss": 3.5333,
      "step": 5259
    },
    {
      "epoch": 2.95,
      "learning_rate": 6.059356966199505e-07,
      "loss": 3.6778,
      "step": 5260
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.997526793075021e-07,
      "loss": 3.3322,
      "step": 5261
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.935696619950536e-07,
      "loss": 3.561,
      "step": 5262
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.873866446826051e-07,
      "loss": 3.5625,
      "step": 5263
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.812036273701566e-07,
      "loss": 3.3405,
      "step": 5264
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.750206100577082e-07,
      "loss": 2.9009,
      "step": 5265
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.688375927452597e-07,
      "loss": 3.1274,
      "step": 5266
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.626545754328112e-07,
      "loss": 2.5694,
      "step": 5267
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.564715581203627e-07,
      "loss": 2.6443,
      "step": 5268
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.502885408079143e-07,
      "loss": 4.1818,
      "step": 5269
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.441055234954658e-07,
      "loss": 3.8775,
      "step": 5270
    },
    {
      "epoch": 2.95,
      "learning_rate": 5.379225061830173e-07,
      "loss": 3.9808,
      "step": 5271
    },
    {
      "epoch": 2.96,
      "learning_rate": 5.317394888705688e-07,
      "loss": 4.0399,
      "step": 5272
    },
    {
      "epoch": 2.96,
      "learning_rate": 5.255564715581203e-07,
      "loss": 3.7086,
      "step": 5273
    },
    {
      "epoch": 2.96,
      "learning_rate": 5.193734542456719e-07,
      "loss": 3.9558,
      "step": 5274
    },
    {
      "epoch": 2.96,
      "learning_rate": 5.131904369332234e-07,
      "loss": 3.7198,
      "step": 5275
    },
    {
      "epoch": 2.96,
      "learning_rate": 5.070074196207749e-07,
      "loss": 3.7034,
      "step": 5276
    },
    {
      "epoch": 2.96,
      "learning_rate": 5.008244023083265e-07,
      "loss": 3.9827,
      "step": 5277
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.94641384995878e-07,
      "loss": 3.9607,
      "step": 5278
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.884583676834295e-07,
      "loss": 3.5996,
      "step": 5279
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.82275350370981e-07,
      "loss": 3.5657,
      "step": 5280
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.7609233305853253e-07,
      "loss": 3.6678,
      "step": 5281
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.6990931574608404e-07,
      "loss": 3.7803,
      "step": 5282
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.6372629843363566e-07,
      "loss": 3.8721,
      "step": 5283
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.5754328112118717e-07,
      "loss": 3.92,
      "step": 5284
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.513602638087387e-07,
      "loss": 3.7026,
      "step": 5285
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.451772464962902e-07,
      "loss": 3.699,
      "step": 5286
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.3899422918384175e-07,
      "loss": 3.7904,
      "step": 5287
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.3281121187139327e-07,
      "loss": 3.6828,
      "step": 5288
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.266281945589448e-07,
      "loss": 3.6557,
      "step": 5289
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.2044517724649634e-07,
      "loss": 3.8247,
      "step": 5290
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.1426215993404785e-07,
      "loss": 3.8212,
      "step": 5291
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.0807914262159936e-07,
      "loss": 3.7603,
      "step": 5292
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.0189612530915087e-07,
      "loss": 3.6663,
      "step": 5293
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.9571310799670243e-07,
      "loss": 3.8549,
      "step": 5294
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.8953009068425394e-07,
      "loss": 3.9585,
      "step": 5295
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.8334707337180545e-07,
      "loss": 3.8046,
      "step": 5296
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.7716405605935696e-07,
      "loss": 3.5638,
      "step": 5297
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.7098103874690853e-07,
      "loss": 3.9879,
      "step": 5298
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.6479802143446004e-07,
      "loss": 3.7505,
      "step": 5299
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.5861500412201155e-07,
      "loss": 3.788,
      "step": 5300
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.5243198680956306e-07,
      "loss": 4.1454,
      "step": 5301
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.462489694971146e-07,
      "loss": 3.825,
      "step": 5302
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.4006595218466613e-07,
      "loss": 3.8784,
      "step": 5303
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.3388293487221764e-07,
      "loss": 3.9132,
      "step": 5304
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.2769991755976915e-07,
      "loss": 3.3373,
      "step": 5305
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.215169002473207e-07,
      "loss": 3.5046,
      "step": 5306
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.1533388293487223e-07,
      "loss": 3.5957,
      "step": 5307
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.0915086562242374e-07,
      "loss": 3.5655,
      "step": 5308
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.0296784830997525e-07,
      "loss": 3.6355,
      "step": 5309
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.967848309975268e-07,
      "loss": 3.8775,
      "step": 5310
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.906018136850783e-07,
      "loss": 3.5886,
      "step": 5311
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.8441879637262983e-07,
      "loss": 3.6388,
      "step": 5312
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.7823577906018134e-07,
      "loss": 3.8451,
      "step": 5313
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.720527617477329e-07,
      "loss": 3.3668,
      "step": 5314
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.658697444352844e-07,
      "loss": 3.2016,
      "step": 5315
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.5968672712283593e-07,
      "loss": 2.8242,
      "step": 5316
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.5350370981038744e-07,
      "loss": 2.8942,
      "step": 5317
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.47320692497939e-07,
      "loss": 2.5908,
      "step": 5318
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.411376751854905e-07,
      "loss": 3.9936,
      "step": 5319
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.3495465787304202e-07,
      "loss": 3.9456,
      "step": 5320
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.2877164056059359e-07,
      "loss": 3.8474,
      "step": 5321
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.225886232481451e-07,
      "loss": 3.9123,
      "step": 5322
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.1640560593569663e-07,
      "loss": 3.9043,
      "step": 5323
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.1022258862324817e-07,
      "loss": 3.7966,
      "step": 5324
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.0403957131079968e-07,
      "loss": 3.9104,
      "step": 5325
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.9785655399835122e-07,
      "loss": 3.6912,
      "step": 5326
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.9167353668590273e-07,
      "loss": 3.7893,
      "step": 5327
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.8549051937345426e-07,
      "loss": 3.8957,
      "step": 5328
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.7930750206100577e-07,
      "loss": 4.0351,
      "step": 5329
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.731244847485573e-07,
      "loss": 3.8351,
      "step": 5330
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.6694146743610882e-07,
      "loss": 3.9905,
      "step": 5331
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.6075845012366036e-07,
      "loss": 3.9011,
      "step": 5332
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.5457543281121187e-07,
      "loss": 3.8105,
      "step": 5333
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.483924154987634e-07,
      "loss": 3.5278,
      "step": 5334
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.4220939818631492e-07,
      "loss": 3.768,
      "step": 5335
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.3602638087386645e-07,
      "loss": 3.5879,
      "step": 5336
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.2984336356141796e-07,
      "loss": 3.8286,
      "step": 5337
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.236603462489695e-07,
      "loss": 3.6917,
      "step": 5338
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.1747732893652101e-07,
      "loss": 3.901,
      "step": 5339
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.1129431162407255e-07,
      "loss": 3.7741,
      "step": 5340
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.0511129431162408e-07,
      "loss": 3.6495,
      "step": 5341
    },
    {
      "epoch": 2.99,
      "learning_rate": 9.892827699917561e-08,
      "loss": 3.5161,
      "step": 5342
    },
    {
      "epoch": 2.99,
      "learning_rate": 9.274525968672713e-08,
      "loss": 3.5559,
      "step": 5343
    },
    {
      "epoch": 3.0,
      "learning_rate": 8.656224237427866e-08,
      "loss": 3.6556,
      "step": 5344
    },
    {
      "epoch": 3.0,
      "learning_rate": 8.037922506183018e-08,
      "loss": 3.6051,
      "step": 5345
    },
    {
      "epoch": 3.0,
      "learning_rate": 7.41962077493817e-08,
      "loss": 3.6621,
      "step": 5346
    },
    {
      "epoch": 3.0,
      "learning_rate": 6.801319043693323e-08,
      "loss": 3.5508,
      "step": 5347
    },
    {
      "epoch": 3.0,
      "learning_rate": 6.183017312448475e-08,
      "loss": 3.6915,
      "step": 5348
    },
    {
      "epoch": 3.0,
      "learning_rate": 5.5647155812036274e-08,
      "loss": 3.3452,
      "step": 5349
    },
    {
      "epoch": 3.0,
      "learning_rate": 4.9464138499587804e-08,
      "loss": 2.9456,
      "step": 5350
    },
    {
      "epoch": 3.0,
      "learning_rate": 4.328112118713933e-08,
      "loss": 3.2073,
      "step": 5351
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.709810387469085e-08,
      "loss": 2.6908,
      "step": 5352
    },
    {
      "epoch": 3.0,
      "step": 5352,
      "total_flos": 0.0,
      "train_loss": 4.476328243990116,
      "train_runtime": 28742.2766,
      "train_samples_per_second": 2.979,
      "train_steps_per_second": 0.186
    }
  ],
  "max_steps": 5352,
  "num_train_epochs": 3,
  "total_flos": 0.0,
  "trial_name": null,
  "trial_params": null
}