{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 5293,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.2578616352201258e-07,
      "loss": 2.9003,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.5157232704402517e-07,
      "loss": 2.8605,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.773584905660378e-07,
      "loss": 2.8961,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.031446540880503e-07,
      "loss": 3.0197,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.28930817610063e-07,
      "loss": 3.0022,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.547169811320755e-07,
      "loss": 3.3174,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.805031446540881e-07,
      "loss": 2.6333,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.0062893081761007e-06,
      "loss": 2.8719,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.1320754716981133e-06,
      "loss": 2.6209,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.257861635220126e-06,
      "loss": 2.3449,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.3836477987421384e-06,
      "loss": 2.2187,
      "step": 11
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.509433962264151e-06,
      "loss": 2.199,
      "step": 12
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.6352201257861635e-06,
      "loss": 1.3869,
      "step": 13
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.7610062893081762e-06,
      "loss": 1.4613,
      "step": 14
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.8867924528301889e-06,
      "loss": 1.3137,
      "step": 15
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.0125786163522013e-06,
      "loss": 1.3615,
      "step": 16
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.138364779874214e-06,
      "loss": 1.2526,
      "step": 17
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.2641509433962266e-06,
      "loss": 1.1311,
      "step": 18
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.389937106918239e-06,
      "loss": 0.8397,
      "step": 19
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.515723270440252e-06,
      "loss": 0.9525,
      "step": 20
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.6415094339622644e-06,
      "loss": 0.9427,
      "step": 21
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.767295597484277e-06,
      "loss": 0.9563,
      "step": 22
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.8930817610062893e-06,
      "loss": 0.9302,
      "step": 23
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.018867924528302e-06,
      "loss": 0.915,
      "step": 24
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.1446540880503146e-06,
      "loss": 0.8127,
      "step": 25
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.270440251572327e-06,
      "loss": 0.7934,
      "step": 26
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.3962264150943395e-06,
      "loss": 0.7398,
      "step": 27
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.5220125786163524e-06,
      "loss": 0.9085,
      "step": 28
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.647798742138365e-06,
      "loss": 0.8116,
      "step": 29
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.7735849056603777e-06,
      "loss": 0.7594,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.89937106918239e-06,
      "loss": 0.7183,
      "step": 31
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.025157232704403e-06,
      "loss": 0.7135,
      "step": 32
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.150943396226416e-06,
      "loss": 0.5953,
      "step": 33
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.276729559748428e-06,
      "loss": 0.7524,
      "step": 34
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.402515723270441e-06,
      "loss": 0.7809,
      "step": 35
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.528301886792453e-06,
      "loss": 0.6991,
      "step": 36
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.654088050314466e-06,
      "loss": 0.775,
      "step": 37
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.779874213836478e-06,
      "loss": 0.7897,
      "step": 38
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.905660377358491e-06,
      "loss": 0.7646,
      "step": 39
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.031446540880504e-06,
      "loss": 0.6718,
      "step": 40
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.157232704402516e-06,
      "loss": 0.8805,
      "step": 41
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.283018867924529e-06,
      "loss": 0.6674,
      "step": 42
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.408805031446541e-06,
      "loss": 0.7801,
      "step": 43
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.534591194968554e-06,
      "loss": 0.7097,
      "step": 44
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.660377358490566e-06,
      "loss": 0.7525,
      "step": 45
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.786163522012579e-06,
      "loss": 0.6345,
      "step": 46
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.911949685534591e-06,
      "loss": 0.6604,
      "step": 47
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.037735849056604e-06,
      "loss": 0.8861,
      "step": 48
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.163522012578617e-06,
      "loss": 0.7344,
      "step": 49
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.289308176100629e-06,
      "loss": 0.7242,
      "step": 50
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.415094339622642e-06,
      "loss": 0.6609,
      "step": 51
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.540880503144654e-06,
      "loss": 0.6181,
      "step": 52
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.666666666666667e-06,
      "loss": 0.5307,
      "step": 53
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.792452830188679e-06,
      "loss": 0.6782,
      "step": 54
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.9182389937106915e-06,
      "loss": 0.6726,
      "step": 55
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.044025157232705e-06,
      "loss": 0.7186,
      "step": 56
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.169811320754717e-06,
      "loss": 0.7728,
      "step": 57
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.29559748427673e-06,
      "loss": 0.6714,
      "step": 58
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.421383647798742e-06,
      "loss": 0.59,
      "step": 59
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.5471698113207555e-06,
      "loss": 0.6865,
      "step": 60
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.672955974842768e-06,
      "loss": 0.726,
      "step": 61
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.79874213836478e-06,
      "loss": 0.541,
      "step": 62
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.924528301886793e-06,
      "loss": 0.614,
      "step": 63
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.050314465408805e-06,
      "loss": 0.7961,
      "step": 64
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.17610062893082e-06,
      "loss": 0.6302,
      "step": 65
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.301886792452832e-06,
      "loss": 0.6586,
      "step": 66
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.427672955974844e-06,
      "loss": 0.6428,
      "step": 67
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.553459119496857e-06,
      "loss": 0.5777,
      "step": 68
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.67924528301887e-06,
      "loss": 0.8179,
      "step": 69
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.805031446540882e-06,
      "loss": 0.5903,
      "step": 70
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.930817610062894e-06,
      "loss": 0.5953,
      "step": 71
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.056603773584907e-06,
      "loss": 0.6665,
      "step": 72
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.182389937106919e-06,
      "loss": 0.7219,
      "step": 73
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.308176100628931e-06,
      "loss": 0.6492,
      "step": 74
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.433962264150944e-06,
      "loss": 0.6834,
      "step": 75
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.559748427672956e-06,
      "loss": 0.5328,
      "step": 76
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.685534591194969e-06,
      "loss": 0.6929,
      "step": 77
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.811320754716981e-06,
      "loss": 0.578,
      "step": 78
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.937106918238994e-06,
      "loss": 0.5392,
      "step": 79
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0062893081761008e-05,
      "loss": 0.6924,
      "step": 80
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.018867924528302e-05,
      "loss": 0.6315,
      "step": 81
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0314465408805033e-05,
      "loss": 0.5293,
      "step": 82
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0440251572327045e-05,
      "loss": 0.6656,
      "step": 83
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0566037735849058e-05,
      "loss": 0.813,
      "step": 84
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.069182389937107e-05,
      "loss": 0.7101,
      "step": 85
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0817610062893083e-05,
      "loss": 0.6531,
      "step": 86
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0943396226415095e-05,
      "loss": 0.5837,
      "step": 87
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1069182389937107e-05,
      "loss": 0.6335,
      "step": 88
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.119496855345912e-05,
      "loss": 0.6192,
      "step": 89
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1320754716981132e-05,
      "loss": 0.5517,
      "step": 90
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1446540880503145e-05,
      "loss": 0.6116,
      "step": 91
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1572327044025157e-05,
      "loss": 0.5987,
      "step": 92
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.169811320754717e-05,
      "loss": 0.6827,
      "step": 93
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1823899371069182e-05,
      "loss": 0.5083,
      "step": 94
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1949685534591196e-05,
      "loss": 0.6555,
      "step": 95
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2075471698113209e-05,
      "loss": 0.614,
      "step": 96
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2201257861635221e-05,
      "loss": 0.7708,
      "step": 97
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2327044025157234e-05,
      "loss": 0.7318,
      "step": 98
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2452830188679246e-05,
      "loss": 0.623,
      "step": 99
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2578616352201259e-05,
      "loss": 0.6502,
      "step": 100
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2704402515723271e-05,
      "loss": 0.5998,
      "step": 101
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2830188679245283e-05,
      "loss": 0.5604,
      "step": 102
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2955974842767296e-05,
      "loss": 0.7929,
      "step": 103
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3081761006289308e-05,
      "loss": 0.7071,
      "step": 104
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.320754716981132e-05,
      "loss": 0.7683,
      "step": 105
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3333333333333333e-05,
      "loss": 0.6646,
      "step": 106
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3459119496855346e-05,
      "loss": 0.6703,
      "step": 107
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3584905660377358e-05,
      "loss": 0.5489,
      "step": 108
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.371069182389937e-05,
      "loss": 0.8126,
      "step": 109
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3836477987421383e-05,
      "loss": 0.8022,
      "step": 110
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.3962264150943397e-05,
      "loss": 0.7625,
      "step": 111
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.408805031446541e-05,
      "loss": 0.8567,
      "step": 112
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4213836477987422e-05,
      "loss": 0.5958,
      "step": 113
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4339622641509435e-05,
      "loss": 0.483,
      "step": 114
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4465408805031447e-05,
      "loss": 0.6908,
      "step": 115
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.459119496855346e-05,
      "loss": 0.7446,
      "step": 116
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4716981132075472e-05,
      "loss": 0.8195,
      "step": 117
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4842767295597484e-05,
      "loss": 0.7398,
      "step": 118
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4968553459119497e-05,
      "loss": 0.8607,
      "step": 119
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5094339622641511e-05,
      "loss": 0.6992,
      "step": 120
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5220125786163525e-05,
      "loss": 0.6016,
      "step": 121
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5345911949685536e-05,
      "loss": 0.6876,
      "step": 122
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.547169811320755e-05,
      "loss": 0.8756,
      "step": 123
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.559748427672956e-05,
      "loss": 0.6193,
      "step": 124
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5723270440251575e-05,
      "loss": 0.6255,
      "step": 125
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5849056603773586e-05,
      "loss": 0.5405,
      "step": 126
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.59748427672956e-05,
      "loss": 0.8002,
      "step": 127
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.610062893081761e-05,
      "loss": 0.5789,
      "step": 128
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6226415094339625e-05,
      "loss": 0.7674,
      "step": 129
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.635220125786164e-05,
      "loss": 0.7165,
      "step": 130
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.647798742138365e-05,
      "loss": 0.6537,
      "step": 131
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6603773584905664e-05,
      "loss": 0.6347,
      "step": 132
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.6729559748427675e-05,
      "loss": 0.6772,
      "step": 133
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.685534591194969e-05,
      "loss": 0.7095,
      "step": 134
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.69811320754717e-05,
      "loss": 0.6785,
      "step": 135
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7106918238993714e-05,
      "loss": 0.6601,
      "step": 136
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7232704402515724e-05,
      "loss": 0.7688,
      "step": 137
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.735849056603774e-05,
      "loss": 0.6459,
      "step": 138
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.748427672955975e-05,
      "loss": 0.5228,
      "step": 139
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7610062893081763e-05,
      "loss": 0.5642,
      "step": 140
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7735849056603774e-05,
      "loss": 0.8328,
      "step": 141
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7861635220125788e-05,
      "loss": 0.5998,
      "step": 142
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.79874213836478e-05,
      "loss": 0.8785,
      "step": 143
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8113207547169813e-05,
      "loss": 0.6707,
      "step": 144
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8238993710691827e-05,
      "loss": 0.6566,
      "step": 145
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8364779874213838e-05,
      "loss": 0.6448,
      "step": 146
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8490566037735852e-05,
      "loss": 0.5819,
      "step": 147
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8616352201257863e-05,
      "loss": 0.6793,
      "step": 148
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8742138364779877e-05,
      "loss": 0.7071,
      "step": 149
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8867924528301888e-05,
      "loss": 0.7584,
      "step": 150
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.8993710691823902e-05,
      "loss": 0.8256,
      "step": 151
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9119496855345913e-05,
      "loss": 0.6851,
      "step": 152
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9245283018867927e-05,
      "loss": 0.667,
      "step": 153
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9371069182389938e-05,
      "loss": 0.6073,
      "step": 154
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9496855345911952e-05,
      "loss": 0.7333,
      "step": 155
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9622641509433963e-05,
      "loss": 0.6845,
      "step": 156
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9748427672955977e-05,
      "loss": 0.5911,
      "step": 157
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9874213836477987e-05,
      "loss": 0.6814,
      "step": 158
    },
    {
      "epoch": 0.03,
      "learning_rate": 2e-05,
      "loss": 0.5802,
      "step": 159
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999998127775143e-05,
      "loss": 0.5972,
      "step": 160
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999992511101266e-05,
      "loss": 0.7765,
      "step": 161
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999983149980473e-05,
      "loss": 0.6315,
      "step": 162
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999970044416273e-05,
      "loss": 0.5626,
      "step": 163
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999995319441357e-05,
      "loss": 0.7166,
      "step": 164
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999932599978674e-05,
      "loss": 0.6474,
      "step": 165
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999908261119295e-05,
      "loss": 0.7463,
      "step": 166
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999880177844552e-05,
      "loss": 0.7002,
      "step": 167
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999848350164958e-05,
      "loss": 0.7019,
      "step": 168
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999812778092425e-05,
      "loss": 0.5973,
      "step": 169
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999773461640282e-05,
      "loss": 0.6132,
      "step": 170
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999730400823244e-05,
      "loss": 0.7176,
      "step": 171
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999683595657436e-05,
      "loss": 0.6751,
      "step": 172
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999633046160386e-05,
      "loss": 0.7415,
      "step": 173
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999578752351022e-05,
      "loss": 0.6113,
      "step": 174
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999520714249672e-05,
      "loss": 0.66,
      "step": 175
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999945893187807e-05,
      "loss": 0.5901,
      "step": 176
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999393405259354e-05,
      "loss": 0.6363,
      "step": 177
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999324134418053e-05,
      "loss": 0.6355,
      "step": 178
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999251119380105e-05,
      "loss": 0.6915,
      "step": 179
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9999174360172855e-05,
      "loss": 0.6234,
      "step": 180
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999909385682504e-05,
      "loss": 0.6215,
      "step": 181
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999900960936681e-05,
      "loss": 0.8868,
      "step": 182
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9998921617829708e-05,
      "loss": 0.6218,
      "step": 183
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9998829882246684e-05,
      "loss": 0.5871,
      "step": 184
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.999873440265208e-05,
      "loss": 0.7843,
      "step": 185
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9998635179081658e-05,
      "loss": 0.7821,
      "step": 186
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9998532211572566e-05,
      "loss": 0.5964,
      "step": 187
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9998425500163363e-05,
      "loss": 0.5649,
      "step": 188
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9998315044894006e-05,
      "loss": 0.6584,
      "step": 189
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999820084580585e-05,
      "loss": 0.8553,
      "step": 190
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999808290294166e-05,
      "loss": 0.6535,
      "step": 191
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.99979612163456e-05,
      "loss": 0.6756,
      "step": 192
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999783578606323e-05,
      "loss": 0.7919,
      "step": 193
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9997706612141522e-05,
      "loss": 0.7333,
      "step": 194
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9997573694628846e-05,
      "loss": 0.5511,
      "step": 195
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9997437033574967e-05,
      "loss": 0.6066,
      "step": 196
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999729662903106e-05,
      "loss": 0.7922,
      "step": 197
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9997152481049697e-05,
      "loss": 0.6688,
      "step": 198
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9997004589684852e-05,
      "loss": 0.7331,
      "step": 199
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9996852954991905e-05,
      "loss": 0.6669,
      "step": 200
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9996697577027636e-05,
      "loss": 0.6727,
      "step": 201
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9996538455850226e-05,
      "loss": 0.7987,
      "step": 202
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9996375591519254e-05,
      "loss": 0.7364,
      "step": 203
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.99962089840957e-05,
      "loss": 0.641,
      "step": 204
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999603863364196e-05,
      "loss": 0.5363,
      "step": 205
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9995864540221816e-05,
      "loss": 0.7245,
      "step": 206
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999568670390045e-05,
      "loss": 0.7697,
      "step": 207
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999550512474446e-05,
      "loss": 0.6599,
      "step": 208
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9995319802821832e-05,
      "loss": 0.6034,
      "step": 209
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9995130738201966e-05,
      "loss": 0.6008,
      "step": 210
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9994937930955652e-05,
      "loss": 0.6824,
      "step": 211
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9994741381155084e-05,
      "loss": 0.6103,
      "step": 212
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9994541088873856e-05,
      "loss": 0.631,
      "step": 213
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9994337054186975e-05,
      "loss": 0.6383,
      "step": 214
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999412927717084e-05,
      "loss": 0.51,
      "step": 215
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9993917757903243e-05,
      "loss": 0.6301,
      "step": 216
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9993702496463395e-05,
      "loss": 0.643,
      "step": 217
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9993483492931895e-05,
      "loss": 0.8141,
      "step": 218
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9993260747390754e-05,
      "loss": 0.7312,
      "step": 219
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999303425992337e-05,
      "loss": 0.5865,
      "step": 220
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9992804030614553e-05,
      "loss": 0.5673,
      "step": 221
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9992570059550513e-05,
      "loss": 0.6901,
      "step": 222
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999233234681886e-05,
      "loss": 0.811,
      "step": 223
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.99920908925086e-05,
      "loss": 0.5411,
      "step": 224
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9991845696710146e-05,
      "loss": 0.5754,
      "step": 225
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9991596759515314e-05,
      "loss": 0.5265,
      "step": 226
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9991344081017312e-05,
      "loss": 0.7087,
      "step": 227
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999108766131076e-05,
      "loss": 0.7973,
      "step": 228
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9990827500491666e-05,
      "loss": 0.5511,
      "step": 229
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.999056359865745e-05,
      "loss": 0.6778,
      "step": 230
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9990295955906927e-05,
      "loss": 0.5695,
      "step": 231
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9990024572340317e-05,
      "loss": 0.619,
      "step": 232
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.998974944805924e-05,
      "loss": 0.6746,
      "step": 233
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.998947058316671e-05,
      "loss": 0.6085,
      "step": 234
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.998918797776715e-05,
      "loss": 0.732,
      "step": 235
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9988901631966373e-05,
      "loss": 0.6765,
      "step": 236
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9988611545871606e-05,
      "loss": 0.6403,
      "step": 237
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9988317719591473e-05,
      "loss": 0.5689,
      "step": 238
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9988020153235996e-05,
      "loss": 0.7555,
      "step": 239
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998771884691659e-05,
      "loss": 0.557,
      "step": 240
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998741380074608e-05,
      "loss": 0.702,
      "step": 241
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998710501483869e-05,
      "loss": 0.6374,
      "step": 242
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9986792489310044e-05,
      "loss": 0.6665,
      "step": 243
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9986476224277167e-05,
      "loss": 0.6836,
      "step": 244
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998615621985848e-05,
      "loss": 0.706,
      "step": 245
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9985832476173807e-05,
      "loss": 0.5046,
      "step": 246
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9985504993344375e-05,
      "loss": 0.6341,
      "step": 247
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9985173771492807e-05,
      "loss": 0.6224,
      "step": 248
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9984838810743124e-05,
      "loss": 0.6806,
      "step": 249
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9984500111220757e-05,
      "loss": 0.5569,
      "step": 250
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9984157673052525e-05,
      "loss": 0.7444,
      "step": 251
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9983811496366654e-05,
      "loss": 0.8125,
      "step": 252
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9983461581292768e-05,
      "loss": 0.6381,
      "step": 253
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9983107927961888e-05,
      "loss": 0.6368,
      "step": 254
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9982750536506445e-05,
      "loss": 0.7953,
      "step": 255
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998238940706026e-05,
      "loss": 0.6563,
      "step": 256
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9982024539758547e-05,
      "loss": 0.7502,
      "step": 257
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998165593473794e-05,
      "loss": 0.8238,
      "step": 258
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998128359213646e-05,
      "loss": 0.6128,
      "step": 259
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998090751209352e-05,
      "loss": 0.6802,
      "step": 260
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9980527694749952e-05,
      "loss": 0.8623,
      "step": 261
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.998014414024797e-05,
      "loss": 0.679,
      "step": 262
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9979756848731195e-05,
      "loss": 0.7058,
      "step": 263
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9979365820344648e-05,
      "loss": 0.6038,
      "step": 264
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997897105523475e-05,
      "loss": 0.6477,
      "step": 265
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9978572553549315e-05,
      "loss": 0.5313,
      "step": 266
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997817031543756e-05,
      "loss": 0.5533,
      "step": 267
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.99777643410501e-05,
      "loss": 0.6897,
      "step": 268
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9977354630538953e-05,
      "loss": 0.5639,
      "step": 269
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9976941184057536e-05,
      "loss": 0.4884,
      "step": 270
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9976524001760653e-05,
      "loss": 0.6924,
      "step": 271
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9976103083804523e-05,
      "loss": 0.6267,
      "step": 272
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9975678430346753e-05,
      "loss": 0.6701,
      "step": 273
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9975250041546354e-05,
      "loss": 0.5491,
      "step": 274
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9974817917563735e-05,
      "loss": 0.6715,
      "step": 275
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.99743820585607e-05,
      "loss": 0.6484,
      "step": 276
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9973942464700456e-05,
      "loss": 0.7303,
      "step": 277
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997349913614761e-05,
      "loss": 0.5503,
      "step": 278
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9973052073068154e-05,
      "loss": 0.8026,
      "step": 279
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9972601275629497e-05,
      "loss": 0.675,
      "step": 280
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9972146744000436e-05,
      "loss": 0.6253,
      "step": 281
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997168847835117e-05,
      "loss": 0.6511,
      "step": 282
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.997122647885329e-05,
      "loss": 0.5213,
      "step": 283
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9970760745679793e-05,
      "loss": 0.6078,
      "step": 284
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9970291279005066e-05,
      "loss": 0.764,
      "step": 285
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9969818079004907e-05,
      "loss": 0.5073,
      "step": 286
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9969341145856493e-05,
      "loss": 0.6617,
      "step": 287
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9968860479738416e-05,
      "loss": 0.5661,
      "step": 288
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9968376080830654e-05,
      "loss": 0.6483,
      "step": 289
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9967887949314595e-05,
      "loss": 0.6387,
      "step": 290
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.996739608537301e-05,
      "loss": 0.5802,
      "step": 291
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9966900489190082e-05,
      "loss": 0.7015,
      "step": 292
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996640116095138e-05,
      "loss": 0.7117,
      "step": 293
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996589810084387e-05,
      "loss": 0.6152,
      "step": 294
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996539130905593e-05,
      "loss": 0.7553,
      "step": 295
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996488078577732e-05,
      "loss": 0.5157,
      "step": 296
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9964366531199205e-05,
      "loss": 0.6229,
      "step": 297
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996384854551415e-05,
      "loss": 0.7195,
      "step": 298
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.99633268289161e-05,
      "loss": 0.6942,
      "step": 299
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996280138160042e-05,
      "loss": 0.6831,
      "step": 300
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9962272203763852e-05,
      "loss": 0.607,
      "step": 301
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9961739295604553e-05,
      "loss": 0.6172,
      "step": 302
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996120265732206e-05,
      "loss": 0.5852,
      "step": 303
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996066228911732e-05,
      "loss": 0.6676,
      "step": 304
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.996011819119267e-05,
      "loss": 0.7202,
      "step": 305
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995957036375184e-05,
      "loss": 0.6109,
      "step": 306
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995901880699997e-05,
      "loss": 0.619,
      "step": 307
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9958463521143584e-05,
      "loss": 0.8135,
      "step": 308
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9957904506390602e-05,
      "loss": 0.5947,
      "step": 309
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9957341762950346e-05,
      "loss": 0.5531,
      "step": 310
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9956775291033534e-05,
      "loss": 0.7493,
      "step": 311
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995620509085228e-05,
      "loss": 0.6118,
      "step": 312
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9955631162620094e-05,
      "loss": 0.835,
      "step": 313
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9955053506551872e-05,
      "loss": 0.6711,
      "step": 314
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995447212286392e-05,
      "loss": 0.7215,
      "step": 315
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9953887011773934e-05,
      "loss": 0.7533,
      "step": 316
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9953298173501007e-05,
      "loss": 0.7127,
      "step": 317
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9952705608265628e-05,
      "loss": 0.7642,
      "step": 318
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995210931628967e-05,
      "loss": 0.4624,
      "step": 319
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9951509297796427e-05,
      "loss": 0.6976,
      "step": 320
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.995090555301056e-05,
      "loss": 0.7235,
      "step": 321
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9950298082158146e-05,
      "loss": 0.6917,
      "step": 322
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9949686885466645e-05,
      "loss": 0.7465,
      "step": 323
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9949071963164918e-05,
      "loss": 0.5137,
      "step": 324
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994845331548322e-05,
      "loss": 0.5533,
      "step": 325
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.99478309426532e-05,
      "loss": 0.6244,
      "step": 326
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9947204844907903e-05,
      "loss": 0.6671,
      "step": 327
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9946575022481764e-05,
      "loss": 0.5778,
      "step": 328
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9945941475610623e-05,
      "loss": 0.6731,
      "step": 329
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994530420453171e-05,
      "loss": 0.6014,
      "step": 330
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9944663209483636e-05,
      "loss": 0.6393,
      "step": 331
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994401849070643e-05,
      "loss": 0.5364,
      "step": 332
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.99433700484415e-05,
      "loss": 0.6886,
      "step": 333
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994271788293165e-05,
      "loss": 0.7886,
      "step": 334
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9942061994421088e-05,
      "loss": 0.7636,
      "step": 335
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.99414023831554e-05,
      "loss": 0.6779,
      "step": 336
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.994073904938157e-05,
      "loss": 0.6993,
      "step": 337
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9940071993347992e-05,
      "loss": 0.7263,
      "step": 338
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9939401215304435e-05,
      "loss": 0.6105,
      "step": 339
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9938726715502067e-05,
      "loss": 0.6277,
      "step": 340
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9938048494193458e-05,
      "loss": 0.699,
      "step": 341
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9937366551632555e-05,
      "loss": 0.5902,
      "step": 342
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9936680888074716e-05,
      "loss": 0.6665,
      "step": 343
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.993599150377668e-05,
      "loss": 0.5722,
      "step": 344
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9935298398996584e-05,
      "loss": 0.6369,
      "step": 345
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.993460157399396e-05,
      "loss": 0.6279,
      "step": 346
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9933901029029732e-05,
      "loss": 0.7698,
      "step": 347
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.993319676436621e-05,
      "loss": 0.6217,
      "step": 348
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9932488780267106e-05,
      "loss": 0.6714,
      "step": 349
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9931777076997516e-05,
      "loss": 0.5437,
      "step": 350
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9931061654823942e-05,
      "loss": 0.7429,
      "step": 351
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9930342514014263e-05,
      "loss": 0.6626,
      "step": 352
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9929619654837765e-05,
      "loss": 0.6474,
      "step": 353
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992889307756511e-05,
      "loss": 0.6279,
      "step": 354
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992816278246837e-05,
      "loss": 0.7801,
      "step": 355
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992742876982099e-05,
      "loss": 0.7405,
      "step": 356
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992669103989783e-05,
      "loss": 0.7276,
      "step": 357
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9925949592975117e-05,
      "loss": 0.6978,
      "step": 358
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9925204429330492e-05,
      "loss": 0.6032,
      "step": 359
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9924455549242972e-05,
      "loss": 0.5869,
      "step": 360
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9923702952992973e-05,
      "loss": 0.621,
      "step": 361
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.99229466408623e-05,
      "loss": 0.6377,
      "step": 362
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9922186613134152e-05,
      "loss": 0.6187,
      "step": 363
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.992142287009312e-05,
      "loss": 0.6818,
      "step": 364
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9920655412025175e-05,
      "loss": 0.8135,
      "step": 365
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9919884239217695e-05,
      "loss": 0.6521,
      "step": 366
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9919109351959444e-05,
      "loss": 0.4523,
      "step": 367
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9918330750540564e-05,
      "loss": 0.6957,
      "step": 368
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9917548435252608e-05,
      "loss": 0.8135,
      "step": 369
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9916762406388507e-05,
      "loss": 0.5928,
      "step": 370
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9915972664242587e-05,
      "loss": 0.6949,
      "step": 371
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.991517920911056e-05,
      "loss": 0.5844,
      "step": 372
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9914382041289533e-05,
      "loss": 0.4758,
      "step": 373
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9913581161078002e-05,
      "loss": 0.7041,
      "step": 374
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9912776568775852e-05,
      "loss": 0.5166,
      "step": 375
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.991196826468436e-05,
      "loss": 0.6685,
      "step": 376
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9911156249106186e-05,
      "loss": 0.7114,
      "step": 377
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9910340522345395e-05,
      "loss": 0.7476,
      "step": 378
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.990952108470742e-05,
      "loss": 0.5799,
      "step": 379
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9908697936499105e-05,
      "loss": 0.6284,
      "step": 380
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.990787107802867e-05,
      "loss": 0.5422,
      "step": 381
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9907040509605725e-05,
      "loss": 0.6316,
      "step": 382
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9906206231541273e-05,
      "loss": 0.6247,
      "step": 383
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9905368244147713e-05,
      "loss": 0.5781,
      "step": 384
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9904526547738812e-05,
      "loss": 0.6742,
      "step": 385
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9903681142629752e-05,
      "loss": 0.6599,
      "step": 386
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9902832029137086e-05,
      "loss": 0.5974,
      "step": 387
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9901979207578755e-05,
      "loss": 0.5805,
      "step": 388
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9901122678274106e-05,
      "loss": 0.6521,
      "step": 389
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9900262441543848e-05,
      "loss": 0.7784,
      "step": 390
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9899398497710098e-05,
      "loss": 0.888,
      "step": 391
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9898530847096354e-05,
      "loss": 0.5475,
      "step": 392
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9897659490027512e-05,
      "loss": 0.6568,
      "step": 393
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9896784426829838e-05,
      "loss": 1.0618,
      "step": 394
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9895905657830995e-05,
      "loss": 0.6606,
      "step": 395
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.989502318336004e-05,
      "loss": 0.5988,
      "step": 396
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9894137003747404e-05,
      "loss": 0.6945,
      "step": 397
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9893247119324916e-05,
      "loss": 0.6397,
      "step": 398
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9892353530425792e-05,
      "loss": 0.7417,
      "step": 399
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9891456237384626e-05,
      "loss": 0.6551,
      "step": 400
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.989055524053741e-05,
      "loss": 0.583,
      "step": 401
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9889650540221514e-05,
      "loss": 0.8043,
      "step": 402
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9888742136775695e-05,
      "loss": 0.5808,
      "step": 403
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.988783003054011e-05,
      "loss": 0.4911,
      "step": 404
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9886914221856285e-05,
      "loss": 0.6328,
      "step": 405
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9885994711067143e-05,
      "loss": 0.5502,
      "step": 406
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.988507149851699e-05,
      "loss": 0.6709,
      "step": 407
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9884144584551517e-05,
      "loss": 0.6058,
      "step": 408
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9883213969517806e-05,
      "loss": 0.598,
      "step": 409
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.988227965376431e-05,
      "loss": 0.7248,
      "step": 410
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9881341637640895e-05,
      "loss": 0.6113,
      "step": 411
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9880399921498786e-05,
      "loss": 0.6465,
      "step": 412
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.987945450569061e-05,
      "loss": 0.5249,
      "step": 413
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.987850539057036e-05,
      "loss": 0.5956,
      "step": 414
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9877552576493442e-05,
      "loss": 0.6303,
      "step": 415
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.987659606381663e-05,
      "loss": 0.6691,
      "step": 416
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.987563585289808e-05,
      "loss": 0.6176,
      "step": 417
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9874671944097338e-05,
      "loss": 0.6232,
      "step": 418
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9873704337775343e-05,
      "loss": 0.6552,
      "step": 419
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9872733034294403e-05,
      "loss": 0.6103,
      "step": 420
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.987175803401822e-05,
      "loss": 0.6751,
      "step": 421
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9870779337311876e-05,
      "loss": 0.5589,
      "step": 422
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9869796944541842e-05,
      "loss": 0.755,
      "step": 423
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9868810856075965e-05,
      "loss": 0.5386,
      "step": 424
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9867821072283492e-05,
      "loss": 0.6365,
      "step": 425
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.986682759353503e-05,
      "loss": 0.5795,
      "step": 426
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9865830420202587e-05,
      "loss": 0.6278,
      "step": 427
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.986482955265955e-05,
      "loss": 0.5914,
      "step": 428
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.986382499128069e-05,
      "loss": 0.5987,
      "step": 429
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.986281673644216e-05,
      "loss": 0.5812,
      "step": 430
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.986180478852149e-05,
      "loss": 0.5928,
      "step": 431
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9860789147897605e-05,
      "loss": 0.4536,
      "step": 432
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.985976981495081e-05,
      "loss": 0.5398,
      "step": 433
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9858746790062778e-05,
      "loss": 0.6403,
      "step": 434
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.985772007361658e-05,
      "loss": 0.6704,
      "step": 435
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9856689665996672e-05,
      "loss": 0.6578,
      "step": 436
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9855655567588877e-05,
      "loss": 0.5833,
      "step": 437
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.985461777878041e-05,
      "loss": 0.5668,
      "step": 438
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9853576299959866e-05,
      "loss": 0.643,
      "step": 439
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.985253113151722e-05,
      "loss": 0.7148,
      "step": 440
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9851482273843833e-05,
      "loss": 0.5307,
      "step": 441
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.985042972733244e-05,
      "loss": 0.6312,
      "step": 442
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9849373492377172e-05,
      "loss": 0.7208,
      "step": 443
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9848313569373518e-05,
      "loss": 0.6216,
      "step": 444
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9847249958718367e-05,
      "loss": 0.6125,
      "step": 445
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9846182660809983e-05,
      "loss": 0.8181,
      "step": 446
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.984511167604801e-05,
      "loss": 0.8014,
      "step": 447
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.984403700483347e-05,
      "loss": 0.7352,
      "step": 448
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.984295864756877e-05,
      "loss": 0.7524,
      "step": 449
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9841876604657703e-05,
      "loss": 0.6212,
      "step": 450
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9840790876505422e-05,
      "loss": 0.6664,
      "step": 451
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.983970146351848e-05,
      "loss": 0.7818,
      "step": 452
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9838608366104796e-05,
      "loss": 0.6461,
      "step": 453
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9837511584673685e-05,
      "loss": 0.6943,
      "step": 454
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.983641111963582e-05,
      "loss": 0.7532,
      "step": 455
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9835306971403275e-05,
      "loss": 0.5039,
      "step": 456
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9834199140389485e-05,
      "loss": 0.6018,
      "step": 457
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9833087627009274e-05,
      "loss": 0.5084,
      "step": 458
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.983197243167884e-05,
      "loss": 0.8088,
      "step": 459
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.983085355481577e-05,
      "loss": 0.6736,
      "step": 460
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.982973099683902e-05,
      "loss": 0.6053,
      "step": 461
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9828604758168918e-05,
      "loss": 0.6537,
      "step": 462
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9827474839227188e-05,
      "loss": 0.842,
      "step": 463
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9826341240436915e-05,
      "loss": 0.6273,
      "step": 464
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9825203962222573e-05,
      "loss": 0.6632,
      "step": 465
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9824063005010008e-05,
      "loss": 0.499,
      "step": 466
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.982291836922645e-05,
      "loss": 0.7698,
      "step": 467
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.98217700553005e-05,
      "loss": 0.6533,
      "step": 468
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9820618063662133e-05,
      "loss": 0.5688,
      "step": 469
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9819462394742715e-05,
      "loss": 0.6132,
      "step": 470
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.981830304897498e-05,
      "loss": 0.5624,
      "step": 471
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.981714002679303e-05,
      "loss": 0.7985,
      "step": 472
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9815973328632362e-05,
      "loss": 0.5627,
      "step": 473
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9814802954929838e-05,
      "loss": 0.7494,
      "step": 474
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9813628906123696e-05,
      "loss": 0.6231,
      "step": 475
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9812451182653554e-05,
      "loss": 0.5332,
      "step": 476
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9811269784960404e-05,
      "loss": 0.5872,
      "step": 477
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9810084713486617e-05,
      "loss": 0.5988,
      "step": 478
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9808895968675937e-05,
      "loss": 0.6907,
      "step": 479
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9807703550973477e-05,
      "loss": 0.59,
      "step": 480
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.980650746082574e-05,
      "loss": 0.5273,
      "step": 481
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9805307698680592e-05,
      "loss": 0.5967,
      "step": 482
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9804104264987278e-05,
      "loss": 0.6169,
      "step": 483
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.980289716019642e-05,
      "loss": 0.5768,
      "step": 484
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.980168638476001e-05,
      "loss": 0.5397,
      "step": 485
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9800471939131418e-05,
      "loss": 0.5614,
      "step": 486
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9799253823765383e-05,
      "loss": 0.6865,
      "step": 487
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9798032039118028e-05,
      "loss": 0.7262,
      "step": 488
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.979680658564684e-05,
      "loss": 0.5734,
      "step": 489
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9795577463810686e-05,
      "loss": 0.589,
      "step": 490
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9794344674069807e-05,
      "loss": 0.6562,
      "step": 491
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.979310821688581e-05,
      "loss": 0.7511,
      "step": 492
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.979186809272168e-05,
      "loss": 0.6116,
      "step": 493
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.979062430204178e-05,
      "loss": 0.7909,
      "step": 494
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.978937684531184e-05,
      "loss": 0.5213,
      "step": 495
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9788125722998958e-05,
      "loss": 0.4475,
      "step": 496
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9786870935571617e-05,
      "loss": 0.6951,
      "step": 497
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9785612483499666e-05,
      "loss": 0.5588,
      "step": 498
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9784350367254322e-05,
      "loss": 0.7768,
      "step": 499
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.978308458730818e-05,
      "loss": 0.7764,
      "step": 500
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9781815144135204e-05,
      "loss": 0.7041,
      "step": 501
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9780542038210733e-05,
      "loss": 0.6452,
      "step": 502
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9779265270011474e-05,
      "loss": 0.6745,
      "step": 503
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9777984840015505e-05,
      "loss": 0.761,
      "step": 504
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9776700748702275e-05,
      "loss": 0.701,
      "step": 505
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9775412996552612e-05,
      "loss": 0.5742,
      "step": 506
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.97741215840487e-05,
      "loss": 0.5596,
      "step": 507
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.977282651167411e-05,
      "loss": 0.6739,
      "step": 508
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9771527779913774e-05,
      "loss": 0.7232,
      "step": 509
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.977022538925399e-05,
      "loss": 0.6018,
      "step": 510
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9768919340182437e-05,
      "loss": 0.6422,
      "step": 511
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.976760963318815e-05,
      "loss": 0.6974,
      "step": 512
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9766296268761555e-05,
      "loss": 0.5416,
      "step": 513
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9764979247394428e-05,
      "loss": 0.4937,
      "step": 514
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9763658569579916e-05,
      "loss": 0.6419,
      "step": 515
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.976233423581255e-05,
      "loss": 0.8821,
      "step": 516
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9761006246588217e-05,
      "loss": 0.681,
      "step": 517
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9759674602404173e-05,
      "loss": 0.7926,
      "step": 518
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9758339303759046e-05,
      "loss": 0.7327,
      "step": 519
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9757000351152834e-05,
      "loss": 0.6987,
      "step": 520
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9755657745086898e-05,
      "loss": 0.834,
      "step": 521
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9754311486063973e-05,
      "loss": 0.5376,
      "step": 522
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.975296157458816e-05,
      "loss": 0.6339,
      "step": 523
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9751608011164922e-05,
      "loss": 0.6089,
      "step": 524
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9750250796301098e-05,
      "loss": 0.6588,
      "step": 525
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9748889930504886e-05,
      "loss": 0.6911,
      "step": 526
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9747525414285863e-05,
      "loss": 0.8009,
      "step": 527
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9746157248154954e-05,
      "loss": 0.6687,
      "step": 528
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9744785432624473e-05,
      "loss": 0.5702,
      "step": 529
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9743409968208084e-05,
      "loss": 0.6122,
      "step": 530
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.974203085542082e-05,
      "loss": 0.8507,
      "step": 531
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.974064809477909e-05,
      "loss": 0.5668,
      "step": 532
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9739261686800662e-05,
      "loss": 0.8123,
      "step": 533
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.973787163200466e-05,
      "loss": 0.5727,
      "step": 534
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9736477930911588e-05,
      "loss": 0.674,
      "step": 535
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9735080584043314e-05,
      "loss": 0.6002,
      "step": 536
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9733679591923062e-05,
      "loss": 0.6238,
      "step": 537
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.973227495507543e-05,
      "loss": 0.7495,
      "step": 538
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9730866674026378e-05,
      "loss": 0.5555,
      "step": 539
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9729454749303227e-05,
      "loss": 0.6144,
      "step": 540
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9728039181434666e-05,
      "loss": 0.7028,
      "step": 541
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9726619970950744e-05,
      "loss": 0.7399,
      "step": 542
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9725197118382884e-05,
      "loss": 0.5762,
      "step": 543
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9723770624263862e-05,
      "loss": 0.7227,
      "step": 544
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9722340489127824e-05,
      "loss": 0.6282,
      "step": 545
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.972090671351027e-05,
      "loss": 0.639,
      "step": 546
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9719469297948076e-05,
      "loss": 0.8125,
      "step": 547
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9718028242979478e-05,
      "loss": 0.624,
      "step": 548
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9716583549144067e-05,
      "loss": 0.7245,
      "step": 549
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.97151352169828e-05,
      "loss": 0.5894,
      "step": 550
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9713683247038e-05,
      "loss": 0.6551,
      "step": 551
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9712227639853352e-05,
      "loss": 0.5875,
      "step": 552
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9710768395973897e-05,
      "loss": 0.5908,
      "step": 553
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9709305515946042e-05,
      "loss": 0.6481,
      "step": 554
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9707839000317557e-05,
      "loss": 0.7637,
      "step": 555
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.970636884963757e-05,
      "loss": 0.4863,
      "step": 556
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9704895064456573e-05,
      "loss": 0.7046,
      "step": 557
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9703417645326418e-05,
      "loss": 0.6052,
      "step": 558
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.970193659280031e-05,
      "loss": 0.7102,
      "step": 559
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9700451907432835e-05,
      "loss": 0.7297,
      "step": 560
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9698963589779915e-05,
      "loss": 0.7229,
      "step": 561
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.969747164039885e-05,
      "loss": 0.6624,
      "step": 562
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9695976059848282e-05,
      "loss": 0.7479,
      "step": 563
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9694476848688236e-05,
      "loss": 0.5644,
      "step": 564
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.969297400748008e-05,
      "loss": 0.7066,
      "step": 565
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.969146753678654e-05,
      "loss": 0.5552,
      "step": 566
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.968995743717171e-05,
      "loss": 0.65,
      "step": 567
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9688443709201043e-05,
      "loss": 0.7314,
      "step": 568
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9686926353441347e-05,
      "loss": 0.7631,
      "step": 569
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.968540537046078e-05,
      "loss": 0.7732,
      "step": 570
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9683880760828874e-05,
      "loss": 0.5251,
      "step": 571
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.968235252511651e-05,
      "loss": 0.6859,
      "step": 572
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.968082066389592e-05,
      "loss": 0.7438,
      "step": 573
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9679285177740715e-05,
      "loss": 0.5103,
      "step": 574
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9677746067225843e-05,
      "loss": 0.7103,
      "step": 575
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9676203332927616e-05,
      "loss": 0.3896,
      "step": 576
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9674656975423704e-05,
      "loss": 0.6445,
      "step": 577
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9673106995293135e-05,
      "loss": 0.5784,
      "step": 578
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9671553393116286e-05,
      "loss": 0.5864,
      "step": 579
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9669996169474903e-05,
      "loss": 0.6886,
      "step": 580
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9668435324952076e-05,
      "loss": 0.7056,
      "step": 581
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.966687086013225e-05,
      "loss": 0.6895,
      "step": 582
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.966530277560124e-05,
      "loss": 0.6188,
      "step": 583
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9663731071946207e-05,
      "loss": 0.5383,
      "step": 584
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.966215574975566e-05,
      "loss": 0.5419,
      "step": 585
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9660576809619475e-05,
      "loss": 0.7643,
      "step": 586
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9658994252128884e-05,
      "loss": 0.5683,
      "step": 587
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9657408077876458e-05,
      "loss": 0.5302,
      "step": 588
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9655818287456133e-05,
      "loss": 0.8155,
      "step": 589
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9654224881463205e-05,
      "loss": 0.6518,
      "step": 590
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9652627860494313e-05,
      "loss": 0.5929,
      "step": 591
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.965102722514745e-05,
      "loss": 0.7371,
      "step": 592
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.964942297602197e-05,
      "loss": 0.4912,
      "step": 593
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9647815113718577e-05,
      "loss": 0.8094,
      "step": 594
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9646203638839327e-05,
      "loss": 0.7346,
      "step": 595
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9644588551987623e-05,
      "loss": 0.8466,
      "step": 596
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.964296985376823e-05,
      "loss": 0.647,
      "step": 597
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.964134754478726e-05,
      "loss": 0.622,
      "step": 598
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9639721625652184e-05,
      "loss": 0.6798,
      "step": 599
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.963809209697181e-05,
      "loss": 0.6253,
      "step": 600
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.963645895935632e-05,
      "loss": 0.6333,
      "step": 601
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9634822213417216e-05,
      "loss": 0.6672,
      "step": 602
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9633181859767385e-05,
      "loss": 0.6169,
      "step": 603
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.963153789902104e-05,
      "loss": 0.7079,
      "step": 604
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9629890331793762e-05,
      "loss": 0.6191,
      "step": 605
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9628239158702468e-05,
      "loss": 0.7398,
      "step": 606
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.962658438036543e-05,
      "loss": 0.6511,
      "step": 607
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.962492599740228e-05,
      "loss": 0.4896,
      "step": 608
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.962326401043398e-05,
      "loss": 0.5733,
      "step": 609
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9621598420082863e-05,
      "loss": 0.6114,
      "step": 610
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9619929226972593e-05,
      "loss": 0.7648,
      "step": 611
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.961825643172819e-05,
      "loss": 0.666,
      "step": 612
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9616580034976037e-05,
      "loss": 0.5617,
      "step": 613
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9614900037343837e-05,
      "loss": 0.636,
      "step": 614
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9613216439460664e-05,
      "loss": 0.6065,
      "step": 615
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9611529241956933e-05,
      "loss": 0.6504,
      "step": 616
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9609838445464406e-05,
      "loss": 0.6761,
      "step": 617
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9608144050616192e-05,
      "loss": 0.6545,
      "step": 618
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.960644605804675e-05,
      "loss": 0.5833,
      "step": 619
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9604744468391882e-05,
      "loss": 0.855,
      "step": 620
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.960303928228874e-05,
      "loss": 0.7429,
      "step": 621
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9601330500375827e-05,
      "loss": 0.6052,
      "step": 622
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9599618123292985e-05,
      "loss": 0.5438,
      "step": 623
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.95979021516814e-05,
      "loss": 0.5654,
      "step": 624
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.959618258618362e-05,
      "loss": 0.6089,
      "step": 625
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.959445942744352e-05,
      "loss": 0.7795,
      "step": 626
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.959273267610633e-05,
      "loss": 0.5702,
      "step": 627
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.959100233281862e-05,
      "loss": 0.6607,
      "step": 628
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9589268398228313e-05,
      "loss": 0.6178,
      "step": 629
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9587530872984672e-05,
      "loss": 0.6211,
      "step": 630
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9585789757738298e-05,
      "loss": 0.5015,
      "step": 631
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.958404505314115e-05,
      "loss": 0.6951,
      "step": 632
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.958229675984652e-05,
      "loss": 0.7686,
      "step": 633
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9580544878509054e-05,
      "loss": 0.6494,
      "step": 634
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9578789409784727e-05,
      "loss": 0.4908,
      "step": 635
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.957703035433087e-05,
      "loss": 0.7011,
      "step": 636
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9575267712806152e-05,
      "loss": 0.5187,
      "step": 637
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9573501485870585e-05,
      "loss": 0.6435,
      "step": 638
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9571731674185523e-05,
      "loss": 0.7211,
      "step": 639
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.956995827841366e-05,
      "loss": 0.5494,
      "step": 640
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9568181299219042e-05,
      "loss": 0.6652,
      "step": 641
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9566400737267045e-05,
      "loss": 0.7224,
      "step": 642
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9564616593224396e-05,
      "loss": 0.6485,
      "step": 643
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9562828867759154e-05,
      "loss": 0.5622,
      "step": 644
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9561037561540722e-05,
      "loss": 0.4832,
      "step": 645
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9559242675239855e-05,
      "loss": 0.5593,
      "step": 646
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.955744420952863e-05,
      "loss": 0.5924,
      "step": 647
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9555642165080475e-05,
      "loss": 0.6536,
      "step": 648
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.955383654257016e-05,
      "loss": 0.5222,
      "step": 649
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.955202734267379e-05,
      "loss": 0.7254,
      "step": 650
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.955021456606881e-05,
      "loss": 0.5802,
      "step": 651
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.954839821343401e-05,
      "loss": 0.6827,
      "step": 652
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9546578285449504e-05,
      "loss": 0.7597,
      "step": 653
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.954475478279676e-05,
      "loss": 0.5999,
      "step": 654
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.954292770615858e-05,
      "loss": 0.7118,
      "step": 655
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9541097056219108e-05,
      "loss": 0.5568,
      "step": 656
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9539262833663813e-05,
      "loss": 0.6132,
      "step": 657
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9537425039179518e-05,
      "loss": 0.5714,
      "step": 658
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9535583673454365e-05,
      "loss": 0.6229,
      "step": 659
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.953373873717786e-05,
      "loss": 0.5683,
      "step": 660
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9531890231040814e-05,
      "loss": 0.5276,
      "step": 661
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9530038155735405e-05,
      "loss": 0.6613,
      "step": 662
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9528182511955124e-05,
      "loss": 0.7438,
      "step": 663
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9526323300394813e-05,
      "loss": 0.4759,
      "step": 664
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.952446052175064e-05,
      "loss": 0.7326,
      "step": 665
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.952259417672011e-05,
      "loss": 0.541,
      "step": 666
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9520724266002078e-05,
      "loss": 0.6963,
      "step": 667
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9518850790296712e-05,
      "loss": 0.5858,
      "step": 668
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.951697375030553e-05,
      "loss": 0.6852,
      "step": 669
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.951509314673138e-05,
      "loss": 0.6302,
      "step": 670
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.951320898027845e-05,
      "loss": 0.5549,
      "step": 671
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9511321251652243e-05,
      "loss": 0.5985,
      "step": 672
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9509429961559624e-05,
      "loss": 0.7871,
      "step": 673
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9507535110708766e-05,
      "loss": 0.4962,
      "step": 674
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9505636699809192e-05,
      "loss": 0.6802,
      "step": 675
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9503734729571752e-05,
      "loss": 0.6865,
      "step": 676
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9501829200708627e-05,
      "loss": 0.7255,
      "step": 677
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9499920113933335e-05,
      "loss": 0.6856,
      "step": 678
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.949800746996072e-05,
      "loss": 0.5618,
      "step": 679
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9496091269506968e-05,
      "loss": 0.6032,
      "step": 680
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9494171513289587e-05,
      "loss": 0.64,
      "step": 681
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9492248202027425e-05,
      "loss": 0.5501,
      "step": 682
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9490321336440646e-05,
      "loss": 0.7113,
      "step": 683
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9488390917250763e-05,
      "loss": 0.7672,
      "step": 684
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.948645694518061e-05,
      "loss": 0.6695,
      "step": 685
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9484519420954356e-05,
      "loss": 0.6951,
      "step": 686
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.948257834529749e-05,
      "loss": 0.7459,
      "step": 687
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9480633718936844e-05,
      "loss": 0.5781,
      "step": 688
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.947868554260057e-05,
      "loss": 0.6469,
      "step": 689
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9476733817018155e-05,
      "loss": 0.7176,
      "step": 690
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9474778542920413e-05,
      "loss": 0.6698,
      "step": 691
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.947281972103949e-05,
      "loss": 0.5304,
      "step": 692
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9470857352108848e-05,
      "loss": 0.5724,
      "step": 693
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.946889143686329e-05,
      "loss": 0.5637,
      "step": 694
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9466921976038947e-05,
      "loss": 0.6861,
      "step": 695
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9464948970373272e-05,
      "loss": 0.5744,
      "step": 696
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9462972420605045e-05,
      "loss": 0.587,
      "step": 697
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9460992327474373e-05,
      "loss": 0.6041,
      "step": 698
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9459008691722697e-05,
      "loss": 0.6414,
      "step": 699
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.945702151409278e-05,
      "loss": 0.7627,
      "step": 700
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9455030795328706e-05,
      "loss": 0.5796,
      "step": 701
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.945303653617589e-05,
      "loss": 0.7071,
      "step": 702
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9451038737381078e-05,
      "loss": 0.5868,
      "step": 703
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.944903739969233e-05,
      "loss": 0.5489,
      "step": 704
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.944703252385904e-05,
      "loss": 0.6015,
      "step": 705
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9445024110631922e-05,
      "loss": 0.5639,
      "step": 706
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9443012160763014e-05,
      "loss": 0.5135,
      "step": 707
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9440996675005683e-05,
      "loss": 0.682,
      "step": 708
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9438977654114616e-05,
      "loss": 0.5945,
      "step": 709
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9436955098845832e-05,
      "loss": 0.6656,
      "step": 710
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9434929009956657e-05,
      "loss": 0.7992,
      "step": 711
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9432899388205756e-05,
      "loss": 0.7891,
      "step": 712
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9430866234353113e-05,
      "loss": 0.5022,
      "step": 713
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9428829549160024e-05,
      "loss": 0.6958,
      "step": 714
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.942678933338912e-05,
      "loss": 0.666,
      "step": 715
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.942474558780435e-05,
      "loss": 0.5324,
      "step": 716
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9422698313170982e-05,
      "loss": 0.6685,
      "step": 717
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.942064751025561e-05,
      "loss": 0.6058,
      "step": 718
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9418593179826143e-05,
      "loss": 0.6526,
      "step": 719
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.941653532265182e-05,
      "loss": 0.5976,
      "step": 720
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9414473939503193e-05,
      "loss": 0.6152,
      "step": 721
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9412409031152136e-05,
      "loss": 0.6327,
      "step": 722
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9410340598371845e-05,
      "loss": 0.6101,
      "step": 723
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9408268641936832e-05,
      "loss": 0.7224,
      "step": 724
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9406193162622934e-05,
      "loss": 0.7438,
      "step": 725
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.94041141612073e-05,
      "loss": 0.6451,
      "step": 726
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9402031638468407e-05,
      "loss": 0.4848,
      "step": 727
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9399945595186037e-05,
      "loss": 0.6546,
      "step": 728
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9397856032141305e-05,
      "loss": 0.5252,
      "step": 729
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9395762950116634e-05,
      "loss": 0.6216,
      "step": 730
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9393666349895772e-05,
      "loss": 0.5868,
      "step": 731
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9391566232263776e-05,
      "loss": 0.7081,
      "step": 732
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9389462598007027e-05,
      "loss": 0.838,
      "step": 733
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.938735544791322e-05,
      "loss": 0.6864,
      "step": 734
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9385244782771365e-05,
      "loss": 0.5882,
      "step": 735
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.938313060337179e-05,
      "loss": 0.5637,
      "step": 736
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9381012910506146e-05,
      "loss": 0.5931,
      "step": 737
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.937889170496738e-05,
      "loss": 0.455,
      "step": 738
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9376766987549775e-05,
      "loss": 0.5748,
      "step": 739
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.937463875904892e-05,
      "loss": 0.5796,
      "step": 740
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9372507020261716e-05,
      "loss": 0.6434,
      "step": 741
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9370371771986388e-05,
      "loss": 0.5936,
      "step": 742
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9368233015022464e-05,
      "loss": 0.6566,
      "step": 743
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9366090750170793e-05,
      "loss": 0.4777,
      "step": 744
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9363944978233533e-05,
      "loss": 0.5978,
      "step": 745
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.936179570001416e-05,
      "loss": 0.7114,
      "step": 746
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.935964291631746e-05,
      "loss": 0.5794,
      "step": 747
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.935748662794953e-05,
      "loss": 0.5913,
      "step": 748
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.935532683571778e-05,
      "loss": 0.6191,
      "step": 749
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9353163540430942e-05,
      "loss": 0.5596,
      "step": 750
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9350996742899043e-05,
      "loss": 0.68,
      "step": 751
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.934882644393343e-05,
      "loss": 0.6164,
      "step": 752
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9346652644346763e-05,
      "loss": 0.781,
      "step": 753
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.934447534495301e-05,
      "loss": 0.6546,
      "step": 754
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.934229454656745e-05,
      "loss": 0.66,
      "step": 755
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9340110250006672e-05,
      "loss": 0.9669,
      "step": 756
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.933792245608857e-05,
      "loss": 0.6082,
      "step": 757
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.933573116563236e-05,
      "loss": 0.5477,
      "step": 758
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.933353637945856e-05,
      "loss": 0.6458,
      "step": 759
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9331338098388985e-05,
      "loss": 0.6161,
      "step": 760
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.932913632324678e-05,
      "loss": 0.7132,
      "step": 761
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.932693105485639e-05,
      "loss": 0.5627,
      "step": 762
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.932472229404356e-05,
      "loss": 0.4438,
      "step": 763
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9322510041635353e-05,
      "loss": 0.8241,
      "step": 764
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.932029429846014e-05,
      "loss": 0.5992,
      "step": 765
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9318075065347584e-05,
      "loss": 0.5083,
      "step": 766
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9315852343128677e-05,
      "loss": 0.6782,
      "step": 767
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9313626132635702e-05,
      "loss": 0.5897,
      "step": 768
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9311396434702247e-05,
      "loss": 0.6798,
      "step": 769
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9309163250163222e-05,
      "loss": 0.6842,
      "step": 770
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.930692657985482e-05,
      "loss": 0.6186,
      "step": 771
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.930468642461456e-05,
      "loss": 0.769,
      "step": 772
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9302442785281252e-05,
      "loss": 0.6532,
      "step": 773
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.930019566269502e-05,
      "loss": 0.5093,
      "step": 774
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.929794505769728e-05,
      "loss": 0.5877,
      "step": 775
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9295690971130763e-05,
      "loss": 0.5852,
      "step": 776
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9293433403839506e-05,
      "loss": 0.5103,
      "step": 777
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9291172356668843e-05,
      "loss": 0.6663,
      "step": 778
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9288907830465403e-05,
      "loss": 0.6642,
      "step": 779
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9286639826077128e-05,
      "loss": 0.4564,
      "step": 780
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9284368344353273e-05,
      "loss": 0.6327,
      "step": 781
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.928209338614437e-05,
      "loss": 0.5594,
      "step": 782
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9279814952302266e-05,
      "loss": 0.6267,
      "step": 783
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.927753304368012e-05,
      "loss": 0.4985,
      "step": 784
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9275247661132373e-05,
      "loss": 0.5941,
      "step": 785
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9272958805514775e-05,
      "loss": 0.6993,
      "step": 786
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9270666477684375e-05,
      "loss": 0.5463,
      "step": 787
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.926837067849953e-05,
      "loss": 0.685,
      "step": 788
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.926607140881989e-05,
      "loss": 0.5844,
      "step": 789
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9263768669506398e-05,
      "loss": 0.8152,
      "step": 790
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9261462461421308e-05,
      "loss": 0.5646,
      "step": 791
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9259152785428165e-05,
      "loss": 0.7259,
      "step": 792
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9256839642391822e-05,
      "loss": 0.5941,
      "step": 793
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9254523033178416e-05,
      "loss": 0.6839,
      "step": 794
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9252202958655393e-05,
      "loss": 0.4982,
      "step": 795
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9249879419691492e-05,
      "loss": 0.7201,
      "step": 796
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9247552417156758e-05,
      "loss": 0.7528,
      "step": 797
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9245221951922515e-05,
      "loss": 0.6746,
      "step": 798
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.92428880248614e-05,
      "loss": 0.6949,
      "step": 799
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9240550636847335e-05,
      "loss": 0.6712,
      "step": 800
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.923820978875555e-05,
      "loss": 0.5549,
      "step": 801
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9235865481462557e-05,
      "loss": 0.6585,
      "step": 802
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9233517715846176e-05,
      "loss": 0.5416,
      "step": 803
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9231166492785514e-05,
      "loss": 0.6663,
      "step": 804
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9228811813160972e-05,
      "loss": 0.5909,
      "step": 805
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9226453677854255e-05,
      "loss": 0.804,
      "step": 806
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9224092087748344e-05,
      "loss": 0.6092,
      "step": 807
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9221727043727534e-05,
      "loss": 0.6703,
      "step": 808
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.92193585466774e-05,
      "loss": 0.6617,
      "step": 809
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9216986597484814e-05,
      "loss": 0.6041,
      "step": 810
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9214611197037943e-05,
      "loss": 0.6311,
      "step": 811
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9212232346226238e-05,
      "loss": 0.6619,
      "step": 812
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9209850045940452e-05,
      "loss": 0.6272,
      "step": 813
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9207464297072625e-05,
      "loss": 0.6699,
      "step": 814
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9205075100516087e-05,
      "loss": 0.5532,
      "step": 815
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9202682457165463e-05,
      "loss": 0.6767,
      "step": 816
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.920028636791667e-05,
      "loss": 0.5421,
      "step": 817
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9197886833666903e-05,
      "loss": 0.5874,
      "step": 818
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9195483855314656e-05,
      "loss": 0.6839,
      "step": 819
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.919307743375972e-05,
      "loss": 0.6249,
      "step": 820
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.919066756990316e-05,
      "loss": 0.6728,
      "step": 821
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9188254264647338e-05,
      "loss": 0.7826,
      "step": 822
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.918583751889591e-05,
      "loss": 0.5555,
      "step": 823
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9183417333553812e-05,
      "loss": 0.6756,
      "step": 824
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9180993709527268e-05,
      "loss": 0.6456,
      "step": 825
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9178566647723787e-05,
      "loss": 0.568,
      "step": 826
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9176136149052184e-05,
      "loss": 0.7508,
      "step": 827
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9173702214422533e-05,
      "loss": 0.813,
      "step": 828
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9171264844746215e-05,
      "loss": 0.7126,
      "step": 829
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.916882404093589e-05,
      "loss": 0.6835,
      "step": 830
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.916637980390551e-05,
      "loss": 0.6673,
      "step": 831
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9163932134570298e-05,
      "loss": 0.7244,
      "step": 832
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9161481033846773e-05,
      "loss": 0.6283,
      "step": 833
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.915902650265274e-05,
      "loss": 0.6931,
      "step": 834
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9156568541907293e-05,
      "loss": 0.4884,
      "step": 835
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9154107152530792e-05,
      "loss": 0.6362,
      "step": 836
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9151642335444894e-05,
      "loss": 0.6063,
      "step": 837
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.914917409157254e-05,
      "loss": 0.6472,
      "step": 838
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9146702421837952e-05,
      "loss": 0.4123,
      "step": 839
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.914422732716663e-05,
      "loss": 0.6746,
      "step": 840
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9141748808485367e-05,
      "loss": 0.6176,
      "step": 841
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9139266866722227e-05,
      "loss": 0.5804,
      "step": 842
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.913678150280656e-05,
      "loss": 0.6943,
      "step": 843
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9134292717669002e-05,
      "loss": 0.6927,
      "step": 844
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9131800512241466e-05,
      "loss": 0.8294,
      "step": 845
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.912930488745714e-05,
      "loss": 0.5905,
      "step": 846
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9126805844250507e-05,
      "loss": 0.8265,
      "step": 847
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9124303383557316e-05,
      "loss": 0.5976,
      "step": 848
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.91217975063146e-05,
      "loss": 0.6046,
      "step": 849
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.911928821346067e-05,
      "loss": 0.6501,
      "step": 850
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9116775505935122e-05,
      "loss": 0.6571,
      "step": 851
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.911425938467883e-05,
      "loss": 0.8515,
      "step": 852
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9111739850633934e-05,
      "loss": 0.5963,
      "step": 853
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.910921690474387e-05,
      "loss": 0.5839,
      "step": 854
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9106690547953336e-05,
      "loss": 0.5777,
      "step": 855
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.910416078120832e-05,
      "loss": 0.6018,
      "step": 856
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.910162760545607e-05,
      "loss": 0.5631,
      "step": 857
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9099091021645134e-05,
      "loss": 0.7011,
      "step": 858
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9096551030725312e-05,
      "loss": 0.6195,
      "step": 859
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9094007633647695e-05,
      "loss": 0.6846,
      "step": 860
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.909146083136465e-05,
      "loss": 0.6347,
      "step": 861
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9088910624829805e-05,
      "loss": 0.736,
      "step": 862
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9086357014998077e-05,
      "loss": 0.7483,
      "step": 863
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.908380000282565e-05,
      "loss": 0.7445,
      "step": 864
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9081239589269992e-05,
      "loss": 0.7819,
      "step": 865
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9078675775289828e-05,
      "loss": 0.6087,
      "step": 866
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9076108561845167e-05,
      "loss": 0.6911,
      "step": 867
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9073537949897293e-05,
      "loss": 0.6111,
      "step": 868
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9070963940408758e-05,
      "loss": 0.5844,
      "step": 869
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9068386534343383e-05,
      "loss": 0.5167,
      "step": 870
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.906580573266627e-05,
      "loss": 0.5836,
      "step": 871
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.906322153634378e-05,
      "loss": 0.6087,
      "step": 872
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.906063394634356e-05,
      "loss": 0.5723,
      "step": 873
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9058042963634518e-05,
      "loss": 0.4785,
      "step": 874
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.905544858918683e-05,
      "loss": 0.6577,
      "step": 875
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9052850823971952e-05,
      "loss": 0.5798,
      "step": 876
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.90502496689626e-05,
      "loss": 0.6885,
      "step": 877
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9047645125132767e-05,
      "loss": 0.6114,
      "step": 878
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.904503719345771e-05,
      "loss": 0.7577,
      "step": 879
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9042425874913952e-05,
      "loss": 0.5028,
      "step": 880
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9039811170479296e-05,
      "loss": 0.5928,
      "step": 881
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9037193081132796e-05,
      "loss": 0.6201,
      "step": 882
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9034571607854793e-05,
      "loss": 0.6095,
      "step": 883
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9031946751626872e-05,
      "loss": 0.7417,
      "step": 884
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9029318513431908e-05,
      "loss": 0.5673,
      "step": 885
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9026686894254025e-05,
      "loss": 0.719,
      "step": 886
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.902405189507862e-05,
      "loss": 0.6307,
      "step": 887
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.902141351689236e-05,
      "loss": 0.746,
      "step": 888
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.901877176068317e-05,
      "loss": 0.617,
      "step": 889
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.901612662744024e-05,
      "loss": 0.5678,
      "step": 890
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.901347811815403e-05,
      "loss": 0.7067,
      "step": 891
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9010826233816258e-05,
      "loss": 0.6585,
      "step": 892
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.900817097541991e-05,
      "loss": 0.6748,
      "step": 893
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9005512343959235e-05,
      "loss": 0.6226,
      "step": 894
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9002850340429745e-05,
      "loss": 0.5234,
      "step": 895
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.900018496582821e-05,
      "loss": 0.642,
      "step": 896
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.899751622115267e-05,
      "loss": 0.7369,
      "step": 897
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8994844107402423e-05,
      "loss": 0.7871,
      "step": 898
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8992168625578023e-05,
      "loss": 0.494,
      "step": 899
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.89894897766813e-05,
      "loss": 0.6086,
      "step": 900
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8986807561715325e-05,
      "loss": 0.6571,
      "step": 901
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8984121981684448e-05,
      "loss": 0.6841,
      "step": 902
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8981433037594268e-05,
      "loss": 0.6146,
      "step": 903
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8978740730451647e-05,
      "loss": 0.6822,
      "step": 904
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8976045061264706e-05,
      "loss": 0.657,
      "step": 905
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8973346031042825e-05,
      "loss": 0.6619,
      "step": 906
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8970643640796642e-05,
      "loss": 0.6548,
      "step": 907
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8967937891538054e-05,
      "loss": 0.8225,
      "step": 908
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8965228784280212e-05,
      "loss": 0.5337,
      "step": 909
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.896251632003753e-05,
      "loss": 0.6117,
      "step": 910
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.895980049982568e-05,
      "loss": 0.7221,
      "step": 911
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.895708132466158e-05,
      "loss": 0.5542,
      "step": 912
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8954358795563418e-05,
      "loss": 0.6472,
      "step": 913
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8951632913550625e-05,
      "loss": 0.7083,
      "step": 914
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.89489036796439e-05,
      "loss": 0.8085,
      "step": 915
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.894617109486519e-05,
      "loss": 0.6214,
      "step": 916
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8943435160237693e-05,
      "loss": 0.4737,
      "step": 917
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.894069587678587e-05,
      "loss": 0.8245,
      "step": 918
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.893795324553543e-05,
      "loss": 0.5847,
      "step": 919
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.893520726751334e-05,
      "loss": 0.8312,
      "step": 920
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8932457943747814e-05,
      "loss": 0.6219,
      "step": 921
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8929705275268326e-05,
      "loss": 0.6198,
      "step": 922
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8926949263105595e-05,
      "loss": 0.6145,
      "step": 923
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.89241899082916e-05,
      "loss": 0.4266,
      "step": 924
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8921427211859568e-05,
      "loss": 0.7579,
      "step": 925
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.8918661174843974e-05,
      "loss": 0.635,
      "step": 926
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8915891798280545e-05,
      "loss": 0.5516,
      "step": 927
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8913119083206264e-05,
      "loss": 0.737,
      "step": 928
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8910343030659355e-05,
      "loss": 0.6503,
      "step": 929
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8907563641679303e-05,
      "loss": 0.5896,
      "step": 930
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8904780917306837e-05,
      "loss": 0.5248,
      "step": 931
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.890199485858393e-05,
      "loss": 0.5724,
      "step": 932
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.88992054665538e-05,
      "loss": 0.6786,
      "step": 933
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8896412742260935e-05,
      "loss": 0.7046,
      "step": 934
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8893616686751053e-05,
      "loss": 0.6563,
      "step": 935
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8890817301071124e-05,
      "loss": 0.5025,
      "step": 936
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8888014586269353e-05,
      "loss": 0.6277,
      "step": 937
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8885208543395212e-05,
      "loss": 0.6108,
      "step": 938
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.888239917349941e-05,
      "loss": 0.6856,
      "step": 939
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.88795864776339e-05,
      "loss": 0.632,
      "step": 940
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.887677045685188e-05,
      "loss": 0.9351,
      "step": 941
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8873951112207795e-05,
      "loss": 0.5587,
      "step": 942
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8871128444757335e-05,
      "loss": 0.783,
      "step": 943
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8868302455557435e-05,
      "loss": 0.4008,
      "step": 944
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8865473145666272e-05,
      "loss": 0.7884,
      "step": 945
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8862640516143264e-05,
      "loss": 0.5406,
      "step": 946
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8859804568049083e-05,
      "loss": 0.6987,
      "step": 947
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8856965302445623e-05,
      "loss": 0.6288,
      "step": 948
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8854122720396043e-05,
      "loss": 0.6703,
      "step": 949
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.885127682296473e-05,
      "loss": 0.6301,
      "step": 950
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8848427611217316e-05,
      "loss": 0.6698,
      "step": 951
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8845575086220672e-05,
      "loss": 0.7547,
      "step": 952
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8842719249042913e-05,
      "loss": 0.6052,
      "step": 953
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8839860100753396e-05,
      "loss": 0.613,
      "step": 954
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.883699764242271e-05,
      "loss": 0.6204,
      "step": 955
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8834131875122692e-05,
      "loss": 0.607,
      "step": 956
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8831262799926412e-05,
      "loss": 0.792,
      "step": 957
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.882839041790818e-05,
      "loss": 0.559,
      "step": 958
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.882551473014355e-05,
      "loss": 0.6348,
      "step": 959
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.88226357377093e-05,
      "loss": 0.5667,
      "step": 960
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8819753441683464e-05,
      "loss": 0.5702,
      "step": 961
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8816867843145296e-05,
      "loss": 0.6723,
      "step": 962
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8813978943175296e-05,
      "loss": 0.6044,
      "step": 963
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.88110867428552e-05,
      "loss": 0.7332,
      "step": 964
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8808191243267977e-05,
      "loss": 0.5585,
      "step": 965
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8805292445497828e-05,
      "loss": 0.6491,
      "step": 966
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.88023903506302e-05,
      "loss": 0.5677,
      "step": 967
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8799484959751767e-05,
      "loss": 0.6494,
      "step": 968
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.879657627395043e-05,
      "loss": 0.5456,
      "step": 969
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.879366429431534e-05,
      "loss": 0.5685,
      "step": 970
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8790749021936873e-05,
      "loss": 0.5719,
      "step": 971
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8787830457906633e-05,
      "loss": 0.5967,
      "step": 972
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8784908603317467e-05,
      "loss": 0.4431,
      "step": 973
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8781983459263443e-05,
      "loss": 0.5898,
      "step": 974
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.877905502683987e-05,
      "loss": 0.7605,
      "step": 975
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8776123307143285e-05,
      "loss": 0.5926,
      "step": 976
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8773188301271458e-05,
      "loss": 0.8481,
      "step": 977
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.877025001032338e-05,
      "loss": 0.6704,
      "step": 978
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8767308435399286e-05,
      "loss": 0.6114,
      "step": 979
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8764363577600628e-05,
      "loss": 0.5793,
      "step": 980
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.87614154380301e-05,
      "loss": 0.5804,
      "step": 981
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8758464017791614e-05,
      "loss": 0.7032,
      "step": 982
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8755509317990315e-05,
      "loss": 0.6574,
      "step": 983
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8752551339732576e-05,
      "loss": 0.7285,
      "step": 984
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8749590084125998e-05,
      "loss": 0.6979,
      "step": 985
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8746625552279403e-05,
      "loss": 0.4682,
      "step": 986
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.874365774530285e-05,
      "loss": 0.5839,
      "step": 987
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8740686664307622e-05,
      "loss": 0.6532,
      "step": 988
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.873771231040622e-05,
      "loss": 0.7945,
      "step": 989
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8734734684712376e-05,
      "loss": 0.5067,
      "step": 990
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8731753788341052e-05,
      "loss": 0.6928,
      "step": 991
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8728769622408423e-05,
      "loss": 0.7042,
      "step": 992
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.87257821880319e-05,
      "loss": 0.6096,
      "step": 993
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.872279148633011e-05,
      "loss": 0.579,
      "step": 994
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.871979751842291e-05,
      "loss": 0.5624,
      "step": 995
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.871680028543137e-05,
      "loss": 0.6938,
      "step": 996
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8713799788477794e-05,
      "loss": 0.7494,
      "step": 997
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8710796028685702e-05,
      "loss": 0.6804,
      "step": 998
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8707789007179835e-05,
      "loss": 0.5586,
      "step": 999
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8704778725086157e-05,
      "loss": 0.6211,
      "step": 1000
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8701765183531855e-05,
      "loss": 0.7616,
      "step": 1001
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8698748383645334e-05,
      "loss": 0.5837,
      "step": 1002
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8695728326556217e-05,
      "loss": 0.6087,
      "step": 1003
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8692705013395354e-05,
      "loss": 0.724,
      "step": 1004
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8689678445294807e-05,
      "loss": 0.6206,
      "step": 1005
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8686648623387856e-05,
      "loss": 0.5135,
      "step": 1006
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8683615548809007e-05,
      "loss": 0.5427,
      "step": 1007
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.868057922269398e-05,
      "loss": 0.5491,
      "step": 1008
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8677539646179706e-05,
      "loss": 0.5754,
      "step": 1009
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8674496820404347e-05,
      "loss": 0.5521,
      "step": 1010
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.867145074650727e-05,
      "loss": 0.7253,
      "step": 1011
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.866840142562906e-05,
      "loss": 0.7106,
      "step": 1012
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8665348858911524e-05,
      "loss": 0.7113,
      "step": 1013
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8662293047497677e-05,
      "loss": 0.5261,
      "step": 1014
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8659233992531753e-05,
      "loss": 0.6985,
      "step": 1015
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.86561716951592e-05,
      "loss": 0.7584,
      "step": 1016
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.865310615652668e-05,
      "loss": 0.566,
      "step": 1017
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8650037377782072e-05,
      "loss": 0.7234,
      "step": 1018
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.864696536007446e-05,
      "loss": 0.7392,
      "step": 1019
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8643890104554144e-05,
      "loss": 0.5283,
      "step": 1020
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8640811612372643e-05,
      "loss": 0.6838,
      "step": 1021
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.863772988468268e-05,
      "loss": 0.6075,
      "step": 1022
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8634644922638193e-05,
      "loss": 0.6556,
      "step": 1023
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.863155672739433e-05,
      "loss": 0.7148,
      "step": 1024
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8628465300107453e-05,
      "loss": 0.7118,
      "step": 1025
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.862537064193513e-05,
      "loss": 0.5628,
      "step": 1026
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.862227275403614e-05,
      "loss": 0.3998,
      "step": 1027
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8619171637570468e-05,
      "loss": 0.6068,
      "step": 1028
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8616067293699316e-05,
      "loss": 0.6896,
      "step": 1029
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.861295972358509e-05,
      "loss": 0.6964,
      "step": 1030
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8609848928391398e-05,
      "loss": 0.646,
      "step": 1031
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8606734909283075e-05,
      "loss": 0.7398,
      "step": 1032
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8603617667426132e-05,
      "loss": 0.5592,
      "step": 1033
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.860049720398782e-05,
      "loss": 0.6808,
      "step": 1034
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8597373520136565e-05,
      "loss": 0.6682,
      "step": 1035
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8594246617042033e-05,
      "loss": 0.6005,
      "step": 1036
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8591116495875065e-05,
      "loss": 0.6298,
      "step": 1037
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.858798315780772e-05,
      "loss": 0.5783,
      "step": 1038
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.858484660401327e-05,
      "loss": 0.6213,
      "step": 1039
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8581706835666167e-05,
      "loss": 0.6614,
      "step": 1040
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8578563853942094e-05,
      "loss": 0.5615,
      "step": 1041
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8575417660017917e-05,
      "loss": 0.7991,
      "step": 1042
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8572268255071718e-05,
      "loss": 0.5536,
      "step": 1043
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.856911564028277e-05,
      "loss": 0.6453,
      "step": 1044
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8565959816831563e-05,
      "loss": 0.641,
      "step": 1045
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.856280078589977e-05,
      "loss": 0.6731,
      "step": 1046
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8559638548670276e-05,
      "loss": 0.4693,
      "step": 1047
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.855647310632717e-05,
      "loss": 0.838,
      "step": 1048
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.855330446005573e-05,
      "loss": 0.662,
      "step": 1049
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.855013261104244e-05,
      "loss": 0.5241,
      "step": 1050
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.854695756047499e-05,
      "loss": 0.7165,
      "step": 1051
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.854377930954225e-05,
      "loss": 0.6542,
      "step": 1052
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8540597859434314e-05,
      "loss": 0.6379,
      "step": 1053
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8537413211342447e-05,
      "loss": 0.5853,
      "step": 1054
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8534225366459133e-05,
      "loss": 0.6787,
      "step": 1055
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.853103432597804e-05,
      "loss": 0.5619,
      "step": 1056
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8527840091094038e-05,
      "loss": 0.5366,
      "step": 1057
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8524642663003195e-05,
      "loss": 0.546,
      "step": 1058
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8521442042902775e-05,
      "loss": 0.6336,
      "step": 1059
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.851823823199122e-05,
      "loss": 0.7357,
      "step": 1060
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.851503123146819e-05,
      "loss": 0.6964,
      "step": 1061
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8511821042534537e-05,
      "loss": 0.5135,
      "step": 1062
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8508607666392285e-05,
      "loss": 0.5448,
      "step": 1063
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8505391104244683e-05,
      "loss": 0.5008,
      "step": 1064
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8502171357296144e-05,
      "loss": 0.6437,
      "step": 1065
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.849894842675229e-05,
      "loss": 0.4832,
      "step": 1066
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.849572231381993e-05,
      "loss": 0.7402,
      "step": 1067
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.849249301970707e-05,
      "loss": 0.6081,
      "step": 1068
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.84892605456229e-05,
      "loss": 0.6185,
      "step": 1069
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8486024892777803e-05,
      "loss": 0.7121,
      "step": 1070
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8482786062383353e-05,
      "loss": 0.7054,
      "step": 1071
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8479544055652313e-05,
      "loss": 0.507,
      "step": 1072
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.847629887379864e-05,
      "loss": 0.6318,
      "step": 1073
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.847305051803747e-05,
      "loss": 0.6549,
      "step": 1074
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8469798989585145e-05,
      "loss": 0.5394,
      "step": 1075
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8466544289659166e-05,
      "loss": 0.6242,
      "step": 1076
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8463286419478256e-05,
      "loss": 0.5879,
      "step": 1077
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.84600253802623e-05,
      "loss": 0.6402,
      "step": 1078
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8456761173232376e-05,
      "loss": 0.7443,
      "step": 1079
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.845349379961075e-05,
      "loss": 0.7271,
      "step": 1080
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.845022326062088e-05,
      "loss": 0.6865,
      "step": 1081
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8446949557487398e-05,
      "loss": 0.6866,
      "step": 1082
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8443672691436127e-05,
      "loss": 0.605,
      "step": 1083
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.844039266369407e-05,
      "loss": 0.5895,
      "step": 1084
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.8437109475489417e-05,
      "loss": 0.6089,
      "step": 1085
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.843382312805155e-05,
      "loss": 0.7216,
      "step": 1086
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.843053362261102e-05,
      "loss": 0.527,
      "step": 1087
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.842724096039956e-05,
      "loss": 0.6968,
      "step": 1088
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8423945142650094e-05,
      "loss": 0.6241,
      "step": 1089
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.842064617059673e-05,
      "loss": 0.6152,
      "step": 1090
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8417344045474747e-05,
      "loss": 0.5978,
      "step": 1091
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.841403876852061e-05,
      "loss": 0.5955,
      "step": 1092
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.841073034097196e-05,
      "loss": 0.5824,
      "step": 1093
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8407418764067627e-05,
      "loss": 0.625,
      "step": 1094
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8404104039047612e-05,
      "loss": 0.5729,
      "step": 1095
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8400786167153092e-05,
      "loss": 0.7498,
      "step": 1096
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8397465149626438e-05,
      "loss": 0.8384,
      "step": 1097
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8394140987711178e-05,
      "loss": 0.5438,
      "step": 1098
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.839081368265203e-05,
      "loss": 0.6708,
      "step": 1099
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8387483235694893e-05,
      "loss": 0.587,
      "step": 1100
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8384149648086832e-05,
      "loss": 0.6099,
      "step": 1101
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8380812921076086e-05,
      "loss": 0.5625,
      "step": 1102
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8377473055912087e-05,
      "loss": 0.6705,
      "step": 1103
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8374130053845423e-05,
      "loss": 0.5287,
      "step": 1104
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8370783916127868e-05,
      "loss": 0.5022,
      "step": 1105
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8367434644012366e-05,
      "loss": 0.5353,
      "step": 1106
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.836408223875303e-05,
      "loss": 0.5345,
      "step": 1107
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.836072670160516e-05,
      "loss": 0.8513,
      "step": 1108
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8357368033825214e-05,
      "loss": 0.6044,
      "step": 1109
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8354006236670825e-05,
      "loss": 0.497,
      "step": 1110
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8350641311400813e-05,
      "loss": 0.6852,
      "step": 1111
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8347273259275144e-05,
      "loss": 0.6819,
      "step": 1112
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8343902081554976e-05,
      "loss": 0.6192,
      "step": 1113
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8340527779502628e-05,
      "loss": 0.6261,
      "step": 1114
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.833715035438159e-05,
      "loss": 0.6714,
      "step": 1115
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8333769807456524e-05,
      "loss": 0.6632,
      "step": 1116
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8330386139993253e-05,
      "loss": 0.6037,
      "step": 1117
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.832699935325878e-05,
      "loss": 0.7395,
      "step": 1118
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.832360944852127e-05,
      "loss": 0.6428,
      "step": 1119
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8320216427050053e-05,
      "loss": 0.5531,
      "step": 1120
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.831682029011563e-05,
      "loss": 0.7085,
      "step": 1121
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8313421038989666e-05,
      "loss": 0.7669,
      "step": 1122
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8310018674944995e-05,
      "loss": 0.5925,
      "step": 1123
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8306613199255618e-05,
      "loss": 0.5693,
      "step": 1124
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8303204613196694e-05,
      "loss": 0.6354,
      "step": 1125
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.829979291804455e-05,
      "loss": 0.7578,
      "step": 1126
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8296378115076683e-05,
      "loss": 0.6739,
      "step": 1127
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8292960205571742e-05,
      "loss": 0.6583,
      "step": 1128
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8289539190809552e-05,
      "loss": 0.5808,
      "step": 1129
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8286115072071095e-05,
      "loss": 0.5759,
      "step": 1130
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.828268785063851e-05,
      "loss": 0.4506,
      "step": 1131
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.827925752779511e-05,
      "loss": 0.5377,
      "step": 1132
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8275824104825353e-05,
      "loss": 0.5511,
      "step": 1133
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8272387583014874e-05,
      "loss": 0.8718,
      "step": 1134
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8268947963650457e-05,
      "loss": 0.9047,
      "step": 1135
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.8265505248020056e-05,
      "loss": 0.7269,
      "step": 1136
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.826205943741277e-05,
      "loss": 0.6187,
      "step": 1137
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8258610533118877e-05,
      "loss": 0.5083,
      "step": 1138
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.825515853642979e-05,
      "loss": 0.6838,
      "step": 1139
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.82517034486381e-05,
      "loss": 0.4996,
      "step": 1140
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8248245271037542e-05,
      "loss": 0.5337,
      "step": 1141
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8244784004923018e-05,
      "loss": 0.5244,
      "step": 1142
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8241319651590577e-05,
      "loss": 0.4718,
      "step": 1143
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.823785221233743e-05,
      "loss": 0.567,
      "step": 1144
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8234381688461943e-05,
      "loss": 0.6693,
      "step": 1145
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.823090808126364e-05,
      "loss": 0.5314,
      "step": 1146
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8227431392043188e-05,
      "loss": 0.6462,
      "step": 1147
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.822395162210242e-05,
      "loss": 0.5107,
      "step": 1148
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.822046877274432e-05,
      "loss": 0.5104,
      "step": 1149
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8216982845273022e-05,
      "loss": 0.5169,
      "step": 1150
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.821349384099381e-05,
      "loss": 0.6941,
      "step": 1151
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8210001761213134e-05,
      "loss": 0.7258,
      "step": 1152
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8206506607238573e-05,
      "loss": 0.6194,
      "step": 1153
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.820300838037888e-05,
      "loss": 0.619,
      "step": 1154
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8199507081943943e-05,
      "loss": 0.5923,
      "step": 1155
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.819600271324481e-05,
      "loss": 0.5896,
      "step": 1156
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8192495275593667e-05,
      "loss": 0.549,
      "step": 1157
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8188984770303866e-05,
      "loss": 0.5684,
      "step": 1158
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.818547119868989e-05,
      "loss": 0.6516,
      "step": 1159
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.818195456206738e-05,
      "loss": 0.6088,
      "step": 1160
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8178434861753128e-05,
      "loss": 0.7894,
      "step": 1161
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.817491209906506e-05,
      "loss": 0.4696,
      "step": 1162
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8171386275322264e-05,
      "loss": 0.5161,
      "step": 1163
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.816785739184496e-05,
      "loss": 0.4949,
      "step": 1164
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8164325449954533e-05,
      "loss": 0.6396,
      "step": 1165
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8160790450973484e-05,
      "loss": 0.588,
      "step": 1166
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8157252396225487e-05,
      "loss": 0.657,
      "step": 1167
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8153711287035348e-05,
      "loss": 0.4273,
      "step": 1168
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.815016712472901e-05,
      "loss": 0.7047,
      "step": 1169
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8146619910633577e-05,
      "loss": 0.6328,
      "step": 1170
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8143069646077282e-05,
      "loss": 0.6533,
      "step": 1171
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.81395163323895e-05,
      "loss": 0.524,
      "step": 1172
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8135959970900753e-05,
      "loss": 0.6528,
      "step": 1173
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8132400562942704e-05,
      "loss": 0.7485,
      "step": 1174
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8128838109848158e-05,
      "loss": 0.6489,
      "step": 1175
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.812527261295105e-05,
      "loss": 0.6087,
      "step": 1176
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.812170407358647e-05,
      "loss": 0.6178,
      "step": 1177
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.811813249309063e-05,
      "loss": 0.7929,
      "step": 1178
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8114557872800906e-05,
      "loss": 0.5257,
      "step": 1179
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.811098021405578e-05,
      "loss": 0.5456,
      "step": 1180
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8107399518194902e-05,
      "loss": 0.5629,
      "step": 1181
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.810381578655904e-05,
      "loss": 0.6053,
      "step": 1182
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.81002290204901e-05,
      "loss": 0.4679,
      "step": 1183
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.809663922133113e-05,
      "loss": 0.5863,
      "step": 1184
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.809304639042632e-05,
      "loss": 0.7561,
      "step": 1185
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.808945052912098e-05,
      "loss": 0.78,
      "step": 1186
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8085851638761564e-05,
      "loss": 0.5826,
      "step": 1187
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8082249720695657e-05,
      "loss": 0.5977,
      "step": 1188
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8078644776271985e-05,
      "loss": 0.6756,
      "step": 1189
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8075036806840388e-05,
      "loss": 0.557,
      "step": 1190
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8071425813751867e-05,
      "loss": 0.6103,
      "step": 1191
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8067811798358535e-05,
      "loss": 0.6093,
      "step": 1192
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.806419476201364e-05,
      "loss": 0.4291,
      "step": 1193
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8060574706071564e-05,
      "loss": 0.5303,
      "step": 1194
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.805695163188782e-05,
      "loss": 0.6524,
      "step": 1195
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8053325540819048e-05,
      "loss": 0.5429,
      "step": 1196
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8049696434223018e-05,
      "loss": 0.7551,
      "step": 1197
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8046064313458635e-05,
      "loss": 0.5625,
      "step": 1198
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8042429179885926e-05,
      "loss": 0.4973,
      "step": 1199
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.803879103486605e-05,
      "loss": 0.683,
      "step": 1200
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8035149879761294e-05,
      "loss": 0.5581,
      "step": 1201
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.803150571593506e-05,
      "loss": 0.575,
      "step": 1202
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.80278585447519e-05,
      "loss": 0.6254,
      "step": 1203
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.802420836757747e-05,
      "loss": 0.5094,
      "step": 1204
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8020555185778566e-05,
      "loss": 0.6753,
      "step": 1205
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.80168990007231e-05,
      "loss": 0.7235,
      "step": 1206
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.801323981378011e-05,
      "loss": 0.6079,
      "step": 1207
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8009577626319768e-05,
      "loss": 0.6635,
      "step": 1208
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8005912439713352e-05,
      "loss": 0.6443,
      "step": 1209
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.800224425533328e-05,
      "loss": 0.5312,
      "step": 1210
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7998573074553086e-05,
      "loss": 0.7175,
      "step": 1211
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.799489889874742e-05,
      "loss": 0.6706,
      "step": 1212
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.799122172929206e-05,
      "loss": 0.6261,
      "step": 1213
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7987541567563905e-05,
      "loss": 0.7684,
      "step": 1214
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.798385841494097e-05,
      "loss": 0.5973,
      "step": 1215
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7980172272802398e-05,
      "loss": 0.514,
      "step": 1216
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.797648314252844e-05,
      "loss": 0.7001,
      "step": 1217
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.797279102550048e-05,
      "loss": 0.5902,
      "step": 1218
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.796909592310101e-05,
      "loss": 0.6479,
      "step": 1219
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7965397836713637e-05,
      "loss": 0.7434,
      "step": 1220
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7961696767723098e-05,
      "loss": 0.6319,
      "step": 1221
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7957992717515234e-05,
      "loss": 0.4695,
      "step": 1222
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7954285687477013e-05,
      "loss": 0.5784,
      "step": 1223
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.795057567899651e-05,
      "loss": 0.5587,
      "step": 1224
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.794686269346292e-05,
      "loss": 0.6629,
      "step": 1225
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7943146732266553e-05,
      "loss": 0.6403,
      "step": 1226
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7939427796798835e-05,
      "loss": 0.7139,
      "step": 1227
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7935705888452295e-05,
      "loss": 0.6358,
      "step": 1228
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7931981008620588e-05,
      "loss": 0.4881,
      "step": 1229
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7928253158698474e-05,
      "loss": 0.8216,
      "step": 1230
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.792452234008183e-05,
      "loss": 0.5825,
      "step": 1231
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.792078855416764e-05,
      "loss": 0.6072,
      "step": 1232
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7917051802354003e-05,
      "loss": 0.6582,
      "step": 1233
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.791331208604013e-05,
      "loss": 0.727,
      "step": 1234
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7909569406626332e-05,
      "loss": 0.502,
      "step": 1235
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.790582376551404e-05,
      "loss": 0.5612,
      "step": 1236
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.790207516410579e-05,
      "loss": 0.6181,
      "step": 1237
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7898323603805228e-05,
      "loss": 0.6488,
      "step": 1238
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.789456908601711e-05,
      "loss": 0.5395,
      "step": 1239
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7890811612147287e-05,
      "loss": 0.6086,
      "step": 1240
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7887051183602736e-05,
      "loss": 0.6238,
      "step": 1241
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.788328780179152e-05,
      "loss": 0.6923,
      "step": 1242
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7879521468122833e-05,
      "loss": 0.7086,
      "step": 1243
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7875752184006945e-05,
      "loss": 0.6572,
      "step": 1244
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7871979950855256e-05,
      "loss": 0.5991,
      "step": 1245
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7868204770080252e-05,
      "loss": 0.6048,
      "step": 1246
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7864426643095537e-05,
      "loss": 0.6229,
      "step": 1247
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.786064557131581e-05,
      "loss": 0.73,
      "step": 1248
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.785686155615687e-05,
      "loss": 0.475,
      "step": 1249
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7853074599035628e-05,
      "loss": 0.5951,
      "step": 1250
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.784928470137009e-05,
      "loss": 0.5214,
      "step": 1251
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.784549186457936e-05,
      "loss": 0.5244,
      "step": 1252
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.784169609008365e-05,
      "loss": 0.6403,
      "step": 1253
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7837897379304267e-05,
      "loss": 0.7071,
      "step": 1254
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7834095733663624e-05,
      "loss": 0.5415,
      "step": 1255
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.783029115458522e-05,
      "loss": 0.6182,
      "step": 1256
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7826483643493664e-05,
      "loss": 0.6117,
      "step": 1257
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7822673201814664e-05,
      "loss": 0.5071,
      "step": 1258
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7818859830975016e-05,
      "loss": 0.5654,
      "step": 1259
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7815043532402616e-05,
      "loss": 0.6708,
      "step": 1260
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7811224307526464e-05,
      "loss": 0.6146,
      "step": 1261
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7807402157776645e-05,
      "loss": 0.5851,
      "step": 1262
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.780357708458434e-05,
      "loss": 0.5154,
      "step": 1263
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7799749089381843e-05,
      "loss": 0.5821,
      "step": 1264
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.779591817360251e-05,
      "loss": 0.6596,
      "step": 1265
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7792084338680816e-05,
      "loss": 0.4192,
      "step": 1266
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7788247586052324e-05,
      "loss": 0.519,
      "step": 1267
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.778440791715368e-05,
      "loss": 0.5634,
      "step": 1268
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7780565333422636e-05,
      "loss": 0.5947,
      "step": 1269
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.777671983629802e-05,
      "loss": 0.523,
      "step": 1270
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7772871427219767e-05,
      "loss": 0.7015,
      "step": 1271
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7769020107628892e-05,
      "loss": 0.5818,
      "step": 1272
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7765165878967497e-05,
      "loss": 0.5754,
      "step": 1273
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7761308742678787e-05,
      "loss": 0.5656,
      "step": 1274
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.775744870020704e-05,
      "loss": 0.7581,
      "step": 1275
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7753585752997638e-05,
      "loss": 0.6464,
      "step": 1276
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.774971990249703e-05,
      "loss": 0.7446,
      "step": 1277
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7745851150152777e-05,
      "loss": 0.5266,
      "step": 1278
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7741979497413505e-05,
      "loss": 0.5522,
      "step": 1279
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7738104945728935e-05,
      "loss": 0.6225,
      "step": 1280
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.773422749654988e-05,
      "loss": 0.8063,
      "step": 1281
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.773034715132823e-05,
      "loss": 0.6506,
      "step": 1282
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.772646391151695e-05,
      "loss": 0.6625,
      "step": 1283
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7722577778570116e-05,
      "loss": 0.5325,
      "step": 1284
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.771868875394286e-05,
      "loss": 0.7432,
      "step": 1285
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7714796839091408e-05,
      "loss": 0.6533,
      "step": 1286
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7710902035473075e-05,
      "loss": 0.7234,
      "step": 1287
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7707004344546246e-05,
      "loss": 0.554,
      "step": 1288
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7703103767770387e-05,
      "loss": 0.4693,
      "step": 1289
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.769920030660606e-05,
      "loss": 0.5559,
      "step": 1290
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7695293962514886e-05,
      "loss": 0.5746,
      "step": 1291
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7691384736959584e-05,
      "loss": 0.5855,
      "step": 1292
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.768747263140394e-05,
      "loss": 0.4356,
      "step": 1293
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.768355764731282e-05,
      "loss": 0.6465,
      "step": 1294
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.767963978615218e-05,
      "loss": 0.6538,
      "step": 1295
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.7675719049389027e-05,
      "loss": 0.6491,
      "step": 1296
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7671795438491476e-05,
      "loss": 0.6001,
      "step": 1297
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7667868954928695e-05,
      "loss": 0.5571,
      "step": 1298
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.766393960017094e-05,
      "loss": 0.4638,
      "step": 1299
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7660007375689537e-05,
      "loss": 0.7924,
      "step": 1300
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7656072282956887e-05,
      "loss": 0.5272,
      "step": 1301
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.765213432344647e-05,
      "loss": 0.5429,
      "step": 1302
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7648193498632828e-05,
      "loss": 0.5592,
      "step": 1303
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.764424980999159e-05,
      "loss": 0.6348,
      "step": 1304
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7640303258999442e-05,
      "loss": 0.6176,
      "step": 1305
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.763635384713416e-05,
      "loss": 0.5416,
      "step": 1306
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.763240157587457e-05,
      "loss": 0.7743,
      "step": 1307
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.762844644670059e-05,
      "loss": 0.6068,
      "step": 1308
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7624488461093197e-05,
      "loss": 0.4343,
      "step": 1309
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7620527620534432e-05,
      "loss": 0.597,
      "step": 1310
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.761656392650742e-05,
      "loss": 0.6763,
      "step": 1311
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7612597380496344e-05,
      "loss": 0.6208,
      "step": 1312
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7608627983986452e-05,
      "loss": 0.5105,
      "step": 1313
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.760465573846407e-05,
      "loss": 0.782,
      "step": 1314
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7600680645416583e-05,
      "loss": 0.7729,
      "step": 1315
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7596702706332446e-05,
      "loss": 0.5772,
      "step": 1316
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.759272192270118e-05,
      "loss": 0.6116,
      "step": 1317
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7588738296013367e-05,
      "loss": 0.5852,
      "step": 1318
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7584751827760655e-05,
      "loss": 0.4363,
      "step": 1319
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7580762519435757e-05,
      "loss": 0.6878,
      "step": 1320
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.757677037253245e-05,
      "loss": 0.5244,
      "step": 1321
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7572775388545577e-05,
      "loss": 0.6731,
      "step": 1322
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7568777568971035e-05,
      "loss": 0.6155,
      "step": 1323
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.756477691530579e-05,
      "loss": 0.5429,
      "step": 1324
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7560773429047862e-05,
      "loss": 0.637,
      "step": 1325
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.755676711169635e-05,
      "loss": 0.4831,
      "step": 1326
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7552757964751375e-05,
      "loss": 0.8057,
      "step": 1327
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7548745989714167e-05,
      "loss": 0.7047,
      "step": 1328
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7544731188086976e-05,
      "loss": 0.5902,
      "step": 1329
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7540713561373127e-05,
      "loss": 0.6175,
      "step": 1330
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7536693111077002e-05,
      "loss": 0.4928,
      "step": 1331
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7532669838704036e-05,
      "loss": 0.7044,
      "step": 1332
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7528643745760724e-05,
      "loss": 0.6148,
      "step": 1333
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7524614833754615e-05,
      "loss": 0.5079,
      "step": 1334
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.752058310419432e-05,
      "loss": 0.5868,
      "step": 1335
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.751654855858949e-05,
      "loss": 0.6933,
      "step": 1336
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.751251119845085e-05,
      "loss": 0.4903,
      "step": 1337
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.750847102529017e-05,
      "loss": 0.7062,
      "step": 1338
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.750442804062026e-05,
      "loss": 0.357,
      "step": 1339
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7500382245955007e-05,
      "loss": 0.6088,
      "step": 1340
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7496333642809337e-05,
      "loss": 0.7182,
      "step": 1341
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.749228223269922e-05,
      "loss": 0.6369,
      "step": 1342
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.74882280171417e-05,
      "loss": 0.5228,
      "step": 1343
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.748417099765485e-05,
      "loss": 0.6296,
      "step": 1344
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.74801111757578e-05,
      "loss": 0.567,
      "step": 1345
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.747604855297073e-05,
      "loss": 0.4983,
      "step": 1346
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.7471983130814872e-05,
      "loss": 0.7389,
      "step": 1347
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.74679149108125e-05,
      "loss": 0.5638,
      "step": 1348
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.746384389448694e-05,
      "loss": 0.6807,
      "step": 1349
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.745977008336256e-05,
      "loss": 0.6734,
      "step": 1350
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7455693478964782e-05,
      "loss": 0.6912,
      "step": 1351
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.745161408282007e-05,
      "loss": 0.5636,
      "step": 1352
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.744753189645593e-05,
      "loss": 0.4027,
      "step": 1353
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7443446921400924e-05,
      "loss": 0.5479,
      "step": 1354
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.743935915918464e-05,
      "loss": 0.679,
      "step": 1355
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7435268611337727e-05,
      "loss": 0.5915,
      "step": 1356
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7431175279391864e-05,
      "loss": 0.5409,
      "step": 1357
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.742707916487979e-05,
      "loss": 0.6069,
      "step": 1358
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7422980269335262e-05,
      "loss": 0.5858,
      "step": 1359
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7418878594293095e-05,
      "loss": 0.6636,
      "step": 1360
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.741477414128914e-05,
      "loss": 0.6452,
      "step": 1361
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7410666911860298e-05,
      "loss": 0.684,
      "step": 1362
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7406556907544486e-05,
      "loss": 0.3533,
      "step": 1363
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7402444129880683e-05,
      "loss": 0.5252,
      "step": 1364
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7398328580408896e-05,
      "loss": 0.5956,
      "step": 1365
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.739421026067017e-05,
      "loss": 0.5251,
      "step": 1366
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7390089172206594e-05,
      "loss": 0.4391,
      "step": 1367
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7385965316561285e-05,
      "loss": 0.6689,
      "step": 1368
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.73818386952784e-05,
      "loss": 0.5935,
      "step": 1369
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.737770930990313e-05,
      "loss": 0.681,
      "step": 1370
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7373577161981708e-05,
      "loss": 0.6445,
      "step": 1371
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.736944225306139e-05,
      "loss": 0.5348,
      "step": 1372
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7365304584690478e-05,
      "loss": 0.7073,
      "step": 1373
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7361164158418296e-05,
      "loss": 0.6518,
      "step": 1374
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7357020975795208e-05,
      "loss": 0.6345,
      "step": 1375
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.735287503837261e-05,
      "loss": 0.6038,
      "step": 1376
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7348726347702922e-05,
      "loss": 0.6215,
      "step": 1377
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7344574905339608e-05,
      "loss": 0.5527,
      "step": 1378
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7340420712837144e-05,
      "loss": 0.6355,
      "step": 1379
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7336263771751055e-05,
      "loss": 0.4508,
      "step": 1380
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7332104083637887e-05,
      "loss": 0.529,
      "step": 1381
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7327941650055208e-05,
      "loss": 0.6833,
      "step": 1382
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7323776472561625e-05,
      "loss": 0.4829,
      "step": 1383
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.731960855271677e-05,
      "loss": 0.5169,
      "step": 1384
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.731543789208129e-05,
      "loss": 0.589,
      "step": 1385
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.731126449221688e-05,
      "loss": 0.5681,
      "step": 1386
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.730708835468624e-05,
      "loss": 0.4832,
      "step": 1387
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7302909481053106e-05,
      "loss": 0.6413,
      "step": 1388
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.729872787288224e-05,
      "loss": 0.7852,
      "step": 1389
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.729454353173942e-05,
      "loss": 0.6015,
      "step": 1390
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7290356459191447e-05,
      "loss": 0.7533,
      "step": 1391
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7286166656806157e-05,
      "loss": 0.5159,
      "step": 1392
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7281974126152402e-05,
      "loss": 0.6834,
      "step": 1393
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.727777886880005e-05,
      "loss": 0.7594,
      "step": 1394
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.727358088631999e-05,
      "loss": 0.7001,
      "step": 1395
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7269380180284142e-05,
      "loss": 0.493,
      "step": 1396
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7265176752265437e-05,
      "loss": 0.5517,
      "step": 1397
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.726097060383783e-05,
      "loss": 0.5341,
      "step": 1398
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7256761736576284e-05,
      "loss": 0.5668,
      "step": 1399
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7252550152056795e-05,
      "loss": 0.6706,
      "step": 1400
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7248335851856367e-05,
      "loss": 0.5381,
      "step": 1401
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.7244118837553028e-05,
      "loss": 0.6714,
      "step": 1402
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7239899110725815e-05,
      "loss": 0.5611,
      "step": 1403
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.723567667295478e-05,
      "loss": 0.6804,
      "step": 1404
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7231451525821e-05,
      "loss": 0.6177,
      "step": 1405
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7227223670906554e-05,
      "loss": 0.641,
      "step": 1406
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7222993109794547e-05,
      "loss": 0.5489,
      "step": 1407
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7218759844069084e-05,
      "loss": 0.4403,
      "step": 1408
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7214523875315298e-05,
      "loss": 0.688,
      "step": 1409
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.721028520511932e-05,
      "loss": 0.5057,
      "step": 1410
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7206043835068304e-05,
      "loss": 0.4988,
      "step": 1411
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7201799766750405e-05,
      "loss": 0.8098,
      "step": 1412
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7197553001754796e-05,
      "loss": 0.7501,
      "step": 1413
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7193303541671652e-05,
      "loss": 0.7061,
      "step": 1414
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7189051388092167e-05,
      "loss": 0.6106,
      "step": 1415
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7184796542608538e-05,
      "loss": 0.5511,
      "step": 1416
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7180539006813973e-05,
      "loss": 0.5549,
      "step": 1417
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.717627878230268e-05,
      "loss": 0.5948,
      "step": 1418
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7172015870669875e-05,
      "loss": 0.583,
      "step": 1419
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7167750273511796e-05,
      "loss": 0.5884,
      "step": 1420
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7163481992425662e-05,
      "loss": 0.5537,
      "step": 1421
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7159211029009716e-05,
      "loss": 0.5934,
      "step": 1422
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.71549373848632e-05,
      "loss": 0.5111,
      "step": 1423
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7150661061586354e-05,
      "loss": 0.5578,
      "step": 1424
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7146382060780428e-05,
      "loss": 0.6121,
      "step": 1425
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7142100384047672e-05,
      "loss": 0.6706,
      "step": 1426
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7137816032991338e-05,
      "loss": 0.7424,
      "step": 1427
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7133529009215685e-05,
      "loss": 0.723,
      "step": 1428
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.712923931432596e-05,
      "loss": 0.5684,
      "step": 1429
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7124946949928418e-05,
      "loss": 0.552,
      "step": 1430
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7120651917630317e-05,
      "loss": 0.4514,
      "step": 1431
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.711635421903991e-05,
      "loss": 0.5139,
      "step": 1432
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7112053855766446e-05,
      "loss": 0.5619,
      "step": 1433
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7107750829420177e-05,
      "loss": 0.6745,
      "step": 1434
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7103445141612348e-05,
      "loss": 0.6543,
      "step": 1435
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7099136793955205e-05,
      "loss": 0.6353,
      "step": 1436
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7094825788061984e-05,
      "loss": 0.5393,
      "step": 1437
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.709051212554692e-05,
      "loss": 0.774,
      "step": 1438
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7086195808025242e-05,
      "loss": 0.5575,
      "step": 1439
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7081876837113173e-05,
      "loss": 0.5133,
      "step": 1440
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7077555214427933e-05,
      "loss": 0.6224,
      "step": 1441
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.707323094158773e-05,
      "loss": 0.5575,
      "step": 1442
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7068904020211764e-05,
      "loss": 0.641,
      "step": 1443
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.706457445192023e-05,
      "loss": 0.5687,
      "step": 1444
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7060242238334317e-05,
      "loss": 0.6552,
      "step": 1445
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7055907381076196e-05,
      "loss": 0.6392,
      "step": 1446
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7051569881769033e-05,
      "loss": 0.6018,
      "step": 1447
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7047229742036987e-05,
      "loss": 0.5588,
      "step": 1448
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7042886963505196e-05,
      "loss": 0.655,
      "step": 1449
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7038541547799797e-05,
      "loss": 0.6132,
      "step": 1450
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7034193496547903e-05,
      "loss": 0.5211,
      "step": 1451
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.7029842811377623e-05,
      "loss": 0.5417,
      "step": 1452
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.702548949391805e-05,
      "loss": 0.5748,
      "step": 1453
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.702113354579926e-05,
      "loss": 0.5473,
      "step": 1454
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.701677496865232e-05,
      "loss": 0.6001,
      "step": 1455
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.7012413764109274e-05,
      "loss": 0.5881,
      "step": 1456
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.7008049933803153e-05,
      "loss": 0.5174,
      "step": 1457
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.700368347936797e-05,
      "loss": 0.5429,
      "step": 1458
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6999314402438726e-05,
      "loss": 0.6169,
      "step": 1459
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.699494270465139e-05,
      "loss": 0.561,
      "step": 1460
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6990568387642936e-05,
      "loss": 0.4459,
      "step": 1461
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.69861914530513e-05,
      "loss": 0.589,
      "step": 1462
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.69818119025154e-05,
      "loss": 0.6018,
      "step": 1463
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6977429737675136e-05,
      "loss": 0.6237,
      "step": 1464
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6973044960171394e-05,
      "loss": 0.544,
      "step": 1465
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6968657571646027e-05,
      "loss": 0.5871,
      "step": 1466
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.696426757374187e-05,
      "loss": 0.519,
      "step": 1467
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6959874968102736e-05,
      "loss": 0.583,
      "step": 1468
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6955479756373415e-05,
      "loss": 0.6231,
      "step": 1469
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6951081940199676e-05,
      "loss": 0.6945,
      "step": 1470
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6946681521228256e-05,
      "loss": 0.5252,
      "step": 1471
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6942278501106865e-05,
      "loss": 0.7087,
      "step": 1472
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.69378728814842e-05,
      "loss": 0.5507,
      "step": 1473
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6933464664009915e-05,
      "loss": 0.5233,
      "step": 1474
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6929053850334653e-05,
      "loss": 0.5241,
      "step": 1475
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6924640442110016e-05,
      "loss": 0.6595,
      "step": 1476
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6920224440988578e-05,
      "loss": 0.5011,
      "step": 1477
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.69158058486239e-05,
      "loss": 0.5046,
      "step": 1478
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6911384666670493e-05,
      "loss": 0.7385,
      "step": 1479
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6906960896783847e-05,
      "loss": 0.6148,
      "step": 1480
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.690253454062043e-05,
      "loss": 0.7022,
      "step": 1481
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6898105599837654e-05,
      "loss": 0.7872,
      "step": 1482
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6893674076093923e-05,
      "loss": 0.5984,
      "step": 1483
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6889239971048594e-05,
      "loss": 0.508,
      "step": 1484
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6884803286362e-05,
      "loss": 0.735,
      "step": 1485
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6880364023695435e-05,
      "loss": 0.6937,
      "step": 1486
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6875922184711152e-05,
      "loss": 0.6188,
      "step": 1487
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.687147777107238e-05,
      "loss": 0.614,
      "step": 1488
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6867030784443313e-05,
      "loss": 0.5874,
      "step": 1489
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6862581226489093e-05,
      "loss": 0.5484,
      "step": 1490
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.685812909887584e-05,
      "loss": 0.5786,
      "step": 1491
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6853674403270628e-05,
      "loss": 0.6042,
      "step": 1492
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.68492171413415e-05,
      "loss": 0.6611,
      "step": 1493
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.684475731475745e-05,
      "loss": 0.5814,
      "step": 1494
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6840294925188436e-05,
      "loss": 0.7758,
      "step": 1495
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6835829974305385e-05,
      "loss": 0.5991,
      "step": 1496
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6831362463780173e-05,
      "loss": 0.557,
      "step": 1497
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6826892395285632e-05,
      "loss": 0.5493,
      "step": 1498
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.682241977049556e-05,
      "loss": 0.5846,
      "step": 1499
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.681794459108471e-05,
      "loss": 0.633,
      "step": 1500
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.681346685872879e-05,
      "loss": 0.6315,
      "step": 1501
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6808986575104464e-05,
      "loss": 0.6774,
      "step": 1502
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.680450374188935e-05,
      "loss": 0.6875,
      "step": 1503
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6800018360762026e-05,
      "loss": 0.6968,
      "step": 1504
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6795530433402016e-05,
      "loss": 0.6012,
      "step": 1505
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6791039961489805e-05,
      "loss": 0.4998,
      "step": 1506
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6786546946706826e-05,
      "loss": 0.5965,
      "step": 1507
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.6782051390735468e-05,
      "loss": 0.6145,
      "step": 1508
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6777553295259065e-05,
      "loss": 0.6323,
      "step": 1509
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6773052661961912e-05,
      "loss": 0.5479,
      "step": 1510
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6768549492529243e-05,
      "loss": 0.6911,
      "step": 1511
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6764043788647253e-05,
      "loss": 0.7565,
      "step": 1512
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.675953555200307e-05,
      "loss": 0.8648,
      "step": 1513
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6755024784284794e-05,
      "loss": 0.4231,
      "step": 1514
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.675051148718145e-05,
      "loss": 0.4991,
      "step": 1515
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6745995662383022e-05,
      "loss": 0.6622,
      "step": 1516
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6741477311580442e-05,
      "loss": 0.5921,
      "step": 1517
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6736956436465573e-05,
      "loss": 0.5665,
      "step": 1518
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6732433038731245e-05,
      "loss": 0.4669,
      "step": 1519
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6727907120071215e-05,
      "loss": 0.548,
      "step": 1520
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6723378682180193e-05,
      "loss": 0.6628,
      "step": 1521
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.671884772675383e-05,
      "loss": 0.6679,
      "step": 1522
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.671431425548872e-05,
      "loss": 0.4911,
      "step": 1523
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6709778270082395e-05,
      "loss": 0.7051,
      "step": 1524
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.670523977223334e-05,
      "loss": 0.5582,
      "step": 1525
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.670069876364096e-05,
      "loss": 0.4785,
      "step": 1526
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.669615524600562e-05,
      "loss": 0.5972,
      "step": 1527
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.669160922102862e-05,
      "loss": 0.7963,
      "step": 1528
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6687060690412194e-05,
      "loss": 0.5964,
      "step": 1529
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6682509655859515e-05,
      "loss": 0.5174,
      "step": 1530
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6677956119074697e-05,
      "loss": 0.7178,
      "step": 1531
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6673400081762785e-05,
      "loss": 0.4297,
      "step": 1532
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.666884154562977e-05,
      "loss": 0.4984,
      "step": 1533
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6664280512382567e-05,
      "loss": 0.4618,
      "step": 1534
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6659716983729038e-05,
      "loss": 0.5059,
      "step": 1535
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.665515096137797e-05,
      "loss": 0.4335,
      "step": 1536
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6650582447039087e-05,
      "loss": 0.5871,
      "step": 1537
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6646011442423045e-05,
      "loss": 0.6818,
      "step": 1538
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6641437949241436e-05,
      "loss": 0.4751,
      "step": 1539
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.663686196920678e-05,
      "loss": 0.6192,
      "step": 1540
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6632283504032532e-05,
      "loss": 0.6461,
      "step": 1541
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6627702555433075e-05,
      "loss": 0.6176,
      "step": 1542
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.662311912512372e-05,
      "loss": 0.5292,
      "step": 1543
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6618533214820712e-05,
      "loss": 0.6219,
      "step": 1544
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6613944826241216e-05,
      "loss": 0.5155,
      "step": 1545
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.660935396110334e-05,
      "loss": 0.6983,
      "step": 1546
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6604760621126104e-05,
      "loss": 0.5451,
      "step": 1547
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.660016480802946e-05,
      "loss": 0.7285,
      "step": 1548
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6595566523534298e-05,
      "loss": 0.5851,
      "step": 1549
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.659096576936241e-05,
      "loss": 0.5391,
      "step": 1550
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.658636254723653e-05,
      "loss": 0.5732,
      "step": 1551
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.658175685888031e-05,
      "loss": 0.6,
      "step": 1552
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.657714870601833e-05,
      "loss": 0.4738,
      "step": 1553
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6572538090376086e-05,
      "loss": 0.5701,
      "step": 1554
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6567925013680005e-05,
      "loss": 0.7767,
      "step": 1555
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.656330947765742e-05,
      "loss": 0.5152,
      "step": 1556
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.655869148403661e-05,
      "loss": 0.5495,
      "step": 1557
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6554071034546746e-05,
      "loss": 0.4465,
      "step": 1558
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.654944813091794e-05,
      "loss": 0.625,
      "step": 1559
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6544822774881212e-05,
      "loss": 0.588,
      "step": 1560
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.6540194968168506e-05,
      "loss": 0.5481,
      "step": 1561
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.653556471251268e-05,
      "loss": 0.4862,
      "step": 1562
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6530932009647508e-05,
      "loss": 0.5021,
      "step": 1563
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.652629686130768e-05,
      "loss": 0.6206,
      "step": 1564
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6521659269228815e-05,
      "loss": 0.4767,
      "step": 1565
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.651701923514742e-05,
      "loss": 0.6299,
      "step": 1566
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6512376760800943e-05,
      "loss": 0.6486,
      "step": 1567
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6507731847927735e-05,
      "loss": 0.5918,
      "step": 1568
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6503084498267058e-05,
      "loss": 0.7229,
      "step": 1569
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.649843471355909e-05,
      "loss": 0.6286,
      "step": 1570
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6493782495544915e-05,
      "loss": 0.6434,
      "step": 1571
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6489127845966538e-05,
      "loss": 0.6164,
      "step": 1572
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6484470766566866e-05,
      "loss": 0.6057,
      "step": 1573
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.647981125908972e-05,
      "loss": 0.4154,
      "step": 1574
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.647514932527983e-05,
      "loss": 0.6895,
      "step": 1575
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6470484966882834e-05,
      "loss": 0.6217,
      "step": 1576
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.646581818564528e-05,
      "loss": 0.4864,
      "step": 1577
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.646114898331461e-05,
      "loss": 0.5457,
      "step": 1578
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6456477361639193e-05,
      "loss": 0.508,
      "step": 1579
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.645180332236829e-05,
      "loss": 0.6799,
      "step": 1580
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6447126867252083e-05,
      "loss": 0.6916,
      "step": 1581
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6442447998041628e-05,
      "loss": 0.5367,
      "step": 1582
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6437766716488918e-05,
      "loss": 0.5703,
      "step": 1583
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6433083024346827e-05,
      "loss": 0.6,
      "step": 1584
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6428396923369145e-05,
      "loss": 0.6006,
      "step": 1585
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6423708415310563e-05,
      "loss": 0.6146,
      "step": 1586
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.641901750192666e-05,
      "loss": 0.5474,
      "step": 1587
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6414324184973925e-05,
      "loss": 0.5987,
      "step": 1588
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6409628466209756e-05,
      "loss": 0.5569,
      "step": 1589
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.640493034739244e-05,
      "loss": 0.6291,
      "step": 1590
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.640022983028115e-05,
      "loss": 0.6844,
      "step": 1591
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.639552691663599e-05,
      "loss": 0.6538,
      "step": 1592
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.639082160821793e-05,
      "loss": 0.6291,
      "step": 1593
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6386113906788853e-05,
      "loss": 0.6467,
      "step": 1594
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.638140381411153e-05,
      "loss": 0.6628,
      "step": 1595
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6376691331949638e-05,
      "loss": 0.6086,
      "step": 1596
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6371976462067744e-05,
      "loss": 0.707,
      "step": 1597
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6367259206231302e-05,
      "loss": 0.5518,
      "step": 1598
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.636253956620666e-05,
      "loss": 0.5365,
      "step": 1599
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6357817543761072e-05,
      "loss": 0.564,
      "step": 1600
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6353093140662674e-05,
      "loss": 0.5614,
      "step": 1601
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.634836635868049e-05,
      "loss": 0.8269,
      "step": 1602
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6343637199584447e-05,
      "loss": 0.4994,
      "step": 1603
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6338905665145352e-05,
      "loss": 0.5822,
      "step": 1604
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.63341717571349e-05,
      "loss": 0.5145,
      "step": 1605
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6329435477325684e-05,
      "loss": 0.7312,
      "step": 1606
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6324696827491178e-05,
      "loss": 0.5878,
      "step": 1607
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6319955809405748e-05,
      "loss": 0.5483,
      "step": 1608
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.631521242484464e-05,
      "loss": 0.4618,
      "step": 1609
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6310466675583995e-05,
      "loss": 0.5631,
      "step": 1610
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6305718563400836e-05,
      "loss": 0.6718,
      "step": 1611
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.630096809007306e-05,
      "loss": 0.5174,
      "step": 1612
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.629621525737947e-05,
      "loss": 0.573,
      "step": 1613
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.6291460067099733e-05,
      "loss": 0.6085,
      "step": 1614
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6286702521014408e-05,
      "loss": 0.5033,
      "step": 1615
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6281942620904934e-05,
      "loss": 0.519,
      "step": 1616
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6277180368553637e-05,
      "loss": 0.6358,
      "step": 1617
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6272415765743708e-05,
      "loss": 0.6607,
      "step": 1618
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6267648814259235e-05,
      "loss": 0.5621,
      "step": 1619
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.626287951588518e-05,
      "loss": 0.7471,
      "step": 1620
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6258107872407376e-05,
      "loss": 0.6529,
      "step": 1621
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6253333885612553e-05,
      "loss": 0.5798,
      "step": 1622
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6248557557288293e-05,
      "loss": 0.8362,
      "step": 1623
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6243778889223074e-05,
      "loss": 0.5901,
      "step": 1624
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6238997883206248e-05,
      "loss": 0.6155,
      "step": 1625
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6234214541028034e-05,
      "loss": 0.569,
      "step": 1626
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.622942886447953e-05,
      "loss": 0.6978,
      "step": 1627
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6224640855352715e-05,
      "loss": 0.5421,
      "step": 1628
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6219850515440426e-05,
      "loss": 0.5015,
      "step": 1629
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6215057846536385e-05,
      "loss": 0.5267,
      "step": 1630
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6210262850435183e-05,
      "loss": 0.5465,
      "step": 1631
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6205465528932286e-05,
      "loss": 0.5682,
      "step": 1632
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.620066588382402e-05,
      "loss": 0.5096,
      "step": 1633
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6195863916907596e-05,
      "loss": 0.5753,
      "step": 1634
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.619105962998108e-05,
      "loss": 0.7451,
      "step": 1635
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.618625302484341e-05,
      "loss": 0.4534,
      "step": 1636
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6181444103294405e-05,
      "loss": 0.6759,
      "step": 1637
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6176632867134738e-05,
      "loss": 0.5178,
      "step": 1638
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6171819318165946e-05,
      "loss": 0.6596,
      "step": 1639
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6167003458190446e-05,
      "loss": 0.5649,
      "step": 1640
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.616218528901151e-05,
      "loss": 0.4778,
      "step": 1641
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6157364812433275e-05,
      "loss": 0.7073,
      "step": 1642
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6152542030260747e-05,
      "loss": 0.5753,
      "step": 1643
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6147716944299792e-05,
      "loss": 0.5643,
      "step": 1644
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6142889556357136e-05,
      "loss": 0.508,
      "step": 1645
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6138059868240377e-05,
      "loss": 0.6897,
      "step": 1646
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.613322788175796e-05,
      "loss": 0.6888,
      "step": 1647
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.61283935987192e-05,
      "loss": 0.8056,
      "step": 1648
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6123557020934275e-05,
      "loss": 0.4933,
      "step": 1649
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6118718150214213e-05,
      "loss": 0.6334,
      "step": 1650
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6113876988370907e-05,
      "loss": 0.6112,
      "step": 1651
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.61090335372171e-05,
      "loss": 0.5884,
      "step": 1652
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6104187798566405e-05,
      "loss": 0.7712,
      "step": 1653
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6099339774233274e-05,
      "loss": 0.7958,
      "step": 1654
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.609448946603304e-05,
      "loss": 0.559,
      "step": 1655
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.608963687578187e-05,
      "loss": 0.6058,
      "step": 1656
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.608478200529679e-05,
      "loss": 0.4981,
      "step": 1657
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.607992485639568e-05,
      "loss": 0.619,
      "step": 1658
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6075065430897284e-05,
      "loss": 0.506,
      "step": 1659
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.607020373062118e-05,
      "loss": 0.4286,
      "step": 1660
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6065339757387812e-05,
      "loss": 0.5259,
      "step": 1661
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6060473513018467e-05,
      "loss": 0.5578,
      "step": 1662
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.605560499933529e-05,
      "loss": 0.6804,
      "step": 1663
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6050734218161265e-05,
      "loss": 0.4614,
      "step": 1664
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6045861171320242e-05,
      "loss": 0.4903,
      "step": 1665
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6040985860636896e-05,
      "loss": 0.7114,
      "step": 1666
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6036108287936774e-05,
      "loss": 0.6392,
      "step": 1667
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.6031228455046255e-05,
      "loss": 0.5792,
      "step": 1668
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.6026346363792565e-05,
      "loss": 0.5584,
      "step": 1669
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.6021462016003778e-05,
      "loss": 0.6968,
      "step": 1670
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.601657541350882e-05,
      "loss": 0.4564,
      "step": 1671
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.601168655813745e-05,
      "loss": 0.5841,
      "step": 1672
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.6006795451720272e-05,
      "loss": 0.5598,
      "step": 1673
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.6001902096088745e-05,
      "loss": 0.743,
      "step": 1674
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5997006493075154e-05,
      "loss": 0.4822,
      "step": 1675
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5992108644512636e-05,
      "loss": 0.5429,
      "step": 1676
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.598720855223516e-05,
      "loss": 0.5175,
      "step": 1677
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.598230621807755e-05,
      "loss": 0.62,
      "step": 1678
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.597740164387546e-05,
      "loss": 0.7945,
      "step": 1679
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5972494831465376e-05,
      "loss": 0.5644,
      "step": 1680
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5967585782684635e-05,
      "loss": 0.6724,
      "step": 1681
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.59626744993714e-05,
      "loss": 0.4584,
      "step": 1682
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.595776098336468e-05,
      "loss": 0.5387,
      "step": 1683
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.595284523650432e-05,
      "loss": 0.5537,
      "step": 1684
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5947927260630985e-05,
      "loss": 0.7468,
      "step": 1685
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.59430070575862e-05,
      "loss": 0.5811,
      "step": 1686
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5938084629212308e-05,
      "loss": 0.6037,
      "step": 1687
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.593315997735248e-05,
      "loss": 0.8431,
      "step": 1688
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.592823310385073e-05,
      "loss": 0.5807,
      "step": 1689
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5923304010551904e-05,
      "loss": 0.5783,
      "step": 1690
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5918372699301673e-05,
      "loss": 0.6102,
      "step": 1691
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5913439171946542e-05,
      "loss": 0.6449,
      "step": 1692
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.590850343033385e-05,
      "loss": 0.5137,
      "step": 1693
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5903565476311755e-05,
      "loss": 0.563,
      "step": 1694
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.589862531172925e-05,
      "loss": 0.601,
      "step": 1695
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5893682938436158e-05,
      "loss": 0.552,
      "step": 1696
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5888738358283125e-05,
      "loss": 0.6442,
      "step": 1697
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5883791573121622e-05,
      "loss": 0.6054,
      "step": 1698
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.587884258480395e-05,
      "loss": 0.3972,
      "step": 1699
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5873891395183228e-05,
      "loss": 0.5666,
      "step": 1700
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5868938006113407e-05,
      "loss": 0.647,
      "step": 1701
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5863982419449265e-05,
      "loss": 0.6769,
      "step": 1702
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.585902463704639e-05,
      "loss": 0.591,
      "step": 1703
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5854064660761196e-05,
      "loss": 0.7816,
      "step": 1704
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5849102492450924e-05,
      "loss": 0.4772,
      "step": 1705
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.584413813397364e-05,
      "loss": 0.5074,
      "step": 1706
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5839171587188213e-05,
      "loss": 0.5609,
      "step": 1707
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5834202853954342e-05,
      "loss": 0.6974,
      "step": 1708
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5829231936132555e-05,
      "loss": 0.5999,
      "step": 1709
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5824258835584174e-05,
      "loss": 0.4979,
      "step": 1710
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5819283554171356e-05,
      "loss": 0.6459,
      "step": 1711
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5814306093757076e-05,
      "loss": 0.6185,
      "step": 1712
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.580932645620511e-05,
      "loss": 0.589,
      "step": 1713
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5804344643380064e-05,
      "loss": 0.5937,
      "step": 1714
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5799360657147355e-05,
      "loss": 0.4549,
      "step": 1715
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5794374499373202e-05,
      "loss": 0.638,
      "step": 1716
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5789386171924656e-05,
      "loss": 0.4639,
      "step": 1717
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5784395676669567e-05,
      "loss": 0.4877,
      "step": 1718
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.57794030154766e-05,
      "loss": 0.6558,
      "step": 1719
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.5774408190215232e-05,
      "loss": 0.5548,
      "step": 1720
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5769411202755756e-05,
      "loss": 0.6746,
      "step": 1721
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5764412054969264e-05,
      "loss": 0.54,
      "step": 1722
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5759410748727663e-05,
      "loss": 0.6056,
      "step": 1723
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5754407285903657e-05,
      "loss": 0.7075,
      "step": 1724
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5749401668370786e-05,
      "loss": 0.5612,
      "step": 1725
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.574439389800337e-05,
      "loss": 0.5069,
      "step": 1726
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5739383976676538e-05,
      "loss": 0.531,
      "step": 1727
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5734371906266236e-05,
      "loss": 0.5308,
      "step": 1728
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5729357688649205e-05,
      "loss": 0.6665,
      "step": 1729
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5724341325702996e-05,
      "loss": 0.7032,
      "step": 1730
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.571932281930596e-05,
      "loss": 0.5157,
      "step": 1731
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5714302171337254e-05,
      "loss": 0.6679,
      "step": 1732
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5709279383676833e-05,
      "loss": 0.6558,
      "step": 1733
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5704254458205446e-05,
      "loss": 0.67,
      "step": 1734
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5699227396804667e-05,
      "loss": 0.5117,
      "step": 1735
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5694198201356842e-05,
      "loss": 0.6767,
      "step": 1736
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5689166873745133e-05,
      "loss": 0.5885,
      "step": 1737
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5684133415853497e-05,
      "loss": 0.4319,
      "step": 1738
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.567909782956668e-05,
      "loss": 0.5947,
      "step": 1739
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5674060116770234e-05,
      "loss": 0.6532,
      "step": 1740
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5669020279350514e-05,
      "loss": 0.6943,
      "step": 1741
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5663978319194647e-05,
      "loss": 0.7376,
      "step": 1742
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.565893423819058e-05,
      "loss": 0.6424,
      "step": 1743
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5653888038227044e-05,
      "loss": 0.5472,
      "step": 1744
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.564883972119356e-05,
      "loss": 0.6489,
      "step": 1745
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5643789288980446e-05,
      "loss": 0.4147,
      "step": 1746
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5638736743478807e-05,
      "loss": 0.5784,
      "step": 1747
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.563368208658055e-05,
      "loss": 0.6114,
      "step": 1748
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5628625320178362e-05,
      "loss": 0.6054,
      "step": 1749
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5623566446165722e-05,
      "loss": 0.5766,
      "step": 1750
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5618505466436907e-05,
      "loss": 0.4565,
      "step": 1751
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5613442382886966e-05,
      "loss": 0.5759,
      "step": 1752
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.560837719741175e-05,
      "loss": 0.6293,
      "step": 1753
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5603309911907897e-05,
      "loss": 0.6898,
      "step": 1754
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5598240528272815e-05,
      "loss": 0.6254,
      "step": 1755
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5593169048404718e-05,
      "loss": 0.4497,
      "step": 1756
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5588095474202597e-05,
      "loss": 0.6484,
      "step": 1757
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.558301980756622e-05,
      "loss": 0.4421,
      "step": 1758
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5577942050396147e-05,
      "loss": 0.4923,
      "step": 1759
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.557286220459372e-05,
      "loss": 0.5483,
      "step": 1760
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.556778027206107e-05,
      "loss": 0.4994,
      "step": 1761
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.556269625470108e-05,
      "loss": 0.5308,
      "step": 1762
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5557610154417456e-05,
      "loss": 0.5752,
      "step": 1763
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.555252197311465e-05,
      "loss": 0.5097,
      "step": 1764
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.554743171269791e-05,
      "loss": 0.6006,
      "step": 1765
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5542339375073262e-05,
      "loss": 0.5182,
      "step": 1766
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.55372449621475e-05,
      "loss": 0.4459,
      "step": 1767
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5532148475828206e-05,
      "loss": 0.5107,
      "step": 1768
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.552704991802373e-05,
      "loss": 0.6753,
      "step": 1769
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5521949290643203e-05,
      "loss": 0.6457,
      "step": 1770
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5516846595596525e-05,
      "loss": 0.6172,
      "step": 1771
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5511741834794386e-05,
      "loss": 0.4617,
      "step": 1772
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.5506635010148228e-05,
      "loss": 0.6032,
      "step": 1773
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5501526123570277e-05,
      "loss": 0.8055,
      "step": 1774
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5496415176973532e-05,
      "loss": 0.5206,
      "step": 1775
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5491302172271764e-05,
      "loss": 0.5727,
      "step": 1776
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.54861871113795e-05,
      "loss": 0.6873,
      "step": 1777
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5481069996212063e-05,
      "loss": 0.6008,
      "step": 1778
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5475950828685528e-05,
      "loss": 0.6257,
      "step": 1779
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5470829610716735e-05,
      "loss": 0.5878,
      "step": 1780
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5465706344223303e-05,
      "loss": 0.5188,
      "step": 1781
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5460581031123613e-05,
      "loss": 0.5916,
      "step": 1782
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5455453673336813e-05,
      "loss": 0.5638,
      "step": 1783
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5450324272782813e-05,
      "loss": 0.6927,
      "step": 1784
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5445192831382296e-05,
      "loss": 0.6456,
      "step": 1785
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5440059351056704e-05,
      "loss": 0.5446,
      "step": 1786
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5434923833728238e-05,
      "loss": 0.6715,
      "step": 1787
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5429786281319873e-05,
      "loss": 0.6587,
      "step": 1788
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5424646695755337e-05,
      "loss": 0.5771,
      "step": 1789
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5419505078959118e-05,
      "loss": 0.501,
      "step": 1790
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5414361432856475e-05,
      "loss": 0.7051,
      "step": 1791
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5409215759373415e-05,
      "loss": 0.5519,
      "step": 1792
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5404068060436715e-05,
      "loss": 0.5305,
      "step": 1793
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.53989183379739e-05,
      "loss": 0.5829,
      "step": 1794
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.539376659391326e-05,
      "loss": 0.6817,
      "step": 1795
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5388612830183842e-05,
      "loss": 0.6226,
      "step": 1796
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.538345704871544e-05,
      "loss": 0.6398,
      "step": 1797
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.537829925143862e-05,
      "loss": 0.6571,
      "step": 1798
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5373139440284687e-05,
      "loss": 0.602,
      "step": 1799
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.53679776171857e-05,
      "loss": 0.5631,
      "step": 1800
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5362813784074493e-05,
      "loss": 0.84,
      "step": 1801
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5357647942884627e-05,
      "loss": 0.5393,
      "step": 1802
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5352480095550427e-05,
      "loss": 0.5971,
      "step": 1803
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.534731024400697e-05,
      "loss": 0.688,
      "step": 1804
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5342138390190078e-05,
      "loss": 0.6871,
      "step": 1805
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5336964536036328e-05,
      "loss": 0.6208,
      "step": 1806
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.533178868348304e-05,
      "loss": 0.7018,
      "step": 1807
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.532661083446829e-05,
      "loss": 0.5821,
      "step": 1808
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5321430990930893e-05,
      "loss": 0.5127,
      "step": 1809
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.531624915481042e-05,
      "loss": 0.6247,
      "step": 1810
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5311065328047184e-05,
      "loss": 0.5795,
      "step": 1811
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.530587951258224e-05,
      "loss": 0.5779,
      "step": 1812
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.530069171035739e-05,
      "loss": 0.5333,
      "step": 1813
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5295501923315178e-05,
      "loss": 0.6923,
      "step": 1814
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5290310153398906e-05,
      "loss": 0.4923,
      "step": 1815
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5285116402552592e-05,
      "loss": 0.6078,
      "step": 1816
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5279920672721014e-05,
      "loss": 0.7185,
      "step": 1817
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5274722965849692e-05,
      "loss": 0.5658,
      "step": 1818
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5269523283884874e-05,
      "loss": 0.6284,
      "step": 1819
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.526432162877356e-05,
      "loss": 0.4198,
      "step": 1820
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5259118002463484e-05,
      "loss": 0.5104,
      "step": 1821
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5253912406903109e-05,
      "loss": 0.5732,
      "step": 1822
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5248704844041655e-05,
      "loss": 0.6834,
      "step": 1823
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5243495315829062e-05,
      "loss": 0.6216,
      "step": 1824
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5238283824216015e-05,
      "loss": 0.6011,
      "step": 1825
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.5233070371153926e-05,
      "loss": 0.5288,
      "step": 1826
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.522785495859495e-05,
      "loss": 0.5025,
      "step": 1827
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5222637588491971e-05,
      "loss": 0.7065,
      "step": 1828
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.521741826279861e-05,
      "loss": 0.6381,
      "step": 1829
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5212196983469212e-05,
      "loss": 0.6338,
      "step": 1830
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5206973752458857e-05,
      "loss": 0.4671,
      "step": 1831
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5201748571723367e-05,
      "loss": 0.6506,
      "step": 1832
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5196521443219277e-05,
      "loss": 0.5925,
      "step": 1833
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5191292368903865e-05,
      "loss": 0.6347,
      "step": 1834
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5186061350735124e-05,
      "loss": 0.7759,
      "step": 1835
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5180828390671787e-05,
      "loss": 0.7019,
      "step": 1836
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.517559349067331e-05,
      "loss": 0.4756,
      "step": 1837
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5170356652699873e-05,
      "loss": 0.5113,
      "step": 1838
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5165117878712386e-05,
      "loss": 0.6124,
      "step": 1839
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.515987717067248e-05,
      "loss": 0.5485,
      "step": 1840
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5154634530542512e-05,
      "loss": 0.502,
      "step": 1841
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.514938996028556e-05,
      "loss": 0.5099,
      "step": 1842
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.514414346186543e-05,
      "loss": 0.7417,
      "step": 1843
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5138895037246646e-05,
      "loss": 0.4921,
      "step": 1844
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5133644688394456e-05,
      "loss": 0.6719,
      "step": 1845
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5128392417274822e-05,
      "loss": 0.5741,
      "step": 1846
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5123138225854437e-05,
      "loss": 0.6025,
      "step": 1847
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.51178821161007e-05,
      "loss": 0.6541,
      "step": 1848
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.511262408998174e-05,
      "loss": 0.5056,
      "step": 1849
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5107364149466399e-05,
      "loss": 0.5652,
      "step": 1850
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5102102296524228e-05,
      "loss": 0.6504,
      "step": 1851
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5096838533125508e-05,
      "loss": 0.4422,
      "step": 1852
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5091572861241228e-05,
      "loss": 0.5864,
      "step": 1853
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5086305282843089e-05,
      "loss": 0.5793,
      "step": 1854
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5081035799903511e-05,
      "loss": 0.7601,
      "step": 1855
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5075764414395627e-05,
      "loss": 0.6921,
      "step": 1856
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.507049112829328e-05,
      "loss": 0.4504,
      "step": 1857
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5065215943571026e-05,
      "loss": 0.5741,
      "step": 1858
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5059938862204126e-05,
      "loss": 0.631,
      "step": 1859
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5054659886168566e-05,
      "loss": 0.6853,
      "step": 1860
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5049379017441026e-05,
      "loss": 0.4414,
      "step": 1861
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5044096257998902e-05,
      "loss": 0.47,
      "step": 1862
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5038811609820294e-05,
      "loss": 0.5189,
      "step": 1863
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5033525074884017e-05,
      "loss": 0.6342,
      "step": 1864
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5028236655169583e-05,
      "loss": 0.5489,
      "step": 1865
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5022946352657216e-05,
      "loss": 0.6361,
      "step": 1866
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5017654169327847e-05,
      "loss": 0.5531,
      "step": 1867
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5012360107163098e-05,
      "loss": 0.7704,
      "step": 1868
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5007064168145316e-05,
      "loss": 0.581,
      "step": 1869
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.5001766354257528e-05,
      "loss": 0.4632,
      "step": 1870
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4996466667483479e-05,
      "loss": 0.6607,
      "step": 1871
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.499116510980761e-05,
      "loss": 0.6787,
      "step": 1872
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4985861683215059e-05,
      "loss": 0.8264,
      "step": 1873
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4980556389691675e-05,
      "loss": 0.4967,
      "step": 1874
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4975249231223987e-05,
      "loss": 0.6395,
      "step": 1875
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4969940209799248e-05,
      "loss": 0.57,
      "step": 1876
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4964629327405385e-05,
      "loss": 0.4951,
      "step": 1877
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.495931658603103e-05,
      "loss": 0.5687,
      "step": 1878
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4954001987665518e-05,
      "loss": 0.6539,
      "step": 1879
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4948685534298868e-05,
      "loss": 0.5703,
      "step": 1880
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4943367227921805e-05,
      "loss": 0.6911,
      "step": 1881
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4938047070525739e-05,
      "loss": 0.5946,
      "step": 1882
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4932725064102778e-05,
      "loss": 0.5287,
      "step": 1883
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4927401210645718e-05,
      "loss": 0.5784,
      "step": 1884
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4922075512148049e-05,
      "loss": 0.698,
      "step": 1885
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4916747970603952e-05,
      "loss": 0.5247,
      "step": 1886
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4911418588008302e-05,
      "loss": 0.6944,
      "step": 1887
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4906087366356658e-05,
      "loss": 0.8495,
      "step": 1888
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4900754307645264e-05,
      "loss": 0.6692,
      "step": 1889
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4895419413871066e-05,
      "loss": 0.4697,
      "step": 1890
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.489008268703168e-05,
      "loss": 0.5792,
      "step": 1891
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4884744129125421e-05,
      "loss": 0.6068,
      "step": 1892
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4879403742151283e-05,
      "loss": 0.5292,
      "step": 1893
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4874061528108948e-05,
      "loss": 0.6115,
      "step": 1894
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4868717488998785e-05,
      "loss": 0.6788,
      "step": 1895
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4863371626821831e-05,
      "loss": 0.6662,
      "step": 1896
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4858023943579831e-05,
      "loss": 0.6338,
      "step": 1897
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4852674441275192e-05,
      "loss": 0.6752,
      "step": 1898
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4847323121911008e-05,
      "loss": 0.4482,
      "step": 1899
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4841969987491052e-05,
      "loss": 0.5345,
      "step": 1900
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4836615040019781e-05,
      "loss": 0.7101,
      "step": 1901
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4831258281502328e-05,
      "loss": 0.5502,
      "step": 1902
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4825899713944503e-05,
      "loss": 0.4552,
      "step": 1903
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4820539339352795e-05,
      "loss": 0.6086,
      "step": 1904
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4815177159734373e-05,
      "loss": 0.6878,
      "step": 1905
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.480981317709707e-05,
      "loss": 0.5485,
      "step": 1906
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4804447393449408e-05,
      "loss": 0.5138,
      "step": 1907
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4799079810800579e-05,
      "loss": 0.736,
      "step": 1908
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4793710431160444e-05,
      "loss": 0.4627,
      "step": 1909
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4788339256539543e-05,
      "loss": 0.5551,
      "step": 1910
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.478296628894908e-05,
      "loss": 0.6607,
      "step": 1911
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4777591530400943e-05,
      "loss": 0.5358,
      "step": 1912
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4772214982907678e-05,
      "loss": 0.5681,
      "step": 1913
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.476683664848251e-05,
      "loss": 0.4296,
      "step": 1914
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4761456529139324e-05,
      "loss": 0.4882,
      "step": 1915
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4756074626892684e-05,
      "loss": 0.6184,
      "step": 1916
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4750690943757815e-05,
      "loss": 0.5742,
      "step": 1917
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4745305481750607e-05,
      "loss": 0.4511,
      "step": 1918
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4739918242887623e-05,
      "loss": 0.585,
      "step": 1919
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4734529229186082e-05,
      "loss": 0.5163,
      "step": 1920
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4729138442663884e-05,
      "loss": 0.5508,
      "step": 1921
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.472374588533957e-05,
      "loss": 0.546,
      "step": 1922
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4718351559232362e-05,
      "loss": 0.6106,
      "step": 1923
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4712955466362136e-05,
      "loss": 0.621,
      "step": 1924
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4707557608749434e-05,
      "loss": 0.6339,
      "step": 1925
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4702157988415455e-05,
      "loss": 0.7364,
      "step": 1926
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.469675660738206e-05,
      "loss": 0.6693,
      "step": 1927
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4691353467671772e-05,
      "loss": 0.6322,
      "step": 1928
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4685948571307762e-05,
      "loss": 0.5154,
      "step": 1929
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4680541920313876e-05,
      "loss": 0.6501,
      "step": 1930
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4675133516714599e-05,
      "loss": 0.5716,
      "step": 1931
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4669723362535084e-05,
      "loss": 0.6371,
      "step": 1932
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4664311459801139e-05,
      "loss": 0.4324,
      "step": 1933
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4658897810539217e-05,
      "loss": 0.6323,
      "step": 1934
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4653482416776436e-05,
      "loss": 0.5545,
      "step": 1935
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4648065280540562e-05,
      "loss": 0.54,
      "step": 1936
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4642646403860017e-05,
      "loss": 0.4739,
      "step": 1937
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.463722578876387e-05,
      "loss": 0.6942,
      "step": 1938
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4631803437281837e-05,
      "loss": 0.5857,
      "step": 1939
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4626379351444304e-05,
      "loss": 0.5119,
      "step": 1940
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.462095353328228e-05,
      "loss": 0.5066,
      "step": 1941
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4615525984827441e-05,
      "loss": 0.4576,
      "step": 1942
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4610096708112107e-05,
      "loss": 0.5459,
      "step": 1943
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4604665705169239e-05,
      "loss": 0.5528,
      "step": 1944
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4599232978032452e-05,
      "loss": 0.7566,
      "step": 1945
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4593798528735998e-05,
      "loss": 0.8539,
      "step": 1946
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4588362359314787e-05,
      "loss": 0.5186,
      "step": 1947
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4582924471804358e-05,
      "loss": 0.6615,
      "step": 1948
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4577484868240908e-05,
      "loss": 0.5027,
      "step": 1949
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4572043550661262e-05,
      "loss": 0.5968,
      "step": 1950
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4566600521102898e-05,
      "loss": 0.619,
      "step": 1951
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.456115578160393e-05,
      "loss": 0.553,
      "step": 1952
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4555709334203116e-05,
      "loss": 0.4782,
      "step": 1953
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4550261180939846e-05,
      "loss": 0.7012,
      "step": 1954
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4544811323854155e-05,
      "loss": 0.5721,
      "step": 1955
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.453935976498672e-05,
      "loss": 0.5037,
      "step": 1956
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.453390650637884e-05,
      "loss": 0.467,
      "step": 1957
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4528451550072472e-05,
      "loss": 0.6087,
      "step": 1958
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4522994898110188e-05,
      "loss": 0.5373,
      "step": 1959
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4517536552535205e-05,
      "loss": 0.5095,
      "step": 1960
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4512076515391375e-05,
      "loss": 0.4573,
      "step": 1961
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4506614788723181e-05,
      "loss": 0.5063,
      "step": 1962
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4501151374575742e-05,
      "loss": 0.5337,
      "step": 1963
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4495686274994799e-05,
      "loss": 0.4332,
      "step": 1964
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.449021949202674e-05,
      "loss": 0.7649,
      "step": 1965
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4484751027718567e-05,
      "loss": 0.3724,
      "step": 1966
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4479280884117919e-05,
      "loss": 0.609,
      "step": 1967
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.447380906327307e-05,
      "loss": 0.4315,
      "step": 1968
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.446833556723291e-05,
      "loss": 0.42,
      "step": 1969
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4462860398046964e-05,
      "loss": 0.6879,
      "step": 1970
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4457383557765385e-05,
      "loss": 0.5968,
      "step": 1971
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4451905048438943e-05,
      "loss": 0.4947,
      "step": 1972
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4446424872119041e-05,
      "loss": 0.6004,
      "step": 1973
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4440943030857698e-05,
      "loss": 0.5348,
      "step": 1974
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4435459526707574e-05,
      "loss": 0.6951,
      "step": 1975
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4429974361721928e-05,
      "loss": 0.6712,
      "step": 1976
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4424487537954658e-05,
      "loss": 0.6103,
      "step": 1977
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4418999057460277e-05,
      "loss": 0.6401,
      "step": 1978
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4413508922293917e-05,
      "loss": 0.5248,
      "step": 1979
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4408017134511334e-05,
      "loss": 0.6691,
      "step": 1980
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4402523696168898e-05,
      "loss": 0.5254,
      "step": 1981
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.43970286093236e-05,
      "loss": 0.5766,
      "step": 1982
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4391531876033047e-05,
      "loss": 0.563,
      "step": 1983
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.4386033498355464e-05,
      "loss": 0.5582,
      "step": 1984
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4380533478349694e-05,
      "loss": 0.6873,
      "step": 1985
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.437503181807519e-05,
      "loss": 0.4901,
      "step": 1986
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4369528519592016e-05,
      "loss": 0.6593,
      "step": 1987
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.436402358496086e-05,
      "loss": 0.6982,
      "step": 1988
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4358517016243013e-05,
      "loss": 0.5235,
      "step": 1989
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4353008815500387e-05,
      "loss": 0.6964,
      "step": 1990
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4347498984795501e-05,
      "loss": 0.456,
      "step": 1991
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4341987526191475e-05,
      "loss": 0.6528,
      "step": 1992
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4336474441752052e-05,
      "loss": 0.583,
      "step": 1993
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4330959733541581e-05,
      "loss": 0.5862,
      "step": 1994
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4325443403625012e-05,
      "loss": 0.5675,
      "step": 1995
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4319925454067912e-05,
      "loss": 0.5143,
      "step": 1996
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4314405886936444e-05,
      "loss": 0.7891,
      "step": 1997
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4308884704297388e-05,
      "loss": 0.6855,
      "step": 1998
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4303361908218119e-05,
      "loss": 0.5469,
      "step": 1999
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4297837500766619e-05,
      "loss": 0.5859,
      "step": 2000
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.429231148401148e-05,
      "loss": 0.5573,
      "step": 2001
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4286783860021884e-05,
      "loss": 0.6411,
      "step": 2002
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.428125463086763e-05,
      "loss": 0.5588,
      "step": 2003
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4275723798619106e-05,
      "loss": 0.6823,
      "step": 2004
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4270191365347304e-05,
      "loss": 0.5778,
      "step": 2005
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4264657333123814e-05,
      "loss": 0.6493,
      "step": 2006
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.425912170402083e-05,
      "loss": 0.5854,
      "step": 2007
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4253584480111142e-05,
      "loss": 0.5067,
      "step": 2008
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4248045663468129e-05,
      "loss": 0.7334,
      "step": 2009
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4242505256165779e-05,
      "loss": 0.5397,
      "step": 2010
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4236963260278667e-05,
      "loss": 0.7243,
      "step": 2011
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4231419677881966e-05,
      "loss": 0.7491,
      "step": 2012
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4225874511051445e-05,
      "loss": 0.6077,
      "step": 2013
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.422032776186346e-05,
      "loss": 0.5184,
      "step": 2014
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4214779432394965e-05,
      "loss": 0.5572,
      "step": 2015
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4209229524723503e-05,
      "loss": 0.5067,
      "step": 2016
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4203678040927211e-05,
      "loss": 0.525,
      "step": 2017
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4198124983084812e-05,
      "loss": 0.5396,
      "step": 2018
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4192570353275623e-05,
      "loss": 0.5606,
      "step": 2019
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4187014153579548e-05,
      "loss": 0.5509,
      "step": 2020
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4181456386077073e-05,
      "loss": 0.4435,
      "step": 2021
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4175897052849278e-05,
      "loss": 0.6465,
      "step": 2022
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4170336155977833e-05,
      "loss": 0.7233,
      "step": 2023
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.416477369754498e-05,
      "loss": 0.5652,
      "step": 2024
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.415920967963356e-05,
      "loss": 0.6913,
      "step": 2025
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4153644104326985e-05,
      "loss": 0.5659,
      "step": 2026
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.414807697370926e-05,
      "loss": 0.4572,
      "step": 2027
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4142508289864972e-05,
      "loss": 0.5511,
      "step": 2028
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4136938054879284e-05,
      "loss": 0.5156,
      "step": 2029
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4131366270837938e-05,
      "loss": 0.669,
      "step": 2030
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4125792939827268e-05,
      "loss": 0.4658,
      "step": 2031
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4120218063934177e-05,
      "loss": 0.7922,
      "step": 2032
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.411464164524615e-05,
      "loss": 0.5772,
      "step": 2033
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4109063685851245e-05,
      "loss": 0.5447,
      "step": 2034
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4103484187838104e-05,
      "loss": 0.5551,
      "step": 2035
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4097903153295944e-05,
      "loss": 0.5805,
      "step": 2036
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.4092320584314552e-05,
      "loss": 0.4206,
      "step": 2037
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4086736482984294e-05,
      "loss": 0.6699,
      "step": 2038
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4081150851396109e-05,
      "loss": 0.4838,
      "step": 2039
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4075563691641506e-05,
      "loss": 0.4567,
      "step": 2040
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4069975005812573e-05,
      "loss": 0.5208,
      "step": 2041
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4064384796001962e-05,
      "loss": 0.762,
      "step": 2042
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4058793064302902e-05,
      "loss": 0.66,
      "step": 2043
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4053199812809186e-05,
      "loss": 0.5597,
      "step": 2044
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4047605043615183e-05,
      "loss": 0.5755,
      "step": 2045
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.404200875881582e-05,
      "loss": 0.5111,
      "step": 2046
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4036410960506601e-05,
      "loss": 0.603,
      "step": 2047
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4030811650783594e-05,
      "loss": 0.4649,
      "step": 2048
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4025210831743431e-05,
      "loss": 0.5139,
      "step": 2049
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4019608505483314e-05,
      "loss": 0.5652,
      "step": 2050
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4014004674101001e-05,
      "loss": 0.5222,
      "step": 2051
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4008399339694822e-05,
      "loss": 0.6241,
      "step": 2052
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.4002792504363665e-05,
      "loss": 0.5817,
      "step": 2053
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.399718417020698e-05,
      "loss": 0.6278,
      "step": 2054
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3991574339324782e-05,
      "loss": 0.4877,
      "step": 2055
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.398596301381764e-05,
      "loss": 0.6347,
      "step": 2056
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3980350195786691e-05,
      "loss": 0.6038,
      "step": 2057
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3974735887333627e-05,
      "loss": 0.5174,
      "step": 2058
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3969120090560695e-05,
      "loss": 0.639,
      "step": 2059
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3963502807570702e-05,
      "loss": 0.6459,
      "step": 2060
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3957884040467006e-05,
      "loss": 0.5329,
      "step": 2061
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.395226379135354e-05,
      "loss": 0.4267,
      "step": 2062
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3946642062334765e-05,
      "loss": 0.5191,
      "step": 2063
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3941018855515716e-05,
      "loss": 0.4508,
      "step": 2064
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3935394173001972e-05,
      "loss": 0.5123,
      "step": 2065
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3929768016899667e-05,
      "loss": 0.7707,
      "step": 2066
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3924140389315488e-05,
      "loss": 0.6802,
      "step": 2067
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3918511292356673e-05,
      "loss": 0.6645,
      "step": 2068
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3912880728131004e-05,
      "loss": 0.5383,
      "step": 2069
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3907248698746822e-05,
      "loss": 0.7546,
      "step": 2070
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3901615206313013e-05,
      "loss": 0.6405,
      "step": 2071
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3895980252939006e-05,
      "loss": 0.6276,
      "step": 2072
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3890343840734782e-05,
      "loss": 0.4313,
      "step": 2073
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.388470597181087e-05,
      "loss": 0.6365,
      "step": 2074
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.387906664827834e-05,
      "loss": 0.5469,
      "step": 2075
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3873425872248805e-05,
      "loss": 0.5573,
      "step": 2076
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3867783645834428e-05,
      "loss": 0.5712,
      "step": 2077
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3862139971147915e-05,
      "loss": 0.5401,
      "step": 2078
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3856494850302508e-05,
      "loss": 0.5148,
      "step": 2079
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3850848285411994e-05,
      "loss": 0.6839,
      "step": 2080
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3845200278590708e-05,
      "loss": 0.6191,
      "step": 2081
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3839550831953503e-05,
      "loss": 0.5655,
      "step": 2082
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3833899947615801e-05,
      "loss": 0.5445,
      "step": 2083
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3828247627693538e-05,
      "loss": 0.648,
      "step": 2084
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3822593874303201e-05,
      "loss": 0.5886,
      "step": 2085
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3816938689561808e-05,
      "loss": 0.6148,
      "step": 2086
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3811282075586916e-05,
      "loss": 0.5431,
      "step": 2087
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3805624034496612e-05,
      "loss": 0.5258,
      "step": 2088
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3799964568409523e-05,
      "loss": 0.6313,
      "step": 2089
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.3794303679444813e-05,
      "loss": 0.4623,
      "step": 2090
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3788641369722163e-05,
      "loss": 0.5208,
      "step": 2091
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3782977641361803e-05,
      "loss": 0.4881,
      "step": 2092
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.377731249648449e-05,
      "loss": 0.5885,
      "step": 2093
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3771645937211502e-05,
      "loss": 0.5007,
      "step": 2094
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3765977965664659e-05,
      "loss": 0.4677,
      "step": 2095
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3760308583966299e-05,
      "loss": 0.6411,
      "step": 2096
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3754637794239303e-05,
      "loss": 0.6303,
      "step": 2097
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3748965598607057e-05,
      "loss": 0.5626,
      "step": 2098
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3743291999193495e-05,
      "loss": 0.6903,
      "step": 2099
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3737616998123068e-05,
      "loss": 0.5027,
      "step": 2100
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.373194059752075e-05,
      "loss": 0.4535,
      "step": 2101
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3726262799512038e-05,
      "loss": 0.7366,
      "step": 2102
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3720583606222956e-05,
      "loss": 0.7491,
      "step": 2103
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.371490301978005e-05,
      "loss": 0.6673,
      "step": 2104
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3709221042310388e-05,
      "loss": 0.6727,
      "step": 2105
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.370353767594156e-05,
      "loss": 0.781,
      "step": 2106
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3697852922801669e-05,
      "loss": 0.6918,
      "step": 2107
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.369216678501934e-05,
      "loss": 0.6525,
      "step": 2108
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3686479264723728e-05,
      "loss": 0.4852,
      "step": 2109
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3680790364044492e-05,
      "loss": 0.6414,
      "step": 2110
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3675100085111809e-05,
      "loss": 0.5319,
      "step": 2111
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.366940843005638e-05,
      "loss": 0.5301,
      "step": 2112
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3663715401009414e-05,
      "loss": 0.4729,
      "step": 2113
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3658021000102638e-05,
      "loss": 0.5805,
      "step": 2114
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3652325229468288e-05,
      "loss": 0.5481,
      "step": 2115
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3646628091239125e-05,
      "loss": 0.5088,
      "step": 2116
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3640929587548403e-05,
      "loss": 0.5655,
      "step": 2117
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.363522972052991e-05,
      "loss": 0.545,
      "step": 2118
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3629528492317925e-05,
      "loss": 0.5359,
      "step": 2119
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3623825905047245e-05,
      "loss": 0.5717,
      "step": 2120
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3618121960853177e-05,
      "loss": 0.618,
      "step": 2121
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3612416661871532e-05,
      "loss": 0.5926,
      "step": 2122
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3606710010238633e-05,
      "loss": 0.514,
      "step": 2123
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3601002008091302e-05,
      "loss": 0.6786,
      "step": 2124
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3595292657566876e-05,
      "loss": 0.5365,
      "step": 2125
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3589581960803195e-05,
      "loss": 0.4894,
      "step": 2126
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3583869919938597e-05,
      "loss": 0.7511,
      "step": 2127
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3578156537111928e-05,
      "loss": 0.5222,
      "step": 2128
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3572441814462534e-05,
      "loss": 0.6847,
      "step": 2129
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3566725754130265e-05,
      "loss": 0.7435,
      "step": 2130
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.356100835825547e-05,
      "loss": 0.6757,
      "step": 2131
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3555289628979005e-05,
      "loss": 0.657,
      "step": 2132
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3549569568442208e-05,
      "loss": 0.4837,
      "step": 2133
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3543848178786941e-05,
      "loss": 0.4819,
      "step": 2134
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.353812546215554e-05,
      "loss": 0.4873,
      "step": 2135
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3532401420690847e-05,
      "loss": 0.6954,
      "step": 2136
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3526676056536205e-05,
      "loss": 0.5893,
      "step": 2137
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3520949371835448e-05,
      "loss": 0.6295,
      "step": 2138
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.35152213687329e-05,
      "loss": 0.4769,
      "step": 2139
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3509492049373388e-05,
      "loss": 0.4812,
      "step": 2140
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3503761415902221e-05,
      "loss": 0.4917,
      "step": 2141
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3498029470465212e-05,
      "loss": 0.5155,
      "step": 2142
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.3492296215208653e-05,
      "loss": 0.4701,
      "step": 2143
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.348656165227934e-05,
      "loss": 0.7118,
      "step": 2144
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3480825783824545e-05,
      "loss": 0.5449,
      "step": 2145
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3475088611992038e-05,
      "loss": 0.5267,
      "step": 2146
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3469350138930073e-05,
      "loss": 0.6651,
      "step": 2147
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3463610366787392e-05,
      "loss": 0.5834,
      "step": 2148
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3457869297713227e-05,
      "loss": 0.4862,
      "step": 2149
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3452126933857286e-05,
      "loss": 0.5247,
      "step": 2150
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3446383277369777e-05,
      "loss": 0.6938,
      "step": 2151
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3440638330401375e-05,
      "loss": 0.6272,
      "step": 2152
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.343489209510325e-05,
      "loss": 0.6388,
      "step": 2153
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3429144573627055e-05,
      "loss": 0.3972,
      "step": 2154
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3423395768124914e-05,
      "loss": 0.4017,
      "step": 2155
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3417645680749441e-05,
      "loss": 0.4779,
      "step": 2156
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3411894313653727e-05,
      "loss": 0.6229,
      "step": 2157
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3406141668991344e-05,
      "loss": 0.4434,
      "step": 2158
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3400387748916338e-05,
      "loss": 0.5871,
      "step": 2159
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3394632555583238e-05,
      "loss": 0.5689,
      "step": 2160
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3388876091147048e-05,
      "loss": 0.7287,
      "step": 2161
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.338311835776324e-05,
      "loss": 0.7309,
      "step": 2162
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3377359357587776e-05,
      "loss": 0.5997,
      "step": 2163
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3371599092777084e-05,
      "loss": 0.4905,
      "step": 2164
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3365837565488065e-05,
      "loss": 0.574,
      "step": 2165
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3360074777878089e-05,
      "loss": 0.7338,
      "step": 2166
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3354310732105014e-05,
      "loss": 0.5053,
      "step": 2167
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3348545430327146e-05,
      "loss": 0.6964,
      "step": 2168
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.334277887470328e-05,
      "loss": 0.6457,
      "step": 2169
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3337011067392673e-05,
      "loss": 0.5051,
      "step": 2170
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.333124201055505e-05,
      "loss": 0.4775,
      "step": 2171
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3325471706350606e-05,
      "loss": 0.5111,
      "step": 2172
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3319700156940003e-05,
      "loss": 0.5008,
      "step": 2173
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3313927364484368e-05,
      "loss": 0.581,
      "step": 2174
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3308153331145293e-05,
      "loss": 0.5461,
      "step": 2175
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3302378059084835e-05,
      "loss": 0.5714,
      "step": 2176
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3296601550465525e-05,
      "loss": 0.5187,
      "step": 2177
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3290823807450333e-05,
      "loss": 0.6369,
      "step": 2178
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3285044832202716e-05,
      "loss": 0.6136,
      "step": 2179
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3279264626886579e-05,
      "loss": 0.5102,
      "step": 2180
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.327348319366629e-05,
      "loss": 0.6006,
      "step": 2181
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.326770053470668e-05,
      "loss": 0.5622,
      "step": 2182
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3261916652173034e-05,
      "loss": 0.5852,
      "step": 2183
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.32561315482311e-05,
      "loss": 0.6932,
      "step": 2184
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3250345225047078e-05,
      "loss": 0.6941,
      "step": 2185
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3244557684787631e-05,
      "loss": 0.6159,
      "step": 2186
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3238768929619874e-05,
      "loss": 0.4833,
      "step": 2187
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3232978961711376e-05,
      "loss": 0.5224,
      "step": 2188
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.322718778323016e-05,
      "loss": 0.7408,
      "step": 2189
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3221395396344707e-05,
      "loss": 0.5099,
      "step": 2190
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3215601803223945e-05,
      "loss": 0.5155,
      "step": 2191
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3209807006037254e-05,
      "loss": 0.4841,
      "step": 2192
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3204011006954474e-05,
      "loss": 0.5624,
      "step": 2193
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3198213808145879e-05,
      "loss": 0.552,
      "step": 2194
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3192415411782203e-05,
      "loss": 0.5521,
      "step": 2195
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.3186615820034632e-05,
      "loss": 0.4757,
      "step": 2196
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3180815035074786e-05,
      "loss": 0.717,
      "step": 2197
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3175013059074747e-05,
      "loss": 0.7794,
      "step": 2198
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.316920989420703e-05,
      "loss": 0.5903,
      "step": 2199
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3163405542644602e-05,
      "loss": 0.5964,
      "step": 2200
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3157600006560878e-05,
      "loss": 0.6212,
      "step": 2201
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3151793288129705e-05,
      "loss": 0.5085,
      "step": 2202
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3145985389525387e-05,
      "loss": 0.6312,
      "step": 2203
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3140176312922654e-05,
      "loss": 0.5963,
      "step": 2204
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3134366060496693e-05,
      "loss": 0.7328,
      "step": 2205
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3128554634423118e-05,
      "loss": 0.597,
      "step": 2206
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3122742036877994e-05,
      "loss": 0.5104,
      "step": 2207
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3116928270037815e-05,
      "loss": 0.5216,
      "step": 2208
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3111113336079517e-05,
      "loss": 0.6339,
      "step": 2209
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3105297237180472e-05,
      "loss": 0.4872,
      "step": 2210
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.309947997551849e-05,
      "loss": 0.6051,
      "step": 2211
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.309366155327182e-05,
      "loss": 0.5744,
      "step": 2212
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3087841972619133e-05,
      "loss": 0.5541,
      "step": 2213
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3082021235739546e-05,
      "loss": 0.5593,
      "step": 2214
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3076199344812606e-05,
      "loss": 0.5775,
      "step": 2215
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3070376302018287e-05,
      "loss": 0.5996,
      "step": 2216
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3064552109537e-05,
      "loss": 0.5791,
      "step": 2217
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3058726769549584e-05,
      "loss": 0.4657,
      "step": 2218
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3052900284237311e-05,
      "loss": 0.5166,
      "step": 2219
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3047072655781874e-05,
      "loss": 0.7104,
      "step": 2220
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3041243886365402e-05,
      "loss": 0.5763,
      "step": 2221
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3035413978170455e-05,
      "loss": 0.5086,
      "step": 2222
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3029582933379998e-05,
      "loss": 0.6203,
      "step": 2223
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3023750754177449e-05,
      "loss": 0.4986,
      "step": 2224
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3017917442746633e-05,
      "loss": 0.5373,
      "step": 2225
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3012083001271805e-05,
      "loss": 0.5584,
      "step": 2226
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3006247431937644e-05,
      "loss": 0.6486,
      "step": 2227
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.3000410736929246e-05,
      "loss": 0.7272,
      "step": 2228
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2994572918432134e-05,
      "loss": 0.6536,
      "step": 2229
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2988733978632248e-05,
      "loss": 0.5865,
      "step": 2230
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2982893919715958e-05,
      "loss": 0.6606,
      "step": 2231
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2977052743870032e-05,
      "loss": 0.5464,
      "step": 2232
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2971210453281675e-05,
      "loss": 0.6519,
      "step": 2233
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2965367050138507e-05,
      "loss": 0.4899,
      "step": 2234
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2959522536628554e-05,
      "loss": 0.6431,
      "step": 2235
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.295367691494027e-05,
      "loss": 0.6226,
      "step": 2236
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2947830187262514e-05,
      "loss": 0.5144,
      "step": 2237
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.294198235578457e-05,
      "loss": 0.6407,
      "step": 2238
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2936133422696122e-05,
      "loss": 0.4638,
      "step": 2239
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.293028339018728e-05,
      "loss": 0.6761,
      "step": 2240
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2924432260448556e-05,
      "loss": 0.5601,
      "step": 2241
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2918580035670875e-05,
      "loss": 0.7177,
      "step": 2242
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2912726718045576e-05,
      "loss": 0.6005,
      "step": 2243
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2906872309764404e-05,
      "loss": 0.4866,
      "step": 2244
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2901016813019509e-05,
      "loss": 0.699,
      "step": 2245
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2895160230003456e-05,
      "loss": 0.6273,
      "step": 2246
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2889302562909214e-05,
      "loss": 0.5978,
      "step": 2247
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2883443813930153e-05,
      "loss": 0.4585,
      "step": 2248
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.2877583985260054e-05,
      "loss": 0.5506,
      "step": 2249
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2871723079093101e-05,
      "loss": 0.511,
      "step": 2250
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2865861097623881e-05,
      "loss": 0.513,
      "step": 2251
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2859998043047383e-05,
      "loss": 0.4954,
      "step": 2252
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2854133917558996e-05,
      "loss": 0.603,
      "step": 2253
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2848268723354518e-05,
      "loss": 0.6588,
      "step": 2254
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2842402462630137e-05,
      "loss": 0.6231,
      "step": 2255
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2836535137582442e-05,
      "loss": 0.5122,
      "step": 2256
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2830666750408434e-05,
      "loss": 0.4174,
      "step": 2257
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.282479730330549e-05,
      "loss": 0.6223,
      "step": 2258
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2818926798471398e-05,
      "loss": 0.626,
      "step": 2259
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2813055238104344e-05,
      "loss": 0.6053,
      "step": 2260
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.28071826244029e-05,
      "loss": 0.579,
      "step": 2261
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2801308959566034e-05,
      "loss": 0.6077,
      "step": 2262
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2795434245793114e-05,
      "loss": 0.7164,
      "step": 2263
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2789558485283897e-05,
      "loss": 0.734,
      "step": 2264
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2783681680238532e-05,
      "loss": 0.4297,
      "step": 2265
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2777803832857558e-05,
      "loss": 0.6674,
      "step": 2266
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2771924945341906e-05,
      "loss": 0.5654,
      "step": 2267
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2766045019892897e-05,
      "loss": 0.516,
      "step": 2268
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.276016405871224e-05,
      "loss": 0.4901,
      "step": 2269
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2754282064002026e-05,
      "loss": 0.5794,
      "step": 2270
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2748399037964742e-05,
      "loss": 0.5879,
      "step": 2271
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2742514982803257e-05,
      "loss": 0.5863,
      "step": 2272
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2736629900720832e-05,
      "loss": 0.5113,
      "step": 2273
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2730743793921094e-05,
      "loss": 0.6055,
      "step": 2274
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2724856664608077e-05,
      "loss": 0.5831,
      "step": 2275
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2718968514986183e-05,
      "loss": 0.557,
      "step": 2276
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2713079347260198e-05,
      "loss": 0.7647,
      "step": 2277
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2707189163635292e-05,
      "loss": 0.6873,
      "step": 2278
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.270129796631702e-05,
      "loss": 0.5424,
      "step": 2279
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2695405757511303e-05,
      "loss": 0.6672,
      "step": 2280
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2689512539424454e-05,
      "loss": 0.5454,
      "step": 2281
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2683618314263158e-05,
      "loss": 0.546,
      "step": 2282
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2677723084234479e-05,
      "loss": 0.5113,
      "step": 2283
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2671826851545851e-05,
      "loss": 0.502,
      "step": 2284
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2665929618405097e-05,
      "loss": 0.6416,
      "step": 2285
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2660031387020401e-05,
      "loss": 0.5054,
      "step": 2286
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2654132159600327e-05,
      "loss": 0.5501,
      "step": 2287
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.264823193835381e-05,
      "loss": 0.6436,
      "step": 2288
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2642330725490161e-05,
      "loss": 0.4724,
      "step": 2289
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2636428523219058e-05,
      "loss": 0.5366,
      "step": 2290
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2630525333750548e-05,
      "loss": 0.4538,
      "step": 2291
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.262462115929506e-05,
      "loss": 0.5906,
      "step": 2292
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2618716002063368e-05,
      "loss": 0.3925,
      "step": 2293
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2612809864266637e-05,
      "loss": 0.5561,
      "step": 2294
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2606902748116394e-05,
      "loss": 0.6806,
      "step": 2295
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2600994655824523e-05,
      "loss": 0.6468,
      "step": 2296
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2595085589603281e-05,
      "loss": 0.48,
      "step": 2297
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2589175551665287e-05,
      "loss": 0.6967,
      "step": 2298
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.258326454422353e-05,
      "loss": 0.6273,
      "step": 2299
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.257735256949135e-05,
      "loss": 0.5908,
      "step": 2300
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.257143962968246e-05,
      "loss": 0.6058,
      "step": 2301
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.2565525727010932e-05,
      "loss": 0.4691,
      "step": 2302
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2559610863691193e-05,
      "loss": 0.7054,
      "step": 2303
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2553695041938039e-05,
      "loss": 0.4603,
      "step": 2304
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2547778263966617e-05,
      "loss": 0.486,
      "step": 2305
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.254186053199243e-05,
      "loss": 0.5668,
      "step": 2306
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2535941848231352e-05,
      "loss": 0.6506,
      "step": 2307
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.25300222148996e-05,
      "loss": 0.7455,
      "step": 2308
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.252410163421375e-05,
      "loss": 0.3753,
      "step": 2309
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2518180108390732e-05,
      "loss": 0.608,
      "step": 2310
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2512257639647838e-05,
      "loss": 0.5122,
      "step": 2311
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2506334230202698e-05,
      "loss": 0.4807,
      "step": 2312
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2500409882273312e-05,
      "loss": 0.3758,
      "step": 2313
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2494484598078016e-05,
      "loss": 0.6126,
      "step": 2314
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2488558379835507e-05,
      "loss": 0.5271,
      "step": 2315
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2482631229764824e-05,
      "loss": 0.781,
      "step": 2316
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2476703150085356e-05,
      "loss": 0.57,
      "step": 2317
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2470774143016854e-05,
      "loss": 0.6225,
      "step": 2318
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2464844210779392e-05,
      "loss": 0.4379,
      "step": 2319
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2458913355593412e-05,
      "loss": 0.7705,
      "step": 2320
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2452981579679688e-05,
      "loss": 0.4015,
      "step": 2321
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2447048885259347e-05,
      "loss": 0.6594,
      "step": 2322
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2441115274553854e-05,
      "loss": 0.5562,
      "step": 2323
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2435180749785021e-05,
      "loss": 0.46,
      "step": 2324
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2429245313175e-05,
      "loss": 0.5096,
      "step": 2325
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2423308966946284e-05,
      "loss": 0.6311,
      "step": 2326
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2417371713321713e-05,
      "loss": 0.5128,
      "step": 2327
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2411433554524457e-05,
      "loss": 0.4396,
      "step": 2328
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2405494492778034e-05,
      "loss": 0.5601,
      "step": 2329
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2399554530306293e-05,
      "loss": 0.5578,
      "step": 2330
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2393613669333422e-05,
      "loss": 0.5396,
      "step": 2331
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2387671912083947e-05,
      "loss": 0.5198,
      "step": 2332
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2381729260782728e-05,
      "loss": 0.4854,
      "step": 2333
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2375785717654968e-05,
      "loss": 0.599,
      "step": 2334
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.236984128492619e-05,
      "loss": 0.5882,
      "step": 2335
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2363895964822259e-05,
      "loss": 0.6558,
      "step": 2336
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2357949759569372e-05,
      "loss": 0.4994,
      "step": 2337
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2352002671394047e-05,
      "loss": 0.6359,
      "step": 2338
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2346054702523154e-05,
      "loss": 0.5457,
      "step": 2339
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2340105855183872e-05,
      "loss": 0.5162,
      "step": 2340
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2334156131603718e-05,
      "loss": 0.7098,
      "step": 2341
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2328205534010538e-05,
      "loss": 0.5118,
      "step": 2342
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.23222540646325e-05,
      "loss": 0.6384,
      "step": 2343
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.231630172569811e-05,
      "loss": 0.6398,
      "step": 2344
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.231034851943618e-05,
      "loss": 0.4625,
      "step": 2345
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2304394448075864e-05,
      "loss": 0.5919,
      "step": 2346
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2298439513846634e-05,
      "loss": 0.6371,
      "step": 2347
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2292483718978288e-05,
      "loss": 0.5887,
      "step": 2348
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2286527065700938e-05,
      "loss": 0.681,
      "step": 2349
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2280569556245026e-05,
      "loss": 0.4429,
      "step": 2350
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2274611192841309e-05,
      "loss": 0.5462,
      "step": 2351
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2268651977720867e-05,
      "loss": 0.5795,
      "step": 2352
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2262691913115103e-05,
      "loss": 0.6146,
      "step": 2353
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2256731001255723e-05,
      "loss": 0.4699,
      "step": 2354
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.2250769244374772e-05,
      "loss": 0.6616,
      "step": 2355
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2244806644704594e-05,
      "loss": 0.7125,
      "step": 2356
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2238843204477855e-05,
      "loss": 0.6338,
      "step": 2357
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2232878925927533e-05,
      "loss": 0.7599,
      "step": 2358
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2226913811286924e-05,
      "loss": 0.5509,
      "step": 2359
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2220947862789641e-05,
      "loss": 0.5225,
      "step": 2360
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2214981082669593e-05,
      "loss": 0.5384,
      "step": 2361
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2209013473161017e-05,
      "loss": 0.495,
      "step": 2362
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2203045036498456e-05,
      "loss": 0.3722,
      "step": 2363
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2197075774916754e-05,
      "loss": 0.5464,
      "step": 2364
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2191105690651078e-05,
      "loss": 0.5805,
      "step": 2365
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2185134785936892e-05,
      "loss": 0.5471,
      "step": 2366
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2179163063009974e-05,
      "loss": 0.7299,
      "step": 2367
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.21731905241064e-05,
      "loss": 0.5153,
      "step": 2368
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2167217171462566e-05,
      "loss": 0.439,
      "step": 2369
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2161243007315158e-05,
      "loss": 0.5231,
      "step": 2370
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.215526803390117e-05,
      "loss": 0.6858,
      "step": 2371
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.214929225345791e-05,
      "loss": 0.4786,
      "step": 2372
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2143315668222966e-05,
      "loss": 0.5186,
      "step": 2373
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.213733828043425e-05,
      "loss": 0.6726,
      "step": 2374
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2131360092329962e-05,
      "loss": 0.5036,
      "step": 2375
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2125381106148604e-05,
      "loss": 0.5348,
      "step": 2376
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2119401324128976e-05,
      "loss": 0.6589,
      "step": 2377
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.211342074851018e-05,
      "loss": 0.4281,
      "step": 2378
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2107439381531612e-05,
      "loss": 0.4938,
      "step": 2379
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.210145722543296e-05,
      "loss": 0.5383,
      "step": 2380
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.209547428245422e-05,
      "loss": 0.6894,
      "step": 2381
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.208949055483567e-05,
      "loss": 0.7322,
      "step": 2382
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2083506044817889e-05,
      "loss": 0.6169,
      "step": 2383
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2077520754641745e-05,
      "loss": 0.8171,
      "step": 2384
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.20715346865484e-05,
      "loss": 0.5123,
      "step": 2385
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.206554784277931e-05,
      "loss": 0.4449,
      "step": 2386
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2059560225576212e-05,
      "loss": 0.4905,
      "step": 2387
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2053571837181148e-05,
      "loss": 0.5048,
      "step": 2388
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2047582679836432e-05,
      "loss": 0.7448,
      "step": 2389
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2041592755784676e-05,
      "loss": 0.5102,
      "step": 2390
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2035602067268781e-05,
      "loss": 0.5747,
      "step": 2391
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2029610616531926e-05,
      "loss": 0.4384,
      "step": 2392
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.202361840581758e-05,
      "loss": 0.661,
      "step": 2393
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2017625437369498e-05,
      "loss": 0.4807,
      "step": 2394
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2011631713431718e-05,
      "loss": 0.6163,
      "step": 2395
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.2005637236248555e-05,
      "loss": 0.613,
      "step": 2396
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1999642008064612e-05,
      "loss": 0.3455,
      "step": 2397
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1993646031124775e-05,
      "loss": 0.5827,
      "step": 2398
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1987649307674206e-05,
      "loss": 0.5537,
      "step": 2399
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1981651839958349e-05,
      "loss": 0.5313,
      "step": 2400
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1975653630222924e-05,
      "loss": 0.5858,
      "step": 2401
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.196965468071393e-05,
      "loss": 0.6632,
      "step": 2402
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1963654993677645e-05,
      "loss": 0.5366,
      "step": 2403
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.195765457136062e-05,
      "loss": 0.4409,
      "step": 2404
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1951653416009686e-05,
      "loss": 0.4915,
      "step": 2405
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.194565152987194e-05,
      "loss": 0.5773,
      "step": 2406
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1939648915194766e-05,
      "loss": 0.6235,
      "step": 2407
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.1933645574225805e-05,
      "loss": 0.6017,
      "step": 2408
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1927641509212984e-05,
      "loss": 0.3602,
      "step": 2409
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1921636722404493e-05,
      "loss": 0.434,
      "step": 2410
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.191563121604879e-05,
      "loss": 0.54,
      "step": 2411
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1909624992394613e-05,
      "loss": 0.4384,
      "step": 2412
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1903618053690957e-05,
      "loss": 0.7469,
      "step": 2413
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1897610402187097e-05,
      "loss": 0.5281,
      "step": 2414
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1891602040132557e-05,
      "loss": 0.484,
      "step": 2415
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1885592969777149e-05,
      "loss": 0.5625,
      "step": 2416
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1879583193370934e-05,
      "loss": 0.5393,
      "step": 2417
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1873572713164241e-05,
      "loss": 0.6545,
      "step": 2418
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1867561531407668e-05,
      "loss": 0.5308,
      "step": 2419
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1861549650352069e-05,
      "loss": 0.6603,
      "step": 2420
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1855537072248564e-05,
      "loss": 0.4867,
      "step": 2421
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.184952379934853e-05,
      "loss": 0.5423,
      "step": 2422
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.184350983390361e-05,
      "loss": 0.649,
      "step": 2423
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1837495178165706e-05,
      "loss": 0.5333,
      "step": 2424
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1831479834386967e-05,
      "loss": 0.5127,
      "step": 2425
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1825463804819814e-05,
      "loss": 0.5891,
      "step": 2426
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1819447091716918e-05,
      "loss": 0.6252,
      "step": 2427
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.181342969733121e-05,
      "loss": 0.654,
      "step": 2428
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1807411623915865e-05,
      "loss": 0.5431,
      "step": 2429
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.180139287372433e-05,
      "loss": 0.5815,
      "step": 2430
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1795373449010288e-05,
      "loss": 0.6465,
      "step": 2431
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1789353352027683e-05,
      "loss": 0.5623,
      "step": 2432
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1783332585030717e-05,
      "loss": 0.5684,
      "step": 2433
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.177731115027383e-05,
      "loss": 0.5758,
      "step": 2434
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.177128905001172e-05,
      "loss": 0.5583,
      "step": 2435
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.176526628649933e-05,
      "loss": 0.6128,
      "step": 2436
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1759242861991855e-05,
      "loss": 0.5591,
      "step": 2437
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1753218778744736e-05,
      "loss": 0.6615,
      "step": 2438
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.174719403901366e-05,
      "loss": 0.59,
      "step": 2439
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1741168645054566e-05,
      "loss": 0.6926,
      "step": 2440
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1735142599123624e-05,
      "loss": 0.4956,
      "step": 2441
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.172911590347726e-05,
      "loss": 0.58,
      "step": 2442
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.172308856037214e-05,
      "loss": 0.6934,
      "step": 2443
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1717060572065174e-05,
      "loss": 0.4797,
      "step": 2444
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.171103194081351e-05,
      "loss": 0.5762,
      "step": 2445
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1705002668874538e-05,
      "loss": 0.4984,
      "step": 2446
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1698972758505891e-05,
      "loss": 0.6153,
      "step": 2447
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1692942211965434e-05,
      "loss": 0.5156,
      "step": 2448
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1686911031511284e-05,
      "loss": 0.6096,
      "step": 2449
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1680879219401774e-05,
      "loss": 0.5142,
      "step": 2450
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1674846777895494e-05,
      "loss": 0.5321,
      "step": 2451
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1668813709251263e-05,
      "loss": 0.5817,
      "step": 2452
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1662780015728123e-05,
      "loss": 0.4443,
      "step": 2453
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1656745699585373e-05,
      "loss": 0.584,
      "step": 2454
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.165071076308252e-05,
      "loss": 0.6176,
      "step": 2455
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1644675208479329e-05,
      "loss": 0.6359,
      "step": 2456
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1638639038035771e-05,
      "loss": 0.5163,
      "step": 2457
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1632602254012064e-05,
      "loss": 0.6711,
      "step": 2458
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1626564858668657e-05,
      "loss": 0.4363,
      "step": 2459
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.162052685426621e-05,
      "loss": 0.566,
      "step": 2460
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.1614488243065636e-05,
      "loss": 0.4326,
      "step": 2461
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1608449027328056e-05,
      "loss": 0.4422,
      "step": 2462
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1602409209314825e-05,
      "loss": 0.6701,
      "step": 2463
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1596368791287522e-05,
      "loss": 0.617,
      "step": 2464
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1590327775507952e-05,
      "loss": 0.5051,
      "step": 2465
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1584286164238144e-05,
      "loss": 0.5861,
      "step": 2466
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1578243959740345e-05,
      "loss": 0.4784,
      "step": 2467
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1572201164277033e-05,
      "loss": 0.4607,
      "step": 2468
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.15661577801109e-05,
      "loss": 0.5439,
      "step": 2469
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.156011380950486e-05,
      "loss": 0.5271,
      "step": 2470
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.155406925472205e-05,
      "loss": 0.5937,
      "step": 2471
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1548024118025822e-05,
      "loss": 0.8131,
      "step": 2472
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1541978401679746e-05,
      "loss": 0.708,
      "step": 2473
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.153593210794761e-05,
      "loss": 0.5336,
      "step": 2474
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1529885239093422e-05,
      "loss": 0.5452,
      "step": 2475
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1523837797381395e-05,
      "loss": 0.5366,
      "step": 2476
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1517789785075965e-05,
      "loss": 0.6255,
      "step": 2477
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1511741204441789e-05,
      "loss": 0.5804,
      "step": 2478
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1505692057743714e-05,
      "loss": 0.6302,
      "step": 2479
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.149964234724682e-05,
      "loss": 0.7076,
      "step": 2480
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1493592075216391e-05,
      "loss": 0.5908,
      "step": 2481
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.148754124391792e-05,
      "loss": 0.4432,
      "step": 2482
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1481489855617108e-05,
      "loss": 0.5953,
      "step": 2483
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1475437912579868e-05,
      "loss": 0.6158,
      "step": 2484
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1469385417072323e-05,
      "loss": 0.5615,
      "step": 2485
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1463332371360794e-05,
      "loss": 0.4911,
      "step": 2486
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1457278777711816e-05,
      "loss": 0.3982,
      "step": 2487
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.145122463839213e-05,
      "loss": 0.5525,
      "step": 2488
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1445169955668674e-05,
      "loss": 0.5322,
      "step": 2489
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1439114731808594e-05,
      "loss": 0.6147,
      "step": 2490
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1433058969079239e-05,
      "loss": 0.5692,
      "step": 2491
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.142700266974816e-05,
      "loss": 0.6542,
      "step": 2492
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.14209458360831e-05,
      "loss": 0.4109,
      "step": 2493
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1414888470352026e-05,
      "loss": 0.5844,
      "step": 2494
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1408830574823074e-05,
      "loss": 0.4066,
      "step": 2495
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1402772151764596e-05,
      "loss": 0.6047,
      "step": 2496
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.139671320344514e-05,
      "loss": 0.6142,
      "step": 2497
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1390653732133447e-05,
      "loss": 0.4415,
      "step": 2498
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.138459374009846e-05,
      "loss": 0.6104,
      "step": 2499
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1378533229609304e-05,
      "loss": 0.5457,
      "step": 2500
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1372472202935315e-05,
      "loss": 0.7463,
      "step": 2501
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1366410662346008e-05,
      "loss": 0.4772,
      "step": 2502
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.13603486101111e-05,
      "loss": 0.5744,
      "step": 2503
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1354286048500496e-05,
      "loss": 0.4232,
      "step": 2504
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1348222979784289e-05,
      "loss": 0.536,
      "step": 2505
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1342159406232769e-05,
      "loss": 0.5588,
      "step": 2506
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1336095330116406e-05,
      "loss": 0.6071,
      "step": 2507
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1330030753705865e-05,
      "loss": 0.6406,
      "step": 2508
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1323965679271993e-05,
      "loss": 0.5207,
      "step": 2509
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1317900109085835e-05,
      "loss": 0.6309,
      "step": 2510
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1311834045418607e-05,
      "loss": 0.5369,
      "step": 2511
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.130576749054171e-05,
      "loss": 0.5741,
      "step": 2512
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.129970044672675e-05,
      "loss": 0.5945,
      "step": 2513
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.1293632916245487e-05,
      "loss": 0.6166,
      "step": 2514
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1287564901369886e-05,
      "loss": 0.606,
      "step": 2515
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1281496404372082e-05,
      "loss": 0.5601,
      "step": 2516
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.127542742752439e-05,
      "loss": 0.5089,
      "step": 2517
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1269357973099314e-05,
      "loss": 0.512,
      "step": 2518
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1263288043369522e-05,
      "loss": 0.5219,
      "step": 2519
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1257217640607883e-05,
      "loss": 0.4488,
      "step": 2520
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.125114676708741e-05,
      "loss": 0.509,
      "step": 2521
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1245075425081328e-05,
      "loss": 0.5134,
      "step": 2522
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1239003616863011e-05,
      "loss": 0.4998,
      "step": 2523
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1232931344706021e-05,
      "loss": 0.4653,
      "step": 2524
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1226858610884088e-05,
      "loss": 0.5476,
      "step": 2525
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1220785417671116e-05,
      "loss": 0.5864,
      "step": 2526
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1214711767341184e-05,
      "loss": 0.5447,
      "step": 2527
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1208637662168536e-05,
      "loss": 0.4657,
      "step": 2528
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1202563104427596e-05,
      "loss": 0.6225,
      "step": 2529
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1196488096392943e-05,
      "loss": 0.6471,
      "step": 2530
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1190412640339344e-05,
      "loss": 0.5144,
      "step": 2531
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1184336738541714e-05,
      "loss": 0.5635,
      "step": 2532
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.117826039327515e-05,
      "loss": 0.5097,
      "step": 2533
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1172183606814903e-05,
      "loss": 0.5868,
      "step": 2534
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1166106381436396e-05,
      "loss": 0.6885,
      "step": 2535
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1160028719415226e-05,
      "loss": 0.4371,
      "step": 2536
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1153950623027127e-05,
      "loss": 0.567,
      "step": 2537
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1147872094548022e-05,
      "loss": 0.5322,
      "step": 2538
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1141793136253987e-05,
      "loss": 0.4348,
      "step": 2539
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1135713750421248e-05,
      "loss": 0.6973,
      "step": 2540
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1129633939326207e-05,
      "loss": 0.5914,
      "step": 2541
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1123553705245418e-05,
      "loss": 0.5581,
      "step": 2542
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1117473050455594e-05,
      "loss": 0.7031,
      "step": 2543
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1111391977233602e-05,
      "loss": 0.599,
      "step": 2544
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1105310487856477e-05,
      "loss": 0.4786,
      "step": 2545
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1099228584601397e-05,
      "loss": 0.5386,
      "step": 2546
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1093146269745694e-05,
      "loss": 0.5962,
      "step": 2547
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1087063545566872e-05,
      "loss": 0.5773,
      "step": 2548
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.108098041434257e-05,
      "loss": 0.4579,
      "step": 2549
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.107489687835059e-05,
      "loss": 0.7348,
      "step": 2550
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1068812939868876e-05,
      "loss": 0.5018,
      "step": 2551
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1062728601175532e-05,
      "loss": 0.5074,
      "step": 2552
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1056643864548805e-05,
      "loss": 0.5604,
      "step": 2553
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1050558732267095e-05,
      "loss": 0.5107,
      "step": 2554
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1044473206608954e-05,
      "loss": 0.4421,
      "step": 2555
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1038387289853069e-05,
      "loss": 0.417,
      "step": 2556
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1032300984278286e-05,
      "loss": 0.4875,
      "step": 2557
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1026214292163591e-05,
      "loss": 0.5871,
      "step": 2558
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1020127215788112e-05,
      "loss": 0.6317,
      "step": 2559
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1014039757431127e-05,
      "loss": 0.5468,
      "step": 2560
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1007951919372056e-05,
      "loss": 0.4433,
      "step": 2561
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.1001863703890453e-05,
      "loss": 0.5657,
      "step": 2562
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.0995775113266026e-05,
      "loss": 0.6984,
      "step": 2563
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.0989686149778613e-05,
      "loss": 0.6332,
      "step": 2564
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.0983596815708202e-05,
      "loss": 0.5989,
      "step": 2565
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.0977507113334903e-05,
      "loss": 0.4455,
      "step": 2566
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.0971417044938984e-05,
      "loss": 0.8571,
      "step": 2567
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0965326612800835e-05,
      "loss": 0.4933,
      "step": 2568
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.095923581920099e-05,
      "loss": 0.5875,
      "step": 2569
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0953144666420114e-05,
      "loss": 0.5361,
      "step": 2570
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0947053156739016e-05,
      "loss": 0.6399,
      "step": 2571
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0940961292438624e-05,
      "loss": 0.5091,
      "step": 2572
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.09348690758e-05,
      "loss": 0.5135,
      "step": 2573
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.092877650910436e-05,
      "loss": 0.5605,
      "step": 2574
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.092268359463302e-05,
      "loss": 0.4201,
      "step": 2575
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0916590334667451e-05,
      "loss": 0.5038,
      "step": 2576
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.091049673148924e-05,
      "loss": 0.5358,
      "step": 2577
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0904402787380104e-05,
      "loss": 0.4974,
      "step": 2578
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0898308504621893e-05,
      "loss": 0.597,
      "step": 2579
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0892213885496577e-05,
      "loss": 0.556,
      "step": 2580
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0886118932286263e-05,
      "loss": 0.7047,
      "step": 2581
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0880023647273163e-05,
      "loss": 0.628,
      "step": 2582
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0873928032739639e-05,
      "loss": 0.5711,
      "step": 2583
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0867832090968154e-05,
      "loss": 0.445,
      "step": 2584
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0861735824241309e-05,
      "loss": 0.5686,
      "step": 2585
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0855639234841816e-05,
      "loss": 0.4484,
      "step": 2586
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0849542325052514e-05,
      "loss": 0.5434,
      "step": 2587
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.084344509715636e-05,
      "loss": 0.5754,
      "step": 2588
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0837347553436426e-05,
      "loss": 0.3622,
      "step": 2589
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0831249696175918e-05,
      "loss": 0.4423,
      "step": 2590
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0825151527658135e-05,
      "loss": 0.71,
      "step": 2591
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0819053050166515e-05,
      "loss": 0.5361,
      "step": 2592
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0812954265984599e-05,
      "loss": 0.5501,
      "step": 2593
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0806855177396045e-05,
      "loss": 0.4335,
      "step": 2594
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0800755786684627e-05,
      "loss": 0.7452,
      "step": 2595
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0794656096134228e-05,
      "loss": 0.5148,
      "step": 2596
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0788556108028854e-05,
      "loss": 0.5997,
      "step": 2597
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.078245582465261e-05,
      "loss": 0.5782,
      "step": 2598
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0776355248289712e-05,
      "loss": 0.4868,
      "step": 2599
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0770254381224501e-05,
      "loss": 0.6682,
      "step": 2600
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0764153225741403e-05,
      "loss": 0.4333,
      "step": 2601
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0758051784124976e-05,
      "loss": 0.7171,
      "step": 2602
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.075195005865987e-05,
      "loss": 0.5698,
      "step": 2603
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0745848051630845e-05,
      "loss": 0.6157,
      "step": 2604
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0739745765322768e-05,
      "loss": 0.5601,
      "step": 2605
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0733643202020606e-05,
      "loss": 0.7661,
      "step": 2606
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.072754036400944e-05,
      "loss": 0.5412,
      "step": 2607
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0721437253574439e-05,
      "loss": 0.5201,
      "step": 2608
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0715333873000888e-05,
      "loss": 0.6437,
      "step": 2609
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0709230224574165e-05,
      "loss": 0.4681,
      "step": 2610
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.070312631057975e-05,
      "loss": 0.5601,
      "step": 2611
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0697022133303223e-05,
      "loss": 0.7075,
      "step": 2612
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0690917695030265e-05,
      "loss": 0.6852,
      "step": 2613
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0684812998046647e-05,
      "loss": 0.473,
      "step": 2614
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0678708044638247e-05,
      "loss": 0.5388,
      "step": 2615
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0672602837091032e-05,
      "loss": 0.431,
      "step": 2616
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0666497377691067e-05,
      "loss": 0.5561,
      "step": 2617
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.066039166872451e-05,
      "loss": 0.5045,
      "step": 2618
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.0654285712477613e-05,
      "loss": 0.4898,
      "step": 2619
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.064817951123672e-05,
      "loss": 0.4481,
      "step": 2620
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.064207306728827e-05,
      "loss": 0.5216,
      "step": 2621
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0635966382918787e-05,
      "loss": 0.573,
      "step": 2622
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0629859460414892e-05,
      "loss": 0.464,
      "step": 2623
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0623752302063284e-05,
      "loss": 0.5751,
      "step": 2624
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0617644910150766e-05,
      "loss": 0.4379,
      "step": 2625
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.061153728696422e-05,
      "loss": 0.5347,
      "step": 2626
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0605429434790607e-05,
      "loss": 0.499,
      "step": 2627
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0599321355916988e-05,
      "loss": 0.4588,
      "step": 2628
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.05932130526305e-05,
      "loss": 0.4768,
      "step": 2629
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0587104527218368e-05,
      "loss": 0.5436,
      "step": 2630
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0580995781967898e-05,
      "loss": 0.4108,
      "step": 2631
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0574886819166479e-05,
      "loss": 0.4045,
      "step": 2632
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.056877764110158e-05,
      "loss": 0.4543,
      "step": 2633
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0562668250060752e-05,
      "loss": 0.4583,
      "step": 2634
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.055655864833163e-05,
      "loss": 0.3974,
      "step": 2635
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0550448838201918e-05,
      "loss": 0.4443,
      "step": 2636
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0544338821959407e-05,
      "loss": 0.4689,
      "step": 2637
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0538228601891963e-05,
      "loss": 0.452,
      "step": 2638
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0532118180287523e-05,
      "loss": 0.7442,
      "step": 2639
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0526007559434107e-05,
      "loss": 0.4532,
      "step": 2640
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0519896741619803e-05,
      "loss": 0.5434,
      "step": 2641
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0513785729132785e-05,
      "loss": 0.5419,
      "step": 2642
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0507674524261275e-05,
      "loss": 0.4588,
      "step": 2643
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0501563129293598e-05,
      "loss": 0.7093,
      "step": 2644
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0495451546518127e-05,
      "loss": 0.4982,
      "step": 2645
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0489339778223317e-05,
      "loss": 0.5438,
      "step": 2646
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0483227826697686e-05,
      "loss": 0.4707,
      "step": 2647
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0477115694229828e-05,
      "loss": 0.722,
      "step": 2648
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0471003383108394e-05,
      "loss": 0.6375,
      "step": 2649
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0464890895622111e-05,
      "loss": 0.6374,
      "step": 2650
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0458778234059773e-05,
      "loss": 0.5245,
      "step": 2651
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0452665400710228e-05,
      "loss": 0.5664,
      "step": 2652
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0446552397862402e-05,
      "loss": 0.4537,
      "step": 2653
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0440439227805279e-05,
      "loss": 0.4842,
      "step": 2654
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0434325892827896e-05,
      "loss": 0.4891,
      "step": 2655
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.042821239521937e-05,
      "loss": 0.4236,
      "step": 2656
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0422098737268862e-05,
      "loss": 0.44,
      "step": 2657
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.041598492126561e-05,
      "loss": 0.4536,
      "step": 2658
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0409870949498892e-05,
      "loss": 0.4361,
      "step": 2659
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0403756824258057e-05,
      "loss": 0.366,
      "step": 2660
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0397642547832514e-05,
      "loss": 0.5477,
      "step": 2661
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0391528122511712e-05,
      "loss": 0.5323,
      "step": 2662
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0385413550585178e-05,
      "loss": 0.5425,
      "step": 2663
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0379298834342475e-05,
      "loss": 0.6236,
      "step": 2664
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0373183976073234e-05,
      "loss": 0.7944,
      "step": 2665
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0367068978067125e-05,
      "loss": 0.5758,
      "step": 2666
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0360953842613886e-05,
      "loss": 0.4301,
      "step": 2667
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0354838572003296e-05,
      "loss": 0.578,
      "step": 2668
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0348723168525183e-05,
      "loss": 0.5863,
      "step": 2669
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0342607634469439e-05,
      "loss": 0.4548,
      "step": 2670
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0336491972125984e-05,
      "loss": 0.5234,
      "step": 2671
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.0330376183784802e-05,
      "loss": 0.5687,
      "step": 2672
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0324260271735923e-05,
      "loss": 0.5575,
      "step": 2673
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0318144238269412e-05,
      "loss": 0.4902,
      "step": 2674
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0312028085675393e-05,
      "loss": 0.6113,
      "step": 2675
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0305911816244024e-05,
      "loss": 0.4576,
      "step": 2676
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0299795432265516e-05,
      "loss": 0.467,
      "step": 2677
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0293678936030114e-05,
      "loss": 0.4824,
      "step": 2678
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0287562329828112e-05,
      "loss": 0.6147,
      "step": 2679
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0281445615949843e-05,
      "loss": 0.5824,
      "step": 2680
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0275328796685674e-05,
      "loss": 0.5046,
      "step": 2681
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0269211874326024e-05,
      "loss": 0.6117,
      "step": 2682
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0263094851161342e-05,
      "loss": 0.4939,
      "step": 2683
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0256977729482115e-05,
      "loss": 0.5119,
      "step": 2684
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.025086051157887e-05,
      "loss": 0.4434,
      "step": 2685
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0244743199742166e-05,
      "loss": 0.6273,
      "step": 2686
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0238625796262604e-05,
      "loss": 0.5254,
      "step": 2687
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0232508303430811e-05,
      "loss": 0.6646,
      "step": 2688
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0226390723537451e-05,
      "loss": 0.608,
      "step": 2689
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0220273058873225e-05,
      "loss": 0.5349,
      "step": 2690
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.021415531172886e-05,
      "loss": 0.5692,
      "step": 2691
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0208037484395114e-05,
      "loss": 0.5361,
      "step": 2692
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.020191957916278e-05,
      "loss": 0.5337,
      "step": 2693
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0195801598322674e-05,
      "loss": 0.4988,
      "step": 2694
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0189683544165643e-05,
      "loss": 0.6059,
      "step": 2695
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0183565418982563e-05,
      "loss": 0.5928,
      "step": 2696
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0177447225064334e-05,
      "loss": 0.413,
      "step": 2697
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0171328964701885e-05,
      "loss": 0.4608,
      "step": 2698
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0165210640186168e-05,
      "loss": 0.5546,
      "step": 2699
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0159092253808156e-05,
      "loss": 0.5403,
      "step": 2700
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.015297380785885e-05,
      "loss": 0.4064,
      "step": 2701
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.014685530462927e-05,
      "loss": 0.574,
      "step": 2702
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0140736746410462e-05,
      "loss": 0.563,
      "step": 2703
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0134618135493481e-05,
      "loss": 0.4162,
      "step": 2704
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0128499474169422e-05,
      "loss": 0.5671,
      "step": 2705
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0122380764729378e-05,
      "loss": 0.487,
      "step": 2706
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0116262009464475e-05,
      "loss": 0.4393,
      "step": 2707
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0110143210665843e-05,
      "loss": 0.6862,
      "step": 2708
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0104024370624644e-05,
      "loss": 0.5551,
      "step": 2709
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.009790549163204e-05,
      "loss": 0.5777,
      "step": 2710
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0091786575979215e-05,
      "loss": 0.3727,
      "step": 2711
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0085667625957372e-05,
      "loss": 0.3504,
      "step": 2712
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0079548643857718e-05,
      "loss": 0.5351,
      "step": 2713
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0073429631971467e-05,
      "loss": 0.465,
      "step": 2714
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0067310592589867e-05,
      "loss": 0.4409,
      "step": 2715
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.006119152800415e-05,
      "loss": 0.5686,
      "step": 2716
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0055072440505576e-05,
      "loss": 0.5794,
      "step": 2717
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0048953332385401e-05,
      "loss": 0.5605,
      "step": 2718
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0042834205934897e-05,
      "loss": 0.4053,
      "step": 2719
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0036715063445342e-05,
      "loss": 0.6097,
      "step": 2720
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0030595907208012e-05,
      "loss": 0.6077,
      "step": 2721
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0024476739514205e-05,
      "loss": 0.4294,
      "step": 2722
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.0018357562655199e-05,
      "loss": 0.5216,
      "step": 2723
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.00122383789223e-05,
      "loss": 0.404,
      "step": 2724
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.00061191906068e-05,
      "loss": 0.6293,
      "step": 2725
    },
    {
      "epoch": 0.52,
      "learning_rate": 1e-05,
      "loss": 0.6528,
      "step": 2726
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.993880809393203e-06,
      "loss": 0.4957,
      "step": 2727
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.9877616210777e-06,
      "loss": 0.4008,
      "step": 2728
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.981642437344804e-06,
      "loss": 0.5579,
      "step": 2729
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.9755232604858e-06,
      "loss": 0.5913,
      "step": 2730
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.969404092791987e-06,
      "loss": 0.6209,
      "step": 2731
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.963284936554661e-06,
      "loss": 0.5417,
      "step": 2732
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.957165794065107e-06,
      "loss": 0.6953,
      "step": 2733
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.951046667614602e-06,
      "loss": 0.5529,
      "step": 2734
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.944927559494428e-06,
      "loss": 0.3847,
      "step": 2735
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.938808471995856e-06,
      "loss": 0.4279,
      "step": 2736
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.932689407410136e-06,
      "loss": 0.4318,
      "step": 2737
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.926570368028535e-06,
      "loss": 0.4936,
      "step": 2738
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.920451356142289e-06,
      "loss": 0.5305,
      "step": 2739
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.91433237404263e-06,
      "loss": 0.5505,
      "step": 2740
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.908213424020789e-06,
      "loss": 0.4633,
      "step": 2741
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.902094508367962e-06,
      "loss": 0.5418,
      "step": 2742
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.89597562937536e-06,
      "loss": 0.5683,
      "step": 2743
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.889856789334159e-06,
      "loss": 0.547,
      "step": 2744
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.883737990535527e-06,
      "loss": 0.4445,
      "step": 2745
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.877619235270625e-06,
      "loss": 0.5953,
      "step": 2746
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.871500525830581e-06,
      "loss": 0.4677,
      "step": 2747
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.865381864506519e-06,
      "loss": 0.5829,
      "step": 2748
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.859263253589541e-06,
      "loss": 0.4285,
      "step": 2749
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.853144695370735e-06,
      "loss": 0.4165,
      "step": 2750
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.847026192141152e-06,
      "loss": 0.5295,
      "step": 2751
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.840907746191847e-06,
      "loss": 0.4143,
      "step": 2752
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.834789359813834e-06,
      "loss": 0.5231,
      "step": 2753
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.828671035298115e-06,
      "loss": 0.5459,
      "step": 2754
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.822552774935668e-06,
      "loss": 0.4801,
      "step": 2755
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.81643458101744e-06,
      "loss": 0.5834,
      "step": 2756
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.810316455834359e-06,
      "loss": 0.7334,
      "step": 2757
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.804198401677328e-06,
      "loss": 0.6355,
      "step": 2758
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.798080420837224e-06,
      "loss": 0.4824,
      "step": 2759
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.791962515604887e-06,
      "loss": 0.4964,
      "step": 2760
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.785844688271142e-06,
      "loss": 0.5241,
      "step": 2761
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.77972694112678e-06,
      "loss": 0.4813,
      "step": 2762
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.773609276462549e-06,
      "loss": 0.4106,
      "step": 2763
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.767491696569194e-06,
      "loss": 0.5177,
      "step": 2764
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.7613742037374e-06,
      "loss": 0.5305,
      "step": 2765
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.755256800257836e-06,
      "loss": 0.7822,
      "step": 2766
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.749139488421133e-06,
      "loss": 0.444,
      "step": 2767
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.74302227051789e-06,
      "loss": 0.5793,
      "step": 2768
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.736905148838661e-06,
      "loss": 0.5255,
      "step": 2769
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.730788125673977e-06,
      "loss": 0.5149,
      "step": 2770
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.72467120331433e-06,
      "loss": 0.5083,
      "step": 2771
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.718554384050159e-06,
      "loss": 0.4244,
      "step": 2772
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.712437670171892e-06,
      "loss": 0.5039,
      "step": 2773
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.706321063969888e-06,
      "loss": 0.4176,
      "step": 2774
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.700204567734485e-06,
      "loss": 0.5816,
      "step": 2775
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.69408818375598e-06,
      "loss": 0.6281,
      "step": 2776
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.687971914324607e-06,
      "loss": 0.6714,
      "step": 2777
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.68185576173059e-06,
      "loss": 0.5102,
      "step": 2778
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.67573972826408e-06,
      "loss": 0.521,
      "step": 2779
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.669623816215198e-06,
      "loss": 0.4446,
      "step": 2780
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.66350802787402e-06,
      "loss": 0.5782,
      "step": 2781
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.657392365530566e-06,
      "loss": 0.69,
      "step": 2782
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.651276831474817e-06,
      "loss": 0.6719,
      "step": 2783
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.645161427996707e-06,
      "loss": 0.3895,
      "step": 2784
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.639046157386117e-06,
      "loss": 0.4088,
      "step": 2785
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.632931021932877e-06,
      "loss": 0.4428,
      "step": 2786
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.626816023926771e-06,
      "loss": 0.6064,
      "step": 2787
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.620701165657527e-06,
      "loss": 0.5348,
      "step": 2788
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.614586449414824e-06,
      "loss": 0.5343,
      "step": 2789
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.60847187748829e-06,
      "loss": 0.5262,
      "step": 2790
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.602357452167492e-06,
      "loss": 0.4868,
      "step": 2791
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.596243175741944e-06,
      "loss": 0.439,
      "step": 2792
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.59012905050111e-06,
      "loss": 0.5359,
      "step": 2793
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.584015078734395e-06,
      "loss": 0.5442,
      "step": 2794
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.57790126273114e-06,
      "loss": 0.5419,
      "step": 2795
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.571787604780633e-06,
      "loss": 0.5754,
      "step": 2796
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.565674107172109e-06,
      "loss": 0.5224,
      "step": 2797
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.559560772194723e-06,
      "loss": 0.5712,
      "step": 2798
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.5534476021376e-06,
      "loss": 0.6409,
      "step": 2799
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.547334599289775e-06,
      "loss": 0.4633,
      "step": 2800
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.541221765940229e-06,
      "loss": 0.4252,
      "step": 2801
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.535109104377892e-06,
      "loss": 0.497,
      "step": 2802
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.52899661689161e-06,
      "loss": 0.5062,
      "step": 2803
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.522884305770176e-06,
      "loss": 0.4586,
      "step": 2804
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.516772173302316e-06,
      "loss": 0.4393,
      "step": 2805
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.510660221776684e-06,
      "loss": 0.5711,
      "step": 2806
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.504548453481875e-06,
      "loss": 0.5228,
      "step": 2807
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.498436870706406e-06,
      "loss": 0.6446,
      "step": 2808
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.492325475738725e-06,
      "loss": 0.4771,
      "step": 2809
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.486214270867219e-06,
      "loss": 0.551,
      "step": 2810
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.480103258380198e-06,
      "loss": 0.4328,
      "step": 2811
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.473992440565896e-06,
      "loss": 0.427,
      "step": 2812
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.467881819712478e-06,
      "loss": 0.5902,
      "step": 2813
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.46177139810804e-06,
      "loss": 0.4928,
      "step": 2814
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.455661178040593e-06,
      "loss": 0.4944,
      "step": 2815
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.449551161798085e-06,
      "loss": 0.6,
      "step": 2816
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.443441351668375e-06,
      "loss": 0.589,
      "step": 2817
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.43733174993925e-06,
      "loss": 0.5031,
      "step": 2818
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.431222358898423e-06,
      "loss": 0.5462,
      "step": 2819
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.425113180833528e-06,
      "loss": 0.4103,
      "step": 2820
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.419004218032105e-06,
      "loss": 0.5538,
      "step": 2821
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.412895472781635e-06,
      "loss": 0.5299,
      "step": 2822
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.406786947369507e-06,
      "loss": 0.4624,
      "step": 2823
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.400678644083014e-06,
      "loss": 0.4718,
      "step": 2824
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.394570565209398e-06,
      "loss": 0.5571,
      "step": 2825
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.388462713035787e-06,
      "loss": 0.6849,
      "step": 2826
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.382355089849235e-06,
      "loss": 0.5158,
      "step": 2827
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.376247697936719e-06,
      "loss": 0.5603,
      "step": 2828
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.370140539585113e-06,
      "loss": 0.4033,
      "step": 2829
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.364033617081216e-06,
      "loss": 0.543,
      "step": 2830
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.357926932711733e-06,
      "loss": 0.6393,
      "step": 2831
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.351820488763285e-06,
      "loss": 0.5856,
      "step": 2832
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.34571428752239e-06,
      "loss": 0.5697,
      "step": 2833
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.339608331275493e-06,
      "loss": 0.7254,
      "step": 2834
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.333502622308937e-06,
      "loss": 0.521,
      "step": 2835
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.32739716290897e-06,
      "loss": 0.4736,
      "step": 2836
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.321291955361756e-06,
      "loss": 0.6969,
      "step": 2837
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.315187001953354e-06,
      "loss": 0.6021,
      "step": 2838
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.309082304969738e-06,
      "loss": 0.4301,
      "step": 2839
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.302977866696779e-06,
      "loss": 0.498,
      "step": 2840
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.29687368942025e-06,
      "loss": 0.5398,
      "step": 2841
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.290769775425838e-06,
      "loss": 0.3539,
      "step": 2842
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.284666126999116e-06,
      "loss": 0.6601,
      "step": 2843
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.278562746425561e-06,
      "loss": 0.434,
      "step": 2844
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.272459635990563e-06,
      "loss": 0.4883,
      "step": 2845
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.266356797979397e-06,
      "loss": 0.4688,
      "step": 2846
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.260254234677235e-06,
      "loss": 0.4445,
      "step": 2847
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.254151948369158e-06,
      "loss": 0.7588,
      "step": 2848
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.248049941340133e-06,
      "loss": 0.4728,
      "step": 2849
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.241948215875024e-06,
      "loss": 0.5616,
      "step": 2850
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.235846774258599e-06,
      "loss": 0.4253,
      "step": 2851
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.229745618775505e-06,
      "loss": 0.4773,
      "step": 2852
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.223644751710288e-06,
      "loss": 0.489,
      "step": 2853
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.217544175347394e-06,
      "loss": 0.685,
      "step": 2854
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.211443891971151e-06,
      "loss": 0.5987,
      "step": 2855
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.205343903865774e-06,
      "loss": 0.4836,
      "step": 2856
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.199244213315377e-06,
      "loss": 0.5499,
      "step": 2857
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.19314482260396e-06,
      "loss": 0.6017,
      "step": 2858
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.187045734015401e-06,
      "loss": 0.5178,
      "step": 2859
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.180946949833487e-06,
      "loss": 0.5892,
      "step": 2860
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.174848472341867e-06,
      "loss": 0.6489,
      "step": 2861
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.168750303824085e-06,
      "loss": 0.4936,
      "step": 2862
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.162652446563576e-06,
      "loss": 0.5396,
      "step": 2863
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.156554902843645e-06,
      "loss": 0.5427,
      "step": 2864
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.15045767494749e-06,
      "loss": 0.4673,
      "step": 2865
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.144360765158188e-06,
      "loss": 0.6484,
      "step": 2866
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.138264175758693e-06,
      "loss": 0.5655,
      "step": 2867
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.132167909031848e-06,
      "loss": 0.4558,
      "step": 2868
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.126071967260365e-06,
      "loss": 0.4492,
      "step": 2869
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.119976352726837e-06,
      "loss": 0.6602,
      "step": 2870
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.11388106771374e-06,
      "loss": 0.6167,
      "step": 2871
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.107786114503426e-06,
      "loss": 0.6037,
      "step": 2872
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.10169149537811e-06,
      "loss": 0.3803,
      "step": 2873
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.0955972126199e-06,
      "loss": 0.4871,
      "step": 2874
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.089503268510764e-06,
      "loss": 0.4424,
      "step": 2875
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.08340966533255e-06,
      "loss": 0.5097,
      "step": 2876
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.07731640536698e-06,
      "loss": 0.5263,
      "step": 2877
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.071223490895644e-06,
      "loss": 0.4157,
      "step": 2878
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.065130924199998e-06,
      "loss": 0.4861,
      "step": 2879
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.059038707561381e-06,
      "loss": 0.5462,
      "step": 2880
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.05294684326099e-06,
      "loss": 0.4644,
      "step": 2881
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.046855333579884e-06,
      "loss": 0.4375,
      "step": 2882
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.040764180799011e-06,
      "loss": 0.5648,
      "step": 2883
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.03467338719917e-06,
      "loss": 0.6093,
      "step": 2884
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.028582955061016e-06,
      "loss": 0.5243,
      "step": 2885
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.022492886665099e-06,
      "loss": 0.6067,
      "step": 2886
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.016403184291805e-06,
      "loss": 0.4974,
      "step": 2887
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.010313850221387e-06,
      "loss": 0.5623,
      "step": 2888
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.00422488673398e-06,
      "loss": 0.4278,
      "step": 2889
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.998136296109552e-06,
      "loss": 0.4943,
      "step": 2890
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.992048080627947e-06,
      "loss": 0.4367,
      "step": 2891
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.985960242568875e-06,
      "loss": 0.6131,
      "step": 2892
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.979872784211893e-06,
      "loss": 0.4937,
      "step": 2893
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.973785707836414e-06,
      "loss": 0.5192,
      "step": 2894
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.967699015721717e-06,
      "loss": 0.4223,
      "step": 2895
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.961612710146934e-06,
      "loss": 0.4285,
      "step": 2896
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.955526793391049e-06,
      "loss": 0.5222,
      "step": 2897
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.949441267732908e-06,
      "loss": 0.4752,
      "step": 2898
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.943356135451198e-06,
      "loss": 0.5753,
      "step": 2899
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.937271398824471e-06,
      "loss": 0.5015,
      "step": 2900
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.931187060131126e-06,
      "loss": 0.5093,
      "step": 2901
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.925103121649412e-06,
      "loss": 0.5462,
      "step": 2902
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.919019585657432e-06,
      "loss": 0.4924,
      "step": 2903
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.912936454433131e-06,
      "loss": 0.547,
      "step": 2904
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.906853730254306e-06,
      "loss": 0.5586,
      "step": 2905
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.900771415398608e-06,
      "loss": 0.5253,
      "step": 2906
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.894689512143528e-06,
      "loss": 0.5854,
      "step": 2907
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.8886080227664e-06,
      "loss": 0.3988,
      "step": 2908
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.88252694954441e-06,
      "loss": 0.6246,
      "step": 2909
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.876446294754587e-06,
      "loss": 0.4862,
      "step": 2910
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.870366060673793e-06,
      "loss": 0.6121,
      "step": 2911
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.864286249578755e-06,
      "loss": 0.5953,
      "step": 2912
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.858206863746018e-06,
      "loss": 0.4309,
      "step": 2913
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.852127905451978e-06,
      "loss": 0.6244,
      "step": 2914
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.846049376972876e-06,
      "loss": 0.5172,
      "step": 2915
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.83997128058478e-06,
      "loss": 0.6256,
      "step": 2916
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.833893618563604e-06,
      "loss": 0.3789,
      "step": 2917
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.8278163931851e-06,
      "loss": 0.5416,
      "step": 2918
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.821739606724857e-06,
      "loss": 0.6849,
      "step": 2919
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.81566326145829e-06,
      "loss": 0.6048,
      "step": 2920
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.809587359660661e-06,
      "loss": 0.5398,
      "step": 2921
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.803511903607059e-06,
      "loss": 0.5547,
      "step": 2922
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.797436895572407e-06,
      "loss": 0.3728,
      "step": 2923
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.791362337831469e-06,
      "loss": 0.5033,
      "step": 2924
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.785288232658821e-06,
      "loss": 0.6686,
      "step": 2925
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.779214582328887e-06,
      "loss": 0.4512,
      "step": 2926
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.773141389115914e-06,
      "loss": 0.6226,
      "step": 2927
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.767068655293984e-06,
      "loss": 0.4401,
      "step": 2928
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.76099638313699e-06,
      "loss": 0.7129,
      "step": 2929
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.754924574918675e-06,
      "loss": 0.5941,
      "step": 2930
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.748853232912588e-06,
      "loss": 0.5283,
      "step": 2931
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.742782359392122e-06,
      "loss": 0.6601,
      "step": 2932
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.73671195663048e-06,
      "loss": 0.5349,
      "step": 2933
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.73064202690069e-06,
      "loss": 0.4126,
      "step": 2934
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.724572572475613e-06,
      "loss": 0.6247,
      "step": 2935
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.718503595627923e-06,
      "loss": 0.5575,
      "step": 2936
    },
    {
      "epoch": 0.55,
      "learning_rate": 8.712435098630116e-06,
      "loss": 0.4952,
      "step": 2937
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.706367083754516e-06,
      "loss": 0.4761,
      "step": 2938
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.700299553273255e-06,
      "loss": 0.5565,
      "step": 2939
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.69423250945829e-06,
      "loss": 0.4512,
      "step": 2940
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.688165954581398e-06,
      "loss": 0.4271,
      "step": 2941
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.68209989091417e-06,
      "loss": 0.4833,
      "step": 2942
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.676034320728007e-06,
      "loss": 0.3956,
      "step": 2943
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.669969246294137e-06,
      "loss": 0.4187,
      "step": 2944
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.663904669883599e-06,
      "loss": 0.4233,
      "step": 2945
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.657840593767231e-06,
      "loss": 0.5181,
      "step": 2946
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.651777020215713e-06,
      "loss": 0.5666,
      "step": 2947
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.645713951499508e-06,
      "loss": 0.6894,
      "step": 2948
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.6396513898889e-06,
      "loss": 0.522,
      "step": 2949
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.633589337653995e-06,
      "loss": 0.5926,
      "step": 2950
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.62752779706469e-06,
      "loss": 0.4989,
      "step": 2951
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.621466770390698e-06,
      "loss": 0.5764,
      "step": 2952
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.615406259901543e-06,
      "loss": 0.4098,
      "step": 2953
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.609346267866556e-06,
      "loss": 0.3709,
      "step": 2954
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.603286796554864e-06,
      "loss": 0.4361,
      "step": 2955
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.597227848235407e-06,
      "loss": 0.4497,
      "step": 2956
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.591169425176931e-06,
      "loss": 0.4724,
      "step": 2957
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.585111529647976e-06,
      "loss": 0.5088,
      "step": 2958
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.579054163916901e-06,
      "loss": 0.5065,
      "step": 2959
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.572997330251845e-06,
      "loss": 0.6236,
      "step": 2960
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.566941030920763e-06,
      "loss": 0.6003,
      "step": 2961
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.560885268191409e-06,
      "loss": 0.6097,
      "step": 2962
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.554830044331326e-06,
      "loss": 0.5347,
      "step": 2963
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.548775361607872e-06,
      "loss": 0.4447,
      "step": 2964
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.542721222288186e-06,
      "loss": 0.5525,
      "step": 2965
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.536667628639207e-06,
      "loss": 0.5475,
      "step": 2966
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.53061458292768e-06,
      "loss": 0.4177,
      "step": 2967
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.524562087420137e-06,
      "loss": 0.5183,
      "step": 2968
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.518510144382896e-06,
      "loss": 0.4696,
      "step": 2969
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.512458756082084e-06,
      "loss": 0.5318,
      "step": 2970
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.506407924783614e-06,
      "loss": 0.5538,
      "step": 2971
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.50035765275318e-06,
      "loss": 0.5127,
      "step": 2972
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.49430794225629e-06,
      "loss": 0.5981,
      "step": 2973
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.488258795558216e-06,
      "loss": 0.3834,
      "step": 2974
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.482210214924033e-06,
      "loss": 0.5269,
      "step": 2975
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.47616220261861e-06,
      "loss": 0.5368,
      "step": 2976
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.470114760906583e-06,
      "loss": 0.4843,
      "step": 2977
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.464067892052392e-06,
      "loss": 0.5317,
      "step": 2978
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.458021598320255e-06,
      "loss": 0.4098,
      "step": 2979
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.451975881974183e-06,
      "loss": 0.4445,
      "step": 2980
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.445930745277953e-06,
      "loss": 0.4867,
      "step": 2981
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.439886190495142e-06,
      "loss": 0.4445,
      "step": 2982
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.433842219889103e-06,
      "loss": 0.5093,
      "step": 2983
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.427798835722969e-06,
      "loss": 0.5275,
      "step": 2984
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.421756040259658e-06,
      "loss": 0.5373,
      "step": 2985
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.415713835761861e-06,
      "loss": 0.5868,
      "step": 2986
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.409672224492051e-06,
      "loss": 0.5477,
      "step": 2987
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.403631208712481e-06,
      "loss": 0.5404,
      "step": 2988
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.397590790685182e-06,
      "loss": 0.5223,
      "step": 2989
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.391550972671948e-06,
      "loss": 0.5017,
      "step": 2990
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.385511756934367e-06,
      "loss": 0.4057,
      "step": 2991
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.379473145733793e-06,
      "loss": 0.4246,
      "step": 2992
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.373435141331348e-06,
      "loss": 0.5538,
      "step": 2993
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.367397745987938e-06,
      "loss": 0.466,
      "step": 2994
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.36136096196423e-06,
      "loss": 0.4652,
      "step": 2995
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.355324791520675e-06,
      "loss": 0.5004,
      "step": 2996
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.349289236917482e-06,
      "loss": 0.5398,
      "step": 2997
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.343254300414629e-06,
      "loss": 0.5208,
      "step": 2998
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.337219984271879e-06,
      "loss": 0.5345,
      "step": 2999
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.331186290748742e-06,
      "loss": 0.4417,
      "step": 3000
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.325153222104506e-06,
      "loss": 0.4823,
      "step": 3001
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.319120780598228e-06,
      "loss": 0.4919,
      "step": 3002
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.313088968488723e-06,
      "loss": 0.4386,
      "step": 3003
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.307057788034568e-06,
      "loss": 0.4287,
      "step": 3004
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.301027241494112e-06,
      "loss": 0.5095,
      "step": 3005
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.294997331125467e-06,
      "loss": 0.5032,
      "step": 3006
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.28896805918649e-06,
      "loss": 0.5239,
      "step": 3007
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.282939427934828e-06,
      "loss": 0.6591,
      "step": 3008
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.276911439627862e-06,
      "loss": 0.4267,
      "step": 3009
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.27088409652274e-06,
      "loss": 0.501,
      "step": 3010
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.26485740087638e-06,
      "loss": 0.4469,
      "step": 3011
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.25883135494544e-06,
      "loss": 0.4296,
      "step": 3012
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.252805960986338e-06,
      "loss": 0.5235,
      "step": 3013
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.246781221255265e-06,
      "loss": 0.5718,
      "step": 3014
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.240757138008149e-06,
      "loss": 0.6041,
      "step": 3015
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.234733713500673e-06,
      "loss": 0.4478,
      "step": 3016
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.228710949988283e-06,
      "loss": 0.4361,
      "step": 3017
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.222688849726172e-06,
      "loss": 0.5025,
      "step": 3018
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.216667414969285e-06,
      "loss": 0.5559,
      "step": 3019
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.210646647972319e-06,
      "loss": 0.4382,
      "step": 3020
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.204626550989717e-06,
      "loss": 0.6519,
      "step": 3021
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.198607126275674e-06,
      "loss": 0.4549,
      "step": 3022
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.192588376084137e-06,
      "loss": 0.4406,
      "step": 3023
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.186570302668792e-06,
      "loss": 0.5285,
      "step": 3024
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.180552908283083e-06,
      "loss": 0.3716,
      "step": 3025
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.174536195180189e-06,
      "loss": 0.4863,
      "step": 3026
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.168520165613035e-06,
      "loss": 0.5362,
      "step": 3027
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.162504821834296e-06,
      "loss": 0.4767,
      "step": 3028
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.156490166096391e-06,
      "loss": 0.5038,
      "step": 3029
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.15047620065147e-06,
      "loss": 0.5282,
      "step": 3030
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.144462927751439e-06,
      "loss": 0.5963,
      "step": 3031
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.138450349647936e-06,
      "loss": 0.6506,
      "step": 3032
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.132438468592332e-06,
      "loss": 0.5802,
      "step": 3033
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.12642728683576e-06,
      "loss": 0.4579,
      "step": 3034
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.12041680662907e-06,
      "loss": 0.4116,
      "step": 3035
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.114407030222851e-06,
      "loss": 0.8309,
      "step": 3036
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.108397959867445e-06,
      "loss": 0.4523,
      "step": 3037
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.102389597812909e-06,
      "loss": 0.4969,
      "step": 3038
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.096381946309043e-06,
      "loss": 0.5381,
      "step": 3039
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.090375007605388e-06,
      "loss": 0.4538,
      "step": 3040
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.084368783951215e-06,
      "loss": 0.5091,
      "step": 3041
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.078363277595512e-06,
      "loss": 0.5762,
      "step": 3042
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.072358490787018e-06,
      "loss": 0.4071,
      "step": 3043
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.066354425774196e-06,
      "loss": 0.4642,
      "step": 3044
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.060351084805236e-06,
      "loss": 0.4128,
      "step": 3045
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.054348470128063e-06,
      "loss": 0.4752,
      "step": 3046
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.04834658399032e-06,
      "loss": 0.4344,
      "step": 3047
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.042345428639382e-06,
      "loss": 0.5735,
      "step": 3048
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.036345006322358e-06,
      "loss": 0.4243,
      "step": 3049
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.030345319286076e-06,
      "loss": 0.5409,
      "step": 3050
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.024346369777081e-06,
      "loss": 0.4688,
      "step": 3051
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.018348160041656e-06,
      "loss": 0.6312,
      "step": 3052
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.0123506923258e-06,
      "loss": 0.5084,
      "step": 3053
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.006353968875226e-06,
      "loss": 0.6555,
      "step": 3054
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.000357991935393e-06,
      "loss": 0.5515,
      "step": 3055
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.994362763751448e-06,
      "loss": 0.495,
      "step": 3056
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.988368286568287e-06,
      "loss": 0.5238,
      "step": 3057
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.982374562630506e-06,
      "loss": 0.4535,
      "step": 3058
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.97638159418242e-06,
      "loss": 0.4129,
      "step": 3059
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.970389383468077e-06,
      "loss": 0.6314,
      "step": 3060
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.96439793273122e-06,
      "loss": 0.4525,
      "step": 3061
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.958407244215324e-06,
      "loss": 0.5849,
      "step": 3062
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.952417320163572e-06,
      "loss": 0.672,
      "step": 3063
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.946428162818857e-06,
      "loss": 0.6415,
      "step": 3064
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.940439774423788e-06,
      "loss": 0.3893,
      "step": 3065
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.934452157220693e-06,
      "loss": 0.5258,
      "step": 3066
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.928465313451603e-06,
      "loss": 0.7122,
      "step": 3067
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.922479245358257e-06,
      "loss": 0.4717,
      "step": 3068
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.916493955182114e-06,
      "loss": 0.5139,
      "step": 3069
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.910509445164333e-06,
      "loss": 0.5068,
      "step": 3070
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.904525717545781e-06,
      "loss": 0.4205,
      "step": 3071
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.898542774567043e-06,
      "loss": 0.5386,
      "step": 3072
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.892560618468395e-06,
      "loss": 0.3922,
      "step": 3073
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.886579251489822e-06,
      "loss": 0.6506,
      "step": 3074
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.880598675871028e-06,
      "loss": 0.5176,
      "step": 3075
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.874618893851403e-06,
      "loss": 0.5194,
      "step": 3076
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.868639907670042e-06,
      "loss": 0.5078,
      "step": 3077
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.862661719565753e-06,
      "loss": 0.4738,
      "step": 3078
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.856684331777039e-06,
      "loss": 0.5476,
      "step": 3079
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.850707746542093e-06,
      "loss": 0.4622,
      "step": 3080
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.844731966098832e-06,
      "loss": 0.4384,
      "step": 3081
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.838756992684847e-06,
      "loss": 0.4824,
      "step": 3082
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.832782828537437e-06,
      "loss": 0.5829,
      "step": 3083
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.826809475893604e-06,
      "loss": 0.376,
      "step": 3084
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.820836936990031e-06,
      "loss": 0.605,
      "step": 3085
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.814865214063111e-06,
      "loss": 0.4291,
      "step": 3086
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.808894309348925e-06,
      "loss": 0.4467,
      "step": 3087
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.802924225083246e-06,
      "loss": 0.4315,
      "step": 3088
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.796954963501545e-06,
      "loss": 0.4285,
      "step": 3089
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.790986526838985e-06,
      "loss": 0.5254,
      "step": 3090
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.785018917330409e-06,
      "loss": 0.6135,
      "step": 3091
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.77905213721036e-06,
      "loss": 0.4402,
      "step": 3092
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.773086188713078e-06,
      "loss": 0.5399,
      "step": 3093
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.767121074072465e-06,
      "loss": 0.7058,
      "step": 3094
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.761156795522147e-06,
      "loss": 0.5413,
      "step": 3095
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.755193355295409e-06,
      "loss": 0.4109,
      "step": 3096
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.749230755625228e-06,
      "loss": 0.6396,
      "step": 3097
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.743268998744278e-06,
      "loss": 0.4935,
      "step": 3098
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.737308086884902e-06,
      "loss": 0.3488,
      "step": 3099
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.731348022279135e-06,
      "loss": 0.5519,
      "step": 3100
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.725388807158693e-06,
      "loss": 0.5785,
      "step": 3101
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.71943044375498e-06,
      "loss": 0.4332,
      "step": 3102
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.713472934299066e-06,
      "loss": 0.5539,
      "step": 3103
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.707516281021716e-06,
      "loss": 0.5435,
      "step": 3104
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.701560486153367e-06,
      "loss": 0.617,
      "step": 3105
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.695605551924137e-06,
      "loss": 0.5418,
      "step": 3106
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.689651480563824e-06,
      "loss": 0.4609,
      "step": 3107
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.683698274301896e-06,
      "loss": 0.5396,
      "step": 3108
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.6777459353675e-06,
      "loss": 0.5072,
      "step": 3109
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.671794465989464e-06,
      "loss": 0.5391,
      "step": 3110
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.665843868396287e-06,
      "loss": 0.4339,
      "step": 3111
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.659894144816132e-06,
      "loss": 0.5433,
      "step": 3112
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.653945297476849e-06,
      "loss": 0.388,
      "step": 3113
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.647997328605956e-06,
      "loss": 0.4643,
      "step": 3114
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.642050240430632e-06,
      "loss": 0.3923,
      "step": 3115
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.636104035177745e-06,
      "loss": 0.556,
      "step": 3116
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.630158715073813e-06,
      "loss": 0.4252,
      "step": 3117
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.624214282345033e-06,
      "loss": 0.384,
      "step": 3118
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.6182707392172735e-06,
      "loss": 0.5383,
      "step": 3119
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.612328087916053e-06,
      "loss": 0.5599,
      "step": 3120
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.606386330666582e-06,
      "loss": 0.5302,
      "step": 3121
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.600445469693711e-06,
      "loss": 0.6102,
      "step": 3122
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.594505507221967e-06,
      "loss": 0.4188,
      "step": 3123
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.588566445475544e-06,
      "loss": 0.5367,
      "step": 3124
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.582628286678291e-06,
      "loss": 0.4214,
      "step": 3125
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.576691033053717e-06,
      "loss": 0.4915,
      "step": 3126
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.570754686825004e-06,
      "loss": 0.5177,
      "step": 3127
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.5648192502149856e-06,
      "loss": 0.5544,
      "step": 3128
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.558884725446149e-06,
      "loss": 0.4857,
      "step": 3129
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.5529511147406565e-06,
      "loss": 0.4185,
      "step": 3130
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.547018420320314e-06,
      "loss": 0.6089,
      "step": 3131
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.541086644406589e-06,
      "loss": 0.5309,
      "step": 3132
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.5351557892206105e-06,
      "loss": 0.42,
      "step": 3133
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.529225856983151e-06,
      "loss": 0.4956,
      "step": 3134
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.523296849914643e-06,
      "loss": 0.5773,
      "step": 3135
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.51736877023518e-06,
      "loss": 0.5708,
      "step": 3136
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.511441620164499e-06,
      "loss": 0.4824,
      "step": 3137
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.505515401921986e-06,
      "loss": 0.4543,
      "step": 3138
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.4995901177266905e-06,
      "loss": 0.421,
      "step": 3139
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.493665769797306e-06,
      "loss": 0.4855,
      "step": 3140
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.487742360352164e-06,
      "loss": 0.574,
      "step": 3141
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.481819891609271e-06,
      "loss": 0.4424,
      "step": 3142
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.475898365786256e-06,
      "loss": 0.5256,
      "step": 3143
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.469977785100403e-06,
      "loss": 0.5839,
      "step": 3144
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.464058151768652e-06,
      "loss": 0.4622,
      "step": 3145
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.458139468007573e-06,
      "loss": 0.525,
      "step": 3146
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.452221736033387e-06,
      "loss": 0.4739,
      "step": 3147
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.446304958061963e-06,
      "loss": 0.6758,
      "step": 3148
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.440389136308806e-06,
      "loss": 0.417,
      "step": 3149
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.434474272989071e-06,
      "loss": 0.4672,
      "step": 3150
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.428560370317542e-06,
      "loss": 0.6176,
      "step": 3151
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.422647430508652e-06,
      "loss": 0.3036,
      "step": 3152
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.4167354557764735e-06,
      "loss": 0.5978,
      "step": 3153
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.410824448334717e-06,
      "loss": 0.5033,
      "step": 3154
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.404914410396722e-06,
      "loss": 0.3857,
      "step": 3155
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.39900534417548e-06,
      "loss": 0.5037,
      "step": 3156
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.393097251883609e-06,
      "loss": 0.5485,
      "step": 3157
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.387190135733362e-06,
      "loss": 0.468,
      "step": 3158
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.3812839979366366e-06,
      "loss": 0.6598,
      "step": 3159
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.3753788407049475e-06,
      "loss": 0.6112,
      "step": 3160
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.3694746662494535e-06,
      "loss": 0.5631,
      "step": 3161
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.363571476780945e-06,
      "loss": 0.483,
      "step": 3162
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.3576692745098435e-06,
      "loss": 0.4778,
      "step": 3163
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.351768061646192e-06,
      "loss": 0.4241,
      "step": 3164
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.345867840399676e-06,
      "loss": 0.3879,
      "step": 3165
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.339968612979605e-06,
      "loss": 0.4364,
      "step": 3166
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.334070381594904e-06,
      "loss": 0.5433,
      "step": 3167
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.328173148454151e-06,
      "loss": 0.4448,
      "step": 3168
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.322276915765526e-06,
      "loss": 0.4766,
      "step": 3169
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.3163816857368444e-06,
      "loss": 0.4472,
      "step": 3170
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.310487460575548e-06,
      "loss": 0.5455,
      "step": 3171
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.304594242488702e-06,
      "loss": 0.3437,
      "step": 3172
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.298702033682985e-06,
      "loss": 0.44,
      "step": 3173
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.29281083636471e-06,
      "loss": 0.47,
      "step": 3174
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.286920652739808e-06,
      "loss": 0.6174,
      "step": 3175
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.281031485013819e-06,
      "loss": 0.6003,
      "step": 3176
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.275143335391927e-06,
      "loss": 0.434,
      "step": 3177
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.2692562060789085e-06,
      "loss": 0.6152,
      "step": 3178
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.263370099279173e-06,
      "loss": 0.5534,
      "step": 3179
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.257485017196746e-06,
      "loss": 0.6418,
      "step": 3180
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.251600962035258e-06,
      "loss": 0.5835,
      "step": 3181
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.245717935997978e-06,
      "loss": 0.4626,
      "step": 3182
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.239835941287765e-06,
      "loss": 0.5387,
      "step": 3183
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.233954980107103e-06,
      "loss": 0.6547,
      "step": 3184
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.228075054658096e-06,
      "loss": 0.4901,
      "step": 3185
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.222196167142445e-06,
      "loss": 0.6247,
      "step": 3186
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.21631831976147e-06,
      "loss": 0.5879,
      "step": 3187
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.210441514716105e-06,
      "loss": 0.4696,
      "step": 3188
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.20456575420689e-06,
      "loss": 0.4383,
      "step": 3189
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.19869104043397e-06,
      "loss": 0.575,
      "step": 3190
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.192817375597106e-06,
      "loss": 0.4951,
      "step": 3191
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.18694476189566e-06,
      "loss": 0.3872,
      "step": 3192
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.181073201528602e-06,
      "loss": 0.618,
      "step": 3193
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.1752026966945145e-06,
      "loss": 0.4462,
      "step": 3194
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.16933324959157e-06,
      "loss": 0.5992,
      "step": 3195
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.1634648624175575e-06,
      "loss": 0.4699,
      "step": 3196
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.157597537369866e-06,
      "loss": 0.426,
      "step": 3197
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.151731276645486e-06,
      "loss": 0.4749,
      "step": 3198
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.145866082441004e-06,
      "loss": 0.4018,
      "step": 3199
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.14000195695262e-06,
      "loss": 0.5467,
      "step": 3200
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.134138902376124e-06,
      "loss": 0.7808,
      "step": 3201
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.1282769209069005e-06,
      "loss": 0.5285,
      "step": 3202
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.122416014739949e-06,
      "loss": 0.4721,
      "step": 3203
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.116556186069852e-06,
      "loss": 0.4818,
      "step": 3204
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.110697437090789e-06,
      "loss": 0.6775,
      "step": 3205
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.104839769996548e-06,
      "loss": 0.8018,
      "step": 3206
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.098983186980495e-06,
      "loss": 0.4717,
      "step": 3207
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.0931276902356e-06,
      "loss": 0.5495,
      "step": 3208
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.087273281954426e-06,
      "loss": 0.4359,
      "step": 3209
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.0814199643291296e-06,
      "loss": 0.4738,
      "step": 3210
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.075567739551448e-06,
      "loss": 0.5617,
      "step": 3211
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.069716609812722e-06,
      "loss": 0.4124,
      "step": 3212
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.063866577303879e-06,
      "loss": 0.4728,
      "step": 3213
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.058017644215433e-06,
      "loss": 0.5131,
      "step": 3214
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.05216981273749e-06,
      "loss": 0.5255,
      "step": 3215
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.046323085059734e-06,
      "loss": 0.3221,
      "step": 3216
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.040477463371449e-06,
      "loss": 0.5473,
      "step": 3217
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.034632949861497e-06,
      "loss": 0.4032,
      "step": 3218
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.028789546718327e-06,
      "loss": 0.938,
      "step": 3219
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.022947256129973e-06,
      "loss": 0.4234,
      "step": 3220
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.017106080284048e-06,
      "loss": 0.5973,
      "step": 3221
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.011266021367753e-06,
      "loss": 0.5066,
      "step": 3222
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.005427081567868e-06,
      "loss": 0.5323,
      "step": 3223
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.999589263070758e-06,
      "loss": 0.6164,
      "step": 3224
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.99375256806236e-06,
      "loss": 0.5667,
      "step": 3225
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.987916998728197e-06,
      "loss": 0.4044,
      "step": 3226
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.982082557253371e-06,
      "loss": 0.4616,
      "step": 3227
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.976249245822552e-06,
      "loss": 0.4577,
      "step": 3228
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.9704170666200045e-06,
      "loss": 0.6498,
      "step": 3229
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.964586021829551e-06,
      "loss": 0.5295,
      "step": 3230
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.958756113634597e-06,
      "loss": 0.5282,
      "step": 3231
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.95292734421813e-06,
      "loss": 0.3377,
      "step": 3232
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.947099715762694e-06,
      "loss": 0.4739,
      "step": 3233
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.941273230450418e-06,
      "loss": 0.4566,
      "step": 3234
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.935447890463003e-06,
      "loss": 0.4811,
      "step": 3235
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.9296236979817175e-06,
      "loss": 0.5803,
      "step": 3236
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.9238006551873985e-06,
      "loss": 0.3844,
      "step": 3237
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.917978764260457e-06,
      "loss": 0.4073,
      "step": 3238
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.912158027380871e-06,
      "loss": 0.4587,
      "step": 3239
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.9063384467281825e-06,
      "loss": 0.4974,
      "step": 3240
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.900520024481512e-06,
      "loss": 0.4855,
      "step": 3241
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.894702762819531e-06,
      "loss": 0.5904,
      "step": 3242
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.888886663920487e-06,
      "loss": 0.3835,
      "step": 3243
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.883071729962188e-06,
      "loss": 0.6465,
      "step": 3244
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.877257963122006e-06,
      "loss": 0.6476,
      "step": 3245
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.871445365576883e-06,
      "loss": 0.5638,
      "step": 3246
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.86563393950331e-06,
      "loss": 0.5597,
      "step": 3247
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.859823687077347e-06,
      "loss": 0.5136,
      "step": 3248
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.854014610474616e-06,
      "loss": 0.4409,
      "step": 3249
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.848206711870298e-06,
      "loss": 0.4024,
      "step": 3250
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.842399993439125e-06,
      "loss": 0.5812,
      "step": 3251
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.8365944573554e-06,
      "loss": 0.4905,
      "step": 3252
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.8307901057929735e-06,
      "loss": 0.605,
      "step": 3253
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.824986940925256e-06,
      "loss": 0.5233,
      "step": 3254
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.8191849649252174e-06,
      "loss": 0.6124,
      "step": 3255
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.813384179965373e-06,
      "loss": 0.4578,
      "step": 3256
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.807584588217798e-06,
      "loss": 0.4537,
      "step": 3257
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.801786191854125e-06,
      "loss": 0.5043,
      "step": 3258
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.795988993045532e-06,
      "loss": 0.5448,
      "step": 3259
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.790192993962748e-06,
      "loss": 0.5025,
      "step": 3260
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.784398196776059e-06,
      "loss": 0.6605,
      "step": 3261
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.778604603655297e-06,
      "loss": 0.6083,
      "step": 3262
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.7728122167698395e-06,
      "loss": 0.4249,
      "step": 3263
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.767021038288628e-06,
      "loss": 0.5278,
      "step": 3264
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.761231070380129e-06,
      "loss": 0.324,
      "step": 3265
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.755442315212368e-06,
      "loss": 0.4769,
      "step": 3266
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.749654774952925e-06,
      "loss": 0.5361,
      "step": 3267
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.743868451768904e-06,
      "loss": 0.5796,
      "step": 3268
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.738083347826968e-06,
      "loss": 0.637,
      "step": 3269
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.732299465293322e-06,
      "loss": 0.6694,
      "step": 3270
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.726516806333714e-06,
      "loss": 0.4917,
      "step": 3271
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.720735373113424e-06,
      "loss": 0.433,
      "step": 3272
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.714955167797288e-06,
      "loss": 0.4871,
      "step": 3273
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.709176192549668e-06,
      "loss": 0.5476,
      "step": 3274
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.7033984495344785e-06,
      "loss": 0.4455,
      "step": 3275
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.697621940915167e-06,
      "loss": 0.5125,
      "step": 3276
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.691846668854709e-06,
      "loss": 0.4601,
      "step": 3277
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.686072635515636e-06,
      "loss": 0.4738,
      "step": 3278
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.680299843060001e-06,
      "loss": 0.5729,
      "step": 3279
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.674528293649396e-06,
      "loss": 0.6135,
      "step": 3280
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.6687579894449536e-06,
      "loss": 0.441,
      "step": 3281
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.662988932607331e-06,
      "loss": 0.3674,
      "step": 3282
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.657221125296722e-06,
      "loss": 0.3977,
      "step": 3283
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.651454569672856e-06,
      "loss": 0.413,
      "step": 3284
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.645689267894993e-06,
      "loss": 0.5119,
      "step": 3285
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.639925222121911e-06,
      "loss": 0.5819,
      "step": 3286
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.634162434511939e-06,
      "loss": 0.6636,
      "step": 3287
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.62840090722292e-06,
      "loss": 0.7617,
      "step": 3288
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.622640642412224e-06,
      "loss": 0.374,
      "step": 3289
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.616881642236763e-06,
      "loss": 0.4353,
      "step": 3290
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.611123908852959e-06,
      "loss": 0.5051,
      "step": 3291
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.6053674444167635e-06,
      "loss": 0.4328,
      "step": 3292
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.599612251083666e-06,
      "loss": 0.6986,
      "step": 3293
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.59385833100866e-06,
      "loss": 0.6015,
      "step": 3294
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.588105686346276e-06,
      "loss": 0.3935,
      "step": 3295
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.5823543192505615e-06,
      "loss": 0.5522,
      "step": 3296
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.57660423187509e-06,
      "loss": 0.5039,
      "step": 3297
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.5708554263729485e-06,
      "loss": 0.5747,
      "step": 3298
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.56510790489675e-06,
      "loss": 0.5201,
      "step": 3299
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.559361669598627e-06,
      "loss": 0.4286,
      "step": 3300
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.553616722630225e-06,
      "loss": 0.5222,
      "step": 3301
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.547873066142716e-06,
      "loss": 0.5216,
      "step": 3302
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.542130702286778e-06,
      "loss": 0.5534,
      "step": 3303
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.53638963321261e-06,
      "loss": 0.4584,
      "step": 3304
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.5306498610699295e-06,
      "loss": 0.3926,
      "step": 3305
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.524911388007963e-06,
      "loss": 0.4331,
      "step": 3306
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.519174216175458e-06,
      "loss": 0.6936,
      "step": 3307
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.513438347720665e-06,
      "loss": 0.4282,
      "step": 3308
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.507703784791346e-06,
      "loss": 0.5521,
      "step": 3309
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.50197052953479e-06,
      "loss": 0.4501,
      "step": 3310
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.496238584097782e-06,
      "loss": 0.4071,
      "step": 3311
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.490507950626616e-06,
      "loss": 0.4504,
      "step": 3312
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.484778631267101e-06,
      "loss": 0.6151,
      "step": 3313
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.479050628164557e-06,
      "loss": 0.5619,
      "step": 3314
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.473323943463795e-06,
      "loss": 0.4219,
      "step": 3315
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.467598579309156e-06,
      "loss": 0.6085,
      "step": 3316
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.461874537844465e-06,
      "loss": 0.3434,
      "step": 3317
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.4561518212130616e-06,
      "loss": 0.5116,
      "step": 3318
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.4504304315577924e-06,
      "loss": 0.5222,
      "step": 3319
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.444710371021e-06,
      "loss": 0.4914,
      "step": 3320
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.438991641744531e-06,
      "loss": 0.6709,
      "step": 3321
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.433274245869738e-06,
      "loss": 0.5064,
      "step": 3322
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.4275581855374715e-06,
      "loss": 0.4392,
      "step": 3323
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.4218434628880755e-06,
      "loss": 0.8076,
      "step": 3324
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.416130080061406e-06,
      "loss": 0.526,
      "step": 3325
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.410418039196808e-06,
      "loss": 0.5932,
      "step": 3326
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.404707342433123e-06,
      "loss": 0.599,
      "step": 3327
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.398997991908702e-06,
      "loss": 0.4436,
      "step": 3328
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.393289989761373e-06,
      "loss": 0.5047,
      "step": 3329
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.387583338128471e-06,
      "loss": 0.4076,
      "step": 3330
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.3818780391468276e-06,
      "loss": 0.5505,
      "step": 3331
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.376174094952761e-06,
      "loss": 0.4719,
      "step": 3332
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.370471507682079e-06,
      "loss": 0.5007,
      "step": 3333
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.364770279470093e-06,
      "loss": 0.564,
      "step": 3334
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.359070412451598e-06,
      "loss": 0.5434,
      "step": 3335
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.353371908760878e-06,
      "loss": 0.5505,
      "step": 3336
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.347674770531716e-06,
      "loss": 0.503,
      "step": 3337
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.3419789998973655e-06,
      "loss": 0.5584,
      "step": 3338
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.33628459899059e-06,
      "loss": 0.5286,
      "step": 3339
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.330591569943623e-06,
      "loss": 0.4597,
      "step": 3340
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.324899914888192e-06,
      "loss": 0.519,
      "step": 3341
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.319209635955512e-06,
      "loss": 0.407,
      "step": 3342
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.313520735276274e-06,
      "loss": 0.5043,
      "step": 3343
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.307833214980659e-06,
      "loss": 0.3968,
      "step": 3344
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.302147077198334e-06,
      "loss": 0.4639,
      "step": 3345
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.296462324058445e-06,
      "loss": 0.5797,
      "step": 3346
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.2907789576896125e-06,
      "loss": 0.4748,
      "step": 3347
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.285096980219951e-06,
      "loss": 0.4875,
      "step": 3348
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.2794163937770494e-06,
      "loss": 0.5545,
      "step": 3349
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.273737200487964e-06,
      "loss": 0.5478,
      "step": 3350
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.268059402479254e-06,
      "loss": 0.5001,
      "step": 3351
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.262383001876934e-06,
      "loss": 0.6528,
      "step": 3352
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.256708000806504e-06,
      "loss": 0.5567,
      "step": 3353
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.251034401392946e-06,
      "loss": 0.5363,
      "step": 3354
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.245362205760703e-06,
      "loss": 0.4667,
      "step": 3355
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.239691416033702e-06,
      "loss": 0.5025,
      "step": 3356
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.2340220343353455e-06,
      "loss": 0.5575,
      "step": 3357
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.228354062788503e-06,
      "loss": 0.4591,
      "step": 3358
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.222687503515514e-06,
      "loss": 0.4459,
      "step": 3359
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.217022358638199e-06,
      "loss": 0.4457,
      "step": 3360
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.211358630277839e-06,
      "loss": 0.607,
      "step": 3361
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.20569632055519e-06,
      "loss": 0.5043,
      "step": 3362
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.200035431590479e-06,
      "loss": 0.4912,
      "step": 3363
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.194375965503392e-06,
      "loss": 0.5051,
      "step": 3364
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.188717924413088e-06,
      "loss": 0.4857,
      "step": 3365
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.1830613104381945e-06,
      "loss": 0.5091,
      "step": 3366
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.177406125696804e-06,
      "loss": 0.4981,
      "step": 3367
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.1717523723064645e-06,
      "loss": 0.5932,
      "step": 3368
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.166100052384203e-06,
      "loss": 0.5207,
      "step": 3369
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.160449168046496e-06,
      "loss": 0.4803,
      "step": 3370
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.1547997214092966e-06,
      "loss": 0.5571,
      "step": 3371
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.149151714588009e-06,
      "loss": 0.4193,
      "step": 3372
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.143505149697494e-06,
      "loss": 0.4508,
      "step": 3373
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.137860028852088e-06,
      "loss": 0.4994,
      "step": 3374
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.1322163541655765e-06,
      "loss": 0.4291,
      "step": 3375
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.126574127751197e-06,
      "loss": 0.5578,
      "step": 3376
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.120933351721665e-06,
      "loss": 0.455,
      "step": 3377
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.115294028189134e-06,
      "loss": 0.4094,
      "step": 3378
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.109656159265218e-06,
      "loss": 0.4123,
      "step": 3379
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.104019747060998e-06,
      "loss": 0.4723,
      "step": 3380
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.098384793686991e-06,
      "loss": 0.7365,
      "step": 3381
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.092751301253179e-06,
      "loss": 0.3808,
      "step": 3382
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.087119271868997e-06,
      "loss": 0.6537,
      "step": 3383
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.0814887076433325e-06,
      "loss": 0.5426,
      "step": 3384
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.075859610684514e-06,
      "loss": 0.5326,
      "step": 3385
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.070231983100334e-06,
      "loss": 0.5755,
      "step": 3386
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.064605826998031e-06,
      "loss": 0.4518,
      "step": 3387
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.058981144484284e-06,
      "loss": 0.4402,
      "step": 3388
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.053357937665237e-06,
      "loss": 0.4267,
      "step": 3389
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.047736208646464e-06,
      "loss": 0.5992,
      "step": 3390
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.042115959532992e-06,
      "loss": 0.6728,
      "step": 3391
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.036497192429302e-06,
      "loss": 0.4722,
      "step": 3392
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.030879909439311e-06,
      "loss": 0.5192,
      "step": 3393
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.025264112666376e-06,
      "loss": 0.5525,
      "step": 3394
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.01964980421331e-06,
      "loss": 0.503,
      "step": 3395
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.014036986182363e-06,
      "loss": 0.4919,
      "step": 3396
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.00842566067522e-06,
      "loss": 0.6135,
      "step": 3397
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.002815829793024e-06,
      "loss": 0.4958,
      "step": 3398
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.997207495636339e-06,
      "loss": 0.526,
      "step": 3399
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.99160066030518e-06,
      "loss": 0.5612,
      "step": 3400
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.985995325899003e-06,
      "loss": 0.3793,
      "step": 3401
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.980391494516686e-06,
      "loss": 0.6535,
      "step": 3402
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.974789168256569e-06,
      "loss": 0.5313,
      "step": 3403
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.969188349216408e-06,
      "loss": 0.5087,
      "step": 3404
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.963589039493401e-06,
      "loss": 0.5083,
      "step": 3405
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.957991241184184e-06,
      "loss": 0.5091,
      "step": 3406
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.952394956384823e-06,
      "loss": 0.582,
      "step": 3407
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.946800187190816e-06,
      "loss": 0.4688,
      "step": 3408
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9412069356971e-06,
      "loss": 0.5332,
      "step": 3409
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.935615203998042e-06,
      "loss": 0.5023,
      "step": 3410
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.930024994187428e-06,
      "loss": 0.5693,
      "step": 3411
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.924436308358497e-06,
      "loss": 0.4669,
      "step": 3412
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.9188491486038955e-06,
      "loss": 0.5396,
      "step": 3413
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.9132635170157085e-06,
      "loss": 0.4117,
      "step": 3414
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.907679415685452e-06,
      "loss": 0.5417,
      "step": 3415
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.902096846704059e-06,
      "loss": 0.5625,
      "step": 3416
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.896515812161896e-06,
      "loss": 0.3686,
      "step": 3417
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.890936314148756e-06,
      "loss": 0.4596,
      "step": 3418
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.885358354753856e-06,
      "loss": 0.5774,
      "step": 3419
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.879781936065825e-06,
      "loss": 0.5382,
      "step": 3420
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.874207060172734e-06,
      "loss": 0.5118,
      "step": 3421
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.868633729162064e-06,
      "loss": 0.5409,
      "step": 3422
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.863061945120719e-06,
      "loss": 0.5532,
      "step": 3423
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.857491710135032e-06,
      "loss": 0.5417,
      "step": 3424
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8519230262907425e-06,
      "loss": 0.5348,
      "step": 3425
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.846355895673017e-06,
      "loss": 0.442,
      "step": 3426
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.840790320366444e-06,
      "loss": 0.4738,
      "step": 3427
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.835226302455021e-06,
      "loss": 0.604,
      "step": 3428
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.829663844022171e-06,
      "loss": 0.481,
      "step": 3429
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.824102947150722e-06,
      "loss": 0.5533,
      "step": 3430
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.8185436139229305e-06,
      "loss": 0.537,
      "step": 3431
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.812985846420456e-06,
      "loss": 0.6688,
      "step": 3432
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.807429646724378e-06,
      "loss": 0.5341,
      "step": 3433
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.801875016915186e-06,
      "loss": 0.6971,
      "step": 3434
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.796321959072793e-06,
      "loss": 0.4766,
      "step": 3435
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7907704752765015e-06,
      "loss": 0.4798,
      "step": 3436
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7852205676050355e-06,
      "loss": 0.3997,
      "step": 3437
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.779672238136541e-06,
      "loss": 0.4918,
      "step": 3438
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.774125488948561e-06,
      "loss": 0.6842,
      "step": 3439
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.768580322118034e-06,
      "loss": 0.5902,
      "step": 3440
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.763036739721334e-06,
      "loss": 0.4537,
      "step": 3441
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.757494743834226e-06,
      "loss": 0.5297,
      "step": 3442
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.751954336531873e-06,
      "loss": 0.4733,
      "step": 3443
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.746415519888862e-06,
      "loss": 0.4153,
      "step": 3444
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.740878295979172e-06,
      "loss": 0.4148,
      "step": 3445
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.735342666876189e-06,
      "loss": 0.3709,
      "step": 3446
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.7298086346527e-06,
      "loss": 0.5815,
      "step": 3447
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.724276201380898e-06,
      "loss": 0.4302,
      "step": 3448
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.718745369132373e-06,
      "loss": 0.5943,
      "step": 3449
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.713216139978118e-06,
      "loss": 0.551,
      "step": 3450
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.707688515988525e-06,
      "loss": 0.4193,
      "step": 3451
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.702162499233381e-06,
      "loss": 0.5102,
      "step": 3452
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.696638091781888e-06,
      "loss": 0.4024,
      "step": 3453
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.691115295702619e-06,
      "loss": 0.5691,
      "step": 3454
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.685594113063556e-06,
      "loss": 0.58,
      "step": 3455
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.680074545932094e-06,
      "loss": 0.4491,
      "step": 3456
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.674556596374993e-06,
      "loss": 0.4142,
      "step": 3457
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.669040266458421e-06,
      "loss": 0.6153,
      "step": 3458
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.663525558247948e-06,
      "loss": 0.4642,
      "step": 3459
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.658012473808532e-06,
      "loss": 0.6904,
      "step": 3460
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.652501015204503e-06,
      "loss": 0.5197,
      "step": 3461
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.646991184499613e-06,
      "loss": 0.4289,
      "step": 3462
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.641482983756987e-06,
      "loss": 0.4673,
      "step": 3463
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.635976415039143e-06,
      "loss": 0.4503,
      "step": 3464
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.6304714804079866e-06,
      "loss": 0.5544,
      "step": 3465
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.624968181924815e-06,
      "loss": 0.3905,
      "step": 3466
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.619466521650309e-06,
      "loss": 0.5364,
      "step": 3467
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.613966501644537e-06,
      "loss": 0.4373,
      "step": 3468
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.608468123966953e-06,
      "loss": 0.4237,
      "step": 3469
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.602971390676405e-06,
      "loss": 0.393,
      "step": 3470
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.597476303831109e-06,
      "loss": 0.6294,
      "step": 3471
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.591982865488668e-06,
      "loss": 0.554,
      "step": 3472
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5864910777060845e-06,
      "loss": 0.5014,
      "step": 3473
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.581000942539729e-06,
      "loss": 0.5121,
      "step": 3474
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5755124620453424e-06,
      "loss": 0.5158,
      "step": 3475
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.570025638278073e-06,
      "loss": 0.4608,
      "step": 3476
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.564540473292433e-06,
      "loss": 0.4672,
      "step": 3477
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5590569691423015e-06,
      "loss": 0.452,
      "step": 3478
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.553575127880962e-06,
      "loss": 0.5112,
      "step": 3479
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5480949515610605e-06,
      "loss": 0.5649,
      "step": 3480
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.542616442234618e-06,
      "loss": 0.4602,
      "step": 3481
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5371396019530364e-06,
      "loss": 0.5398,
      "step": 3482
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.531664432767092e-06,
      "loss": 0.5054,
      "step": 3483
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.526190936726934e-06,
      "loss": 0.5122,
      "step": 3484
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.5207191158820835e-06,
      "loss": 0.3856,
      "step": 3485
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.515248972281438e-06,
      "loss": 0.5632,
      "step": 3486
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.509780507973266e-06,
      "loss": 0.6288,
      "step": 3487
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.504313725005204e-06,
      "loss": 0.5059,
      "step": 3488
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.498848625424264e-06,
      "loss": 0.554,
      "step": 3489
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4933852112768184e-06,
      "loss": 0.4661,
      "step": 3490
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.487923484608629e-06,
      "loss": 0.424,
      "step": 3491
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4824634474648005e-06,
      "loss": 0.5835,
      "step": 3492
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.477005101889814e-06,
      "loss": 0.4314,
      "step": 3493
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.471548449927531e-06,
      "loss": 0.3086,
      "step": 3494
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.466093493621158e-06,
      "loss": 0.6224,
      "step": 3495
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4606402350132826e-06,
      "loss": 0.5816,
      "step": 3496
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.455188676145846e-06,
      "loss": 0.3441,
      "step": 3497
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.449738819060156e-06,
      "loss": 0.5054,
      "step": 3498
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.444290665796888e-06,
      "loss": 0.4813,
      "step": 3499
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.438844218396072e-06,
      "loss": 0.5564,
      "step": 3500
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.433399478897106e-06,
      "loss": 0.6561,
      "step": 3501
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.4279564493387414e-06,
      "loss": 0.4198,
      "step": 3502
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.422515131759097e-06,
      "loss": 0.4705,
      "step": 3503
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.417075528195641e-06,
      "loss": 0.5703,
      "step": 3504
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.411637640685219e-06,
      "loss": 0.6025,
      "step": 3505
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.406201471264006e-06,
      "loss": 0.4661,
      "step": 3506
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.40076702196755e-06,
      "loss": 0.5251,
      "step": 3507
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.395334294830766e-06,
      "loss": 0.553,
      "step": 3508
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.389903291887898e-06,
      "loss": 0.5171,
      "step": 3509
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.3844740151725586e-06,
      "loss": 0.648,
      "step": 3510
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.379046466717721e-06,
      "loss": 0.4543,
      "step": 3511
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.373620648555702e-06,
      "loss": 0.4749,
      "step": 3512
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.368196562718162e-06,
      "loss": 0.4333,
      "step": 3513
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.362774211236134e-06,
      "loss": 0.5011,
      "step": 3514
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.3573535961399844e-06,
      "loss": 0.5206,
      "step": 3515
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.35193471945944e-06,
      "loss": 0.3573,
      "step": 3516
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.346517583223567e-06,
      "loss": 0.4214,
      "step": 3517
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.3411021894607865e-06,
      "loss": 0.5389,
      "step": 3518
    },
    {
      "epoch": 0.66,
      "learning_rate": 5.3356885401988655e-06,
      "loss": 0.4298,
      "step": 3519
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.330276637464918e-06,
      "loss": 0.4552,
      "step": 3520
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.3248664832854045e-06,
      "loss": 0.4032,
      "step": 3521
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.319458079686129e-06,
      "loss": 0.6496,
      "step": 3522
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.3140514286922394e-06,
      "loss": 0.4265,
      "step": 3523
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.308646532328233e-06,
      "loss": 0.5252,
      "step": 3524
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.3032433926179395e-06,
      "loss": 0.3827,
      "step": 3525
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.297842011584549e-06,
      "loss": 0.4257,
      "step": 3526
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.292442391250567e-06,
      "loss": 0.5167,
      "step": 3527
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.287044533637865e-06,
      "loss": 0.5508,
      "step": 3528
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.28164844076764e-06,
      "loss": 0.5471,
      "step": 3529
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2762541146604316e-06,
      "loss": 0.3377,
      "step": 3530
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.270861557336119e-06,
      "loss": 0.5653,
      "step": 3531
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.265470770813918e-06,
      "loss": 0.5825,
      "step": 3532
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.26008175711238e-06,
      "loss": 0.3938,
      "step": 3533
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.254694518249395e-06,
      "loss": 0.5072,
      "step": 3534
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.249309056242189e-06,
      "loss": 0.4652,
      "step": 3535
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.243925373107318e-06,
      "loss": 0.5568,
      "step": 3536
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.238543470860677e-06,
      "loss": 0.5639,
      "step": 3537
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.233163351517494e-06,
      "loss": 0.6471,
      "step": 3538
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.227785017092322e-06,
      "loss": 0.5682,
      "step": 3539
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.222408469599061e-06,
      "loss": 0.5105,
      "step": 3540
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.217033711050923e-06,
      "loss": 0.7013,
      "step": 3541
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.211660743460458e-06,
      "loss": 0.5538,
      "step": 3542
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.2062895688395595e-06,
      "loss": 0.4278,
      "step": 3543
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.200920189199425e-06,
      "loss": 0.4614,
      "step": 3544
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.195552606550593e-06,
      "loss": 0.5245,
      "step": 3545
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.190186822902932e-06,
      "loss": 0.5252,
      "step": 3546
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.184822840265635e-06,
      "loss": 0.3847,
      "step": 3547
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.179460660647206e-06,
      "loss": 0.5081,
      "step": 3548
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.174100286055499e-06,
      "loss": 0.4585,
      "step": 3549
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.168741718497674e-06,
      "loss": 0.4081,
      "step": 3550
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.163384959980221e-06,
      "loss": 0.4893,
      "step": 3551
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.15803001250895e-06,
      "loss": 0.4001,
      "step": 3552
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.152676878088996e-06,
      "loss": 0.5426,
      "step": 3553
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.147325558724811e-06,
      "loss": 0.6149,
      "step": 3554
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.14197605642017e-06,
      "loss": 0.6608,
      "step": 3555
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.136628373178165e-06,
      "loss": 0.5387,
      "step": 3556
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.131282511001221e-06,
      "loss": 0.4085,
      "step": 3557
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.125938471891054e-06,
      "loss": 0.4146,
      "step": 3558
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.120596257848716e-06,
      "loss": 0.4865,
      "step": 3559
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.1152558708745795e-06,
      "loss": 0.4,
      "step": 3560
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.109917312968325e-06,
      "loss": 0.4896,
      "step": 3561
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.104580586128935e-06,
      "loss": 0.5853,
      "step": 3562
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.099245692354735e-06,
      "loss": 0.4437,
      "step": 3563
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.093912633643348e-06,
      "loss": 0.4146,
      "step": 3564
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.088581411991699e-06,
      "loss": 0.6475,
      "step": 3565
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.0832520293960485e-06,
      "loss": 0.4399,
      "step": 3566
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.077924487851954e-06,
      "loss": 0.7193,
      "step": 3567
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.072598789354286e-06,
      "loss": 0.4709,
      "step": 3568
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.067274935897226e-06,
      "loss": 0.4856,
      "step": 3569
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.061952929474263e-06,
      "loss": 0.5192,
      "step": 3570
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.056632772078198e-06,
      "loss": 0.5042,
      "step": 3571
    },
    {
      "epoch": 0.67,
      "learning_rate": 5.051314465701134e-06,
      "loss": 0.6777,
      "step": 3572
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.045998012334486e-06,
      "loss": 0.5329,
      "step": 3573
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.040683413968974e-06,
      "loss": 0.5675,
      "step": 3574
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.035370672594622e-06,
      "loss": 0.3784,
      "step": 3575
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.0300597902007565e-06,
      "loss": 0.4492,
      "step": 3576
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.024750768776011e-06,
      "loss": 0.4128,
      "step": 3577
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.019443610308331e-06,
      "loss": 0.5922,
      "step": 3578
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.014138316784944e-06,
      "loss": 0.4892,
      "step": 3579
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.008834890192394e-06,
      "loss": 0.4928,
      "step": 3580
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.003533332516524e-06,
      "loss": 0.612,
      "step": 3581
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.998233645742478e-06,
      "loss": 0.5213,
      "step": 3582
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.992935831854687e-06,
      "loss": 0.5113,
      "step": 3583
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.987639892836903e-06,
      "loss": 0.4617,
      "step": 3584
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.982345830672161e-06,
      "loss": 0.5016,
      "step": 3585
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.977053647342785e-06,
      "loss": 0.4038,
      "step": 3586
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.971763344830419e-06,
      "loss": 0.4752,
      "step": 3587
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.966474925115987e-06,
      "loss": 0.3675,
      "step": 3588
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.961188390179709e-06,
      "loss": 0.3985,
      "step": 3589
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.9559037420011025e-06,
      "loss": 0.517,
      "step": 3590
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.950620982558975e-06,
      "loss": 0.5105,
      "step": 3591
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.945340113831437e-06,
      "loss": 0.5471,
      "step": 3592
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.940061137795876e-06,
      "loss": 0.5166,
      "step": 3593
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.934784056428975e-06,
      "loss": 0.4103,
      "step": 3594
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.929508871706724e-06,
      "loss": 0.4926,
      "step": 3595
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.924235585604377e-06,
      "loss": 0.5678,
      "step": 3596
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.91896420009649e-06,
      "loss": 0.346,
      "step": 3597
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.913694717156912e-06,
      "loss": 0.4559,
      "step": 3598
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.908427138758779e-06,
      "loss": 0.5344,
      "step": 3599
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.9031614668744934e-06,
      "loss": 0.4623,
      "step": 3600
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.897897703475773e-06,
      "loss": 0.4759,
      "step": 3601
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.892635850533604e-06,
      "loss": 0.4374,
      "step": 3602
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.88737591001826e-06,
      "loss": 0.3644,
      "step": 3603
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.8821178838993e-06,
      "loss": 0.4018,
      "step": 3604
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.876861774145565e-06,
      "loss": 0.4609,
      "step": 3605
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.871607582725179e-06,
      "loss": 0.5247,
      "step": 3606
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.866355311605547e-06,
      "loss": 0.6381,
      "step": 3607
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.861104962753357e-06,
      "loss": 0.5911,
      "step": 3608
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.855856538134573e-06,
      "loss": 0.5014,
      "step": 3609
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.850610039714444e-06,
      "loss": 0.4734,
      "step": 3610
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.845365469457494e-06,
      "loss": 0.5451,
      "step": 3611
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.8401228293275215e-06,
      "loss": 0.354,
      "step": 3612
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.834882121287618e-06,
      "loss": 0.4395,
      "step": 3613
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.82964334730013e-06,
      "loss": 0.4237,
      "step": 3614
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.824406509326691e-06,
      "loss": 0.4269,
      "step": 3615
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.819171609328214e-06,
      "loss": 0.4605,
      "step": 3616
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.813938649264881e-06,
      "loss": 0.5304,
      "step": 3617
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.808707631096138e-06,
      "loss": 0.5329,
      "step": 3618
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.8034785567807225e-06,
      "loss": 0.5057,
      "step": 3619
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.798251428276635e-06,
      "loss": 0.6011,
      "step": 3620
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.793026247541144e-06,
      "loss": 0.495,
      "step": 3621
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.787803016530792e-06,
      "loss": 0.4956,
      "step": 3622
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.782581737201394e-06,
      "loss": 0.5596,
      "step": 3623
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.777362411508031e-06,
      "loss": 0.4795,
      "step": 3624
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.772145041405053e-06,
      "loss": 0.4693,
      "step": 3625
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.766929628846073e-06,
      "loss": 0.542,
      "step": 3626
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.7617161757839895e-06,
      "loss": 0.5354,
      "step": 3627
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.756504684170942e-06,
      "loss": 0.5771,
      "step": 3628
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.751295155958345e-06,
      "loss": 0.4813,
      "step": 3629
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.746087593096894e-06,
      "loss": 0.598,
      "step": 3630
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.740881997536524e-06,
      "loss": 0.5691,
      "step": 3631
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.7356783712264405e-06,
      "loss": 0.4288,
      "step": 3632
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.730476716115127e-06,
      "loss": 0.4635,
      "step": 3633
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.725277034150314e-06,
      "loss": 0.407,
      "step": 3634
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.720079327278987e-06,
      "loss": 0.3943,
      "step": 3635
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.7148835974474105e-06,
      "loss": 0.5844,
      "step": 3636
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.7096898466010976e-06,
      "loss": 0.3975,
      "step": 3637
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.70449807668482e-06,
      "loss": 0.4844,
      "step": 3638
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6993082896426125e-06,
      "loss": 0.5085,
      "step": 3639
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6941204874177625e-06,
      "loss": 0.3902,
      "step": 3640
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.688934671952818e-06,
      "loss": 0.4441,
      "step": 3641
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.683750845189581e-06,
      "loss": 0.5188,
      "step": 3642
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.678569009069109e-06,
      "loss": 0.4289,
      "step": 3643
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.673389165531714e-06,
      "loss": 0.634,
      "step": 3644
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.668211316516964e-06,
      "loss": 0.6181,
      "step": 3645
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6630354639636775e-06,
      "loss": 0.4493,
      "step": 3646
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.657861609809923e-06,
      "loss": 0.3741,
      "step": 3647
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.652689755993035e-06,
      "loss": 0.4472,
      "step": 3648
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6475199044495765e-06,
      "loss": 0.5604,
      "step": 3649
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.642352057115374e-06,
      "loss": 0.6608,
      "step": 3650
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.637186215925512e-06,
      "loss": 0.4695,
      "step": 3651
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.632022382814299e-06,
      "loss": 0.4525,
      "step": 3652
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.626860559715318e-06,
      "loss": 0.4612,
      "step": 3653
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.621700748561384e-06,
      "loss": 0.5393,
      "step": 3654
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.616542951284561e-06,
      "loss": 0.6791,
      "step": 3655
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.611387169816162e-06,
      "loss": 0.373,
      "step": 3656
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6062334060867416e-06,
      "loss": 0.4408,
      "step": 3657
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.601081662026103e-06,
      "loss": 0.5197,
      "step": 3658
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.595931939563288e-06,
      "loss": 0.5973,
      "step": 3659
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5907842406265864e-06,
      "loss": 0.4853,
      "step": 3660
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5856385671435285e-06,
      "loss": 0.532,
      "step": 3661
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5804949210408856e-06,
      "loss": 0.5437,
      "step": 3662
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.575353304244668e-06,
      "loss": 0.6754,
      "step": 3663
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.570213718680127e-06,
      "loss": 0.4302,
      "step": 3664
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5650761662717645e-06,
      "loss": 0.3984,
      "step": 3665
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.559940648943301e-06,
      "loss": 0.4557,
      "step": 3666
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.554807168617703e-06,
      "loss": 0.4976,
      "step": 3667
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.549675727217186e-06,
      "loss": 0.4543,
      "step": 3668
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.544546326663193e-06,
      "loss": 0.5146,
      "step": 3669
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.539418968876389e-06,
      "loss": 0.4129,
      "step": 3670
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.534293655776699e-06,
      "loss": 0.4823,
      "step": 3671
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5291703892832705e-06,
      "loss": 0.551,
      "step": 3672
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.524049171314475e-06,
      "loss": 0.3644,
      "step": 3673
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.518930003787938e-06,
      "loss": 0.5485,
      "step": 3674
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.513812888620502e-06,
      "loss": 0.4162,
      "step": 3675
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.508697827728242e-06,
      "loss": 0.5323,
      "step": 3676
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5035848230264715e-06,
      "loss": 0.3471,
      "step": 3677
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.498473876429727e-06,
      "loss": 0.4536,
      "step": 3678
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.493364989851776e-06,
      "loss": 0.6671,
      "step": 3679
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.488258165205618e-06,
      "loss": 0.5601,
      "step": 3680
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.483153404403472e-06,
      "loss": 0.3448,
      "step": 3681
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.478050709356802e-06,
      "loss": 0.5614,
      "step": 3682
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.472950081976275e-06,
      "loss": 0.4477,
      "step": 3683
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.467851524171796e-06,
      "loss": 0.4489,
      "step": 3684
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4627550378525e-06,
      "loss": 0.4995,
      "step": 3685
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.457660624926742e-06,
      "loss": 0.3268,
      "step": 3686
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.452568287302088e-06,
      "loss": 0.5036,
      "step": 3687
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.44747802688535e-06,
      "loss": 0.5796,
      "step": 3688
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4423898455825456e-06,
      "loss": 0.613,
      "step": 3689
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.43730374529892e-06,
      "loss": 0.572,
      "step": 3690
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.432219727938936e-06,
      "loss": 0.4195,
      "step": 3691
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.42713779540628e-06,
      "loss": 0.5286,
      "step": 3692
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.422057949603855e-06,
      "loss": 0.4141,
      "step": 3693
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.416980192433784e-06,
      "loss": 0.3977,
      "step": 3694
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.411904525797408e-06,
      "loss": 0.4681,
      "step": 3695
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.4068309515952855e-06,
      "loss": 0.4813,
      "step": 3696
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.40175947172719e-06,
      "loss": 0.4185,
      "step": 3697
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.39669008809211e-06,
      "loss": 0.5414,
      "step": 3698
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.39162280258825e-06,
      "loss": 0.3841,
      "step": 3699
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.386557617113039e-06,
      "loss": 0.4233,
      "step": 3700
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3814945335631e-06,
      "loss": 0.5064,
      "step": 3701
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3764335538342785e-06,
      "loss": 0.4953,
      "step": 3702
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.371374679821639e-06,
      "loss": 0.5123,
      "step": 3703
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.366317913419454e-06,
      "loss": 0.4325,
      "step": 3704
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3612632565211935e-06,
      "loss": 0.4391,
      "step": 3705
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.356210711019556e-06,
      "loss": 0.6194,
      "step": 3706
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.351160278806444e-06,
      "loss": 0.4244,
      "step": 3707
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.346111961772956e-06,
      "loss": 0.5246,
      "step": 3708
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.341065761809418e-06,
      "loss": 0.475,
      "step": 3709
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.336021680805353e-06,
      "loss": 0.4625,
      "step": 3710
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.330979720649491e-06,
      "loss": 0.4219,
      "step": 3711
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3259398832297665e-06,
      "loss": 0.5057,
      "step": 3712
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.32090217043332e-06,
      "loss": 0.5554,
      "step": 3713
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.3158665841465074e-06,
      "loss": 0.4477,
      "step": 3714
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.310833126254868e-06,
      "loss": 0.4001,
      "step": 3715
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.305801798643156e-06,
      "loss": 0.4283,
      "step": 3716
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.300772603195335e-06,
      "loss": 0.5898,
      "step": 3717
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2957455417945535e-06,
      "loss": 0.4937,
      "step": 3718
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.29072061632317e-06,
      "loss": 0.4358,
      "step": 3719
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.285697828662746e-06,
      "loss": 0.4243,
      "step": 3720
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.280677180694043e-06,
      "loss": 0.5556,
      "step": 3721
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.275658674297004e-06,
      "loss": 0.5003,
      "step": 3722
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.270642311350796e-06,
      "loss": 0.3587,
      "step": 3723
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.2656280937337665e-06,
      "loss": 0.4449,
      "step": 3724
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.260616023323464e-06,
      "loss": 0.5068,
      "step": 3725
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.255606101996633e-06,
      "loss": 0.4196,
      "step": 3726
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.250598331629215e-06,
      "loss": 0.549,
      "step": 3727
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.245592714096342e-06,
      "loss": 0.493,
      "step": 3728
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.240589251272342e-06,
      "loss": 0.5204,
      "step": 3729
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.235587945030739e-06,
      "loss": 0.5819,
      "step": 3730
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.230588797244246e-06,
      "loss": 0.4804,
      "step": 3731
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.225591809784769e-06,
      "loss": 0.4089,
      "step": 3732
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.220596984523404e-06,
      "loss": 0.578,
      "step": 3733
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.215604323330435e-06,
      "loss": 0.3765,
      "step": 3734
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.210613828075349e-06,
      "loss": 0.6591,
      "step": 3735
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.2056255006268025e-06,
      "loss": 0.4925,
      "step": 3736
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.200639342852648e-06,
      "loss": 0.5362,
      "step": 3737
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.19565535661994e-06,
      "loss": 0.4369,
      "step": 3738
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1906735437948945e-06,
      "loss": 0.476,
      "step": 3739
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.185693906242927e-06,
      "loss": 0.4103,
      "step": 3740
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.180716445828645e-06,
      "loss": 0.4415,
      "step": 3741
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.175741164415833e-06,
      "loss": 0.3573,
      "step": 3742
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.17076806386745e-06,
      "loss": 0.4523,
      "step": 3743
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.165797146045658e-06,
      "loss": 0.4847,
      "step": 3744
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.160828412811791e-06,
      "loss": 0.4664,
      "step": 3745
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.155861866026364e-06,
      "loss": 0.4131,
      "step": 3746
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.150897507549076e-06,
      "loss": 0.4107,
      "step": 3747
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1459353392388065e-06,
      "loss": 0.4945,
      "step": 3748
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.140975362953615e-06,
      "loss": 0.5322,
      "step": 3749
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.136017580550739e-06,
      "loss": 0.3721,
      "step": 3750
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.13106199388659e-06,
      "loss": 0.4273,
      "step": 3751
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1261086048167766e-06,
      "loss": 0.4735,
      "step": 3752
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.121157415196057e-06,
      "loss": 0.4288,
      "step": 3753
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1162084268783795e-06,
      "loss": 0.3539,
      "step": 3754
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.111261641716876e-06,
      "loss": 0.3638,
      "step": 3755
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.106317061563846e-06,
      "loss": 0.2784,
      "step": 3756
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.101374688270751e-06,
      "loss": 0.39,
      "step": 3757
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0964345236882476e-06,
      "loss": 0.5442,
      "step": 3758
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.091496569666157e-06,
      "loss": 0.4436,
      "step": 3759
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.086560828053459e-06,
      "loss": 0.511,
      "step": 3760
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.08162730069833e-06,
      "loss": 0.4802,
      "step": 3761
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0766959894481e-06,
      "loss": 0.5404,
      "step": 3762
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0717668961492725e-06,
      "loss": 0.4397,
      "step": 3763
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.066840022647525e-06,
      "loss": 0.4152,
      "step": 3764
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.061915370787697e-06,
      "loss": 0.5366,
      "step": 3765
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.056992942413801e-06,
      "loss": 0.3908,
      "step": 3766
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.052072739369015e-06,
      "loss": 0.4283,
      "step": 3767
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0471547634956865e-06,
      "loss": 0.3916,
      "step": 3768
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.04223901663532e-06,
      "loss": 0.5941,
      "step": 3769
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.037325500628605e-06,
      "loss": 0.4506,
      "step": 3770
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0324142173153715e-06,
      "loss": 0.5095,
      "step": 3771
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.027505168534625e-06,
      "loss": 0.4346,
      "step": 3772
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.022598356124545e-06,
      "loss": 0.5739,
      "step": 3773
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.017693781922453e-06,
      "loss": 0.3817,
      "step": 3774
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.01279144776484e-06,
      "loss": 0.3951,
      "step": 3775
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.007891355487368e-06,
      "loss": 0.3372,
      "step": 3776
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0029935069248494e-06,
      "loss": 0.522,
      "step": 3777
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.998097903911258e-06,
      "loss": 0.3482,
      "step": 3778
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.9932045482797285e-06,
      "loss": 0.4726,
      "step": 3779
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.9883134418625535e-06,
      "loss": 0.4388,
      "step": 3780
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.9834245864911824e-06,
      "loss": 0.4862,
      "step": 3781
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.978537983996222e-06,
      "loss": 0.7028,
      "step": 3782
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.973653636207437e-06,
      "loss": 0.6082,
      "step": 3783
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.968771544953748e-06,
      "loss": 0.4589,
      "step": 3784
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.963891712063227e-06,
      "loss": 0.5366,
      "step": 3785
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.959014139363101e-06,
      "loss": 0.6103,
      "step": 3786
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.954138828679762e-06,
      "loss": 0.4146,
      "step": 3787
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.949265781838737e-06,
      "loss": 0.56,
      "step": 3788
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.944395000664713e-06,
      "loss": 0.5134,
      "step": 3789
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.939526486981534e-06,
      "loss": 0.4458,
      "step": 3790
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.934660242612194e-06,
      "loss": 0.46,
      "step": 3791
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.929796269378822e-06,
      "loss": 0.576,
      "step": 3792
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.924934569102719e-06,
      "loss": 0.432,
      "step": 3793
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.9200751436043225e-06,
      "loss": 0.4745,
      "step": 3794
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.915217994703212e-06,
      "loss": 0.4973,
      "step": 3795
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.910363124218132e-06,
      "loss": 0.465,
      "step": 3796
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.905510533966959e-06,
      "loss": 0.6343,
      "step": 3797
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.900660225766725e-06,
      "loss": 0.3217,
      "step": 3798
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8958122014336e-06,
      "loss": 0.4499,
      "step": 3799
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8909664627829034e-06,
      "loss": 0.4408,
      "step": 3800
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.886123011629098e-06,
      "loss": 0.4612,
      "step": 3801
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.881281849785789e-06,
      "loss": 0.4905,
      "step": 3802
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.876442979065727e-06,
      "loss": 0.3577,
      "step": 3803
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.871606401280801e-06,
      "loss": 0.587,
      "step": 3804
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.866772118242044e-06,
      "loss": 0.5098,
      "step": 3805
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.861940131759629e-06,
      "loss": 0.4247,
      "step": 3806
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.857110443642864e-06,
      "loss": 0.5158,
      "step": 3807
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.852283055700213e-06,
      "loss": 0.4944,
      "step": 3808
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.847457969739254e-06,
      "loss": 0.5539,
      "step": 3809
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.842635187566727e-06,
      "loss": 0.478,
      "step": 3810
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.837814710988492e-06,
      "loss": 0.6128,
      "step": 3811
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.832996541809555e-06,
      "loss": 0.5673,
      "step": 3812
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.828180681834055e-06,
      "loss": 0.5087,
      "step": 3813
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.823367132865266e-06,
      "loss": 0.5193,
      "step": 3814
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8185558967055965e-06,
      "loss": 0.5362,
      "step": 3815
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8137469751565912e-06,
      "loss": 0.5483,
      "step": 3816
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.8089403700189254e-06,
      "loss": 0.3834,
      "step": 3817
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.804136083092409e-06,
      "loss": 0.5153,
      "step": 3818
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7993341161759824e-06,
      "loss": 0.5049,
      "step": 3819
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.794534471067718e-06,
      "loss": 0.5341,
      "step": 3820
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7897371495648163e-06,
      "loss": 0.5074,
      "step": 3821
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7849421534636187e-06,
      "loss": 0.483,
      "step": 3822
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7801494845595787e-06,
      "loss": 0.6113,
      "step": 3823
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7753591446472872e-06,
      "loss": 0.3917,
      "step": 3824
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.770571135520472e-06,
      "loss": 0.578,
      "step": 3825
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.76578545897197e-06,
      "loss": 0.4042,
      "step": 3826
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7610021167937526e-06,
      "loss": 0.5865,
      "step": 3827
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.756221110776925e-06,
      "loss": 0.3881,
      "step": 3828
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7514424427117124e-06,
      "loss": 0.5404,
      "step": 3829
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7466661143874504e-06,
      "loss": 0.4527,
      "step": 3830
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7418921275926245e-06,
      "loss": 0.477,
      "step": 3831
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7371204841148233e-06,
      "loss": 0.5471,
      "step": 3832
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7323511857407666e-06,
      "loss": 0.4987,
      "step": 3833
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.727584234256295e-06,
      "loss": 0.6179,
      "step": 3834
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.7228196314463672e-06,
      "loss": 0.4221,
      "step": 3835
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.718057379095067e-06,
      "loss": 0.5962,
      "step": 3836
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.713297478985595e-06,
      "loss": 0.486,
      "step": 3837
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.708539932900267e-06,
      "loss": 0.4185,
      "step": 3838
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.703784742620534e-06,
      "loss": 0.4756,
      "step": 3839
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.699031909926942e-06,
      "loss": 0.4113,
      "step": 3840
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6942814365991673e-06,
      "loss": 0.3678,
      "step": 3841
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.689533324416006e-06,
      "loss": 0.4985,
      "step": 3842
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6847875751553643e-06,
      "loss": 0.456,
      "step": 3843
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.680044190594254e-06,
      "loss": 0.4022,
      "step": 3844
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.675303172508824e-06,
      "loss": 0.4013,
      "step": 3845
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6705645226743216e-06,
      "loss": 0.584,
      "step": 3846
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6658282428651026e-06,
      "loss": 0.4976,
      "step": 3847
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6610943348546524e-06,
      "loss": 0.4388,
      "step": 3848
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6563628004155548e-06,
      "loss": 0.515,
      "step": 3849
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.651633641319511e-06,
      "loss": 0.6309,
      "step": 3850
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.646906859337329e-06,
      "loss": 0.5026,
      "step": 3851
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6421824562389306e-06,
      "loss": 0.5728,
      "step": 3852
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6374604337933428e-06,
      "loss": 0.4941,
      "step": 3853
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6327407937687045e-06,
      "loss": 0.388,
      "step": 3854
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6280235379322593e-06,
      "loss": 0.7063,
      "step": 3855
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.6233086680503593e-06,
      "loss": 0.3833,
      "step": 3856
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.618596185888471e-06,
      "loss": 0.5553,
      "step": 3857
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.613886093211152e-06,
      "loss": 0.4384,
      "step": 3858
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.609178391782072e-06,
      "loss": 0.4815,
      "step": 3859
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.604473083364016e-06,
      "loss": 0.3943,
      "step": 3860
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5997701697188526e-06,
      "loss": 0.4274,
      "step": 3861
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5950696526075646e-06,
      "loss": 0.4683,
      "step": 3862
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5903715337902444e-06,
      "loss": 0.4619,
      "step": 3863
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5856758150260783e-06,
      "loss": 0.425,
      "step": 3864
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5809824980733445e-06,
      "loss": 0.5336,
      "step": 3865
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.576291584689442e-06,
      "loss": 0.4015,
      "step": 3866
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5716030766308553e-06,
      "loss": 0.4468,
      "step": 3867
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.566916975653175e-06,
      "loss": 0.5315,
      "step": 3868
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5622332835110862e-06,
      "loss": 0.4179,
      "step": 3869
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5575520019583753e-06,
      "loss": 0.538,
      "step": 3870
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5528731327479227e-06,
      "loss": 0.4655,
      "step": 3871
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.548196677631709e-06,
      "loss": 0.458,
      "step": 3872
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5435226383608058e-06,
      "loss": 0.5596,
      "step": 3873
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.538851016685394e-06,
      "loss": 0.4399,
      "step": 3874
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5341818143547267e-06,
      "loss": 0.5176,
      "step": 3875
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.529515033117166e-06,
      "loss": 0.418,
      "step": 3876
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5248506747201694e-06,
      "loss": 0.3836,
      "step": 3877
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5201887409102843e-06,
      "loss": 0.5511,
      "step": 3878
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.515529233433136e-06,
      "loss": 0.6809,
      "step": 3879
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5108721540334643e-06,
      "loss": 0.4412,
      "step": 3880
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5062175044550917e-06,
      "loss": 0.4829,
      "step": 3881
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.5015652864409142e-06,
      "loss": 0.4425,
      "step": 3882
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.496915501732945e-06,
      "loss": 0.4752,
      "step": 3883
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4922681520722677e-06,
      "loss": 0.4303,
      "step": 3884
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4876232391990582e-06,
      "loss": 0.4997,
      "step": 3885
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4829807648525827e-06,
      "loss": 0.3038,
      "step": 3886
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4783407307711913e-06,
      "loss": 0.4452,
      "step": 3887
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.4737031386923216e-06,
      "loss": 0.4766,
      "step": 3888
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.469067990352496e-06,
      "loss": 0.4172,
      "step": 3889
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.464435287487323e-06,
      "loss": 0.3605,
      "step": 3890
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.459805031831496e-06,
      "loss": 0.524,
      "step": 3891
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.45517722511879e-06,
      "loss": 0.4426,
      "step": 3892
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.450551869082063e-06,
      "loss": 0.4629,
      "step": 3893
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4459289654532535e-06,
      "loss": 0.5318,
      "step": 3894
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4413085159633963e-06,
      "loss": 0.5196,
      "step": 3895
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4366905223425827e-06,
      "loss": 0.5,
      "step": 3896
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4320749863199987e-06,
      "loss": 0.3308,
      "step": 3897
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.427461909623916e-06,
      "loss": 0.4668,
      "step": 3898
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.422851293981676e-06,
      "loss": 0.3678,
      "step": 3899
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4182431411196924e-06,
      "loss": 0.422,
      "step": 3900
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.413637452763473e-06,
      "loss": 0.3922,
      "step": 3901
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.4090342306375932e-06,
      "loss": 0.5034,
      "step": 3902
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.404433476465706e-06,
      "loss": 0.4181,
      "step": 3903
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3998351919705396e-06,
      "loss": 0.5289,
      "step": 3904
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.395239378873899e-06,
      "loss": 0.3658,
      "step": 3905
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3906460388966633e-06,
      "loss": 0.4191,
      "step": 3906
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3860551737587857e-06,
      "loss": 0.6117,
      "step": 3907
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.381466785179289e-06,
      "loss": 0.4797,
      "step": 3908
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3768808748762836e-06,
      "loss": 0.5778,
      "step": 3909
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.372297444566929e-06,
      "loss": 0.4297,
      "step": 3910
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.367716495967467e-06,
      "loss": 0.4823,
      "step": 3911
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3631380307932228e-06,
      "loss": 0.5823,
      "step": 3912
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.358562050758568e-06,
      "loss": 0.5463,
      "step": 3913
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3539885575769572e-06,
      "loss": 0.4504,
      "step": 3914
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3494175529609163e-06,
      "loss": 0.4671,
      "step": 3915
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3448490386220355e-06,
      "loss": 0.6241,
      "step": 3916
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3402830162709644e-06,
      "loss": 0.4628,
      "step": 3917
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.335719487617434e-06,
      "loss": 0.5251,
      "step": 3918
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3311584543702324e-06,
      "loss": 0.6705,
      "step": 3919
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3265999182372166e-06,
      "loss": 0.5785,
      "step": 3920
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3220438809253065e-06,
      "loss": 0.5323,
      "step": 3921
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.317490344140488e-06,
      "loss": 0.4916,
      "step": 3922
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3129393095878094e-06,
      "loss": 0.5338,
      "step": 3923
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.3083907789713822e-06,
      "loss": 0.4664,
      "step": 3924
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.303844753994382e-06,
      "loss": 0.4136,
      "step": 3925
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2993012363590447e-06,
      "loss": 0.4656,
      "step": 3926
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2947602277666678e-06,
      "loss": 0.652,
      "step": 3927
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.290221729917609e-06,
      "loss": 0.4904,
      "step": 3928
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.285685744511281e-06,
      "loss": 0.5377,
      "step": 3929
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.281152273246174e-06,
      "loss": 0.4793,
      "step": 3930
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2766213178198112e-06,
      "loss": 0.405,
      "step": 3931
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2720928799287865e-06,
      "loss": 0.5592,
      "step": 3932
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2675669612687565e-06,
      "loss": 0.5377,
      "step": 3933
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2630435635344283e-06,
      "loss": 0.4975,
      "step": 3934
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.258522688419562e-06,
      "loss": 0.4816,
      "step": 3935
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2540043376169784e-06,
      "loss": 0.4347,
      "step": 3936
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2494885128185517e-06,
      "loss": 0.4417,
      "step": 3937
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.244975215715208e-06,
      "loss": 0.4166,
      "step": 3938
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2404644479969294e-06,
      "loss": 0.575,
      "step": 3939
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2359562113527508e-06,
      "loss": 0.4471,
      "step": 3940
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2314505074707593e-06,
      "loss": 0.4033,
      "step": 3941
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.226947338038091e-06,
      "loss": 0.4252,
      "step": 3942
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.2224467047409335e-06,
      "loss": 0.3165,
      "step": 3943
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.217948609264535e-06,
      "loss": 0.4165,
      "step": 3944
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.2134530532931764e-06,
      "loss": 0.6771,
      "step": 3945
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.208960038510195e-06,
      "loss": 0.4269,
      "step": 3946
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.2044695665979865e-06,
      "loss": 0.5372,
      "step": 3947
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.199981639237978e-06,
      "loss": 0.4498,
      "step": 3948
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1954962581106495e-06,
      "loss": 0.497,
      "step": 3949
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.191013424895536e-06,
      "loss": 0.4927,
      "step": 3950
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.186533141271213e-06,
      "loss": 0.4376,
      "step": 3951
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.18205540891529e-06,
      "loss": 0.4356,
      "step": 3952
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.17758022950444e-06,
      "loss": 0.4992,
      "step": 3953
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1731076047143704e-06,
      "loss": 0.3876,
      "step": 3954
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1686375362198307e-06,
      "loss": 0.524,
      "step": 3955
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.164170025694617e-06,
      "loss": 0.4753,
      "step": 3956
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1597050748115655e-06,
      "loss": 0.3757,
      "step": 3957
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1552426852425555e-06,
      "loss": 0.5555,
      "step": 3958
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1507828586585053e-06,
      "loss": 0.4316,
      "step": 3959
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.146325596729375e-06,
      "loss": 0.4768,
      "step": 3960
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.141870901124164e-06,
      "loss": 0.5106,
      "step": 3961
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.137418773510911e-06,
      "loss": 0.3733,
      "step": 3962
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1329692155566884e-06,
      "loss": 0.4776,
      "step": 3963
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1285222289276175e-06,
      "loss": 0.6169,
      "step": 3964
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1240778152888518e-06,
      "loss": 0.607,
      "step": 3965
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.119635976304568e-06,
      "loss": 0.4655,
      "step": 3966
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.115196713638e-06,
      "loss": 0.4431,
      "step": 3967
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1107600289514096e-06,
      "loss": 0.5519,
      "step": 3968
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1063259239060795e-06,
      "loss": 0.4935,
      "step": 3969
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1018944001623486e-06,
      "loss": 0.4343,
      "step": 3970
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.097465459379574e-06,
      "loss": 0.4046,
      "step": 3971
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.093039103216152e-06,
      "loss": 0.4091,
      "step": 3972
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.088615333329509e-06,
      "loss": 0.5641,
      "step": 3973
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.084194151376102e-06,
      "loss": 0.4844,
      "step": 3974
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.079775559011422e-06,
      "loss": 0.4174,
      "step": 3975
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0753595578899885e-06,
      "loss": 0.5111,
      "step": 3976
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0709461496653504e-06,
      "loss": 0.4386,
      "step": 3977
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0665353359900864e-06,
      "loss": 0.4519,
      "step": 3978
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0621271185158043e-06,
      "loss": 0.477,
      "step": 3979
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0577214988931368e-06,
      "loss": 0.3561,
      "step": 3980
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0533184787717452e-06,
      "loss": 0.4431,
      "step": 3981
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0489180598003264e-06,
      "loss": 0.5416,
      "step": 3982
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0445202436265864e-06,
      "loss": 0.4741,
      "step": 3983
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0401250318972643e-06,
      "loss": 0.4461,
      "step": 3984
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.035732426258132e-06,
      "loss": 0.4903,
      "step": 3985
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0313424283539796e-06,
      "loss": 0.3884,
      "step": 3986
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0269550398286096e-06,
      "loss": 0.5311,
      "step": 3987
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0225702623248655e-06,
      "loss": 0.3829,
      "step": 3988
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0181880974846067e-06,
      "loss": 0.4991,
      "step": 3989
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0138085469487033e-06,
      "loss": 0.4277,
      "step": 3990
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0094316123570653e-06,
      "loss": 0.3977,
      "step": 3991
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.0050572953486113e-06,
      "loss": 0.5415,
      "step": 3992
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.00068559756128e-06,
      "loss": 0.4422,
      "step": 3993
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.996316520632034e-06,
      "loss": 0.4658,
      "step": 3994
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.9919500661968494e-06,
      "loss": 0.5025,
      "step": 3995
    },
    {
      "epoch": 0.75,
      "learning_rate": 2.9875862358907302e-06,
      "loss": 0.4415,
      "step": 3996
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.983225031347683e-06,
      "loss": 0.3975,
      "step": 3997
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.978866454200737e-06,
      "loss": 0.4308,
      "step": 3998
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9745105060819524e-06,
      "loss": 0.3831,
      "step": 3999
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.97015718862238e-06,
      "loss": 0.499,
      "step": 4000
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.965806503452098e-06,
      "loss": 0.5143,
      "step": 4001
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9614584522002055e-06,
      "loss": 0.4695,
      "step": 4002
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9571130364948077e-06,
      "loss": 0.467,
      "step": 4003
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9527702579630145e-06,
      "loss": 0.4985,
      "step": 4004
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.948430118230967e-06,
      "loss": 0.52,
      "step": 4005
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9440926189238063e-06,
      "loss": 0.5338,
      "step": 4006
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.939757761665686e-06,
      "loss": 0.3799,
      "step": 4007
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9354255480797713e-06,
      "loss": 0.4223,
      "step": 4008
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.93109597978824e-06,
      "loss": 0.4877,
      "step": 4009
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9267690584122745e-06,
      "loss": 0.4866,
      "step": 4010
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9224447855720707e-06,
      "loss": 0.528,
      "step": 4011
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.91812316288683e-06,
      "loss": 0.4669,
      "step": 4012
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.9138041919747617e-06,
      "loss": 0.4451,
      "step": 4013
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.909487874453084e-06,
      "loss": 0.5917,
      "step": 4014
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.90517421193802e-06,
      "loss": 0.3269,
      "step": 4015
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.900863206044795e-06,
      "loss": 0.4495,
      "step": 4016
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8965548583876534e-06,
      "loss": 0.5395,
      "step": 4017
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.892249170579826e-06,
      "loss": 0.5437,
      "step": 4018
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8879461442335545e-06,
      "loss": 0.5069,
      "step": 4019
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8836457809600914e-06,
      "loss": 0.4299,
      "step": 4020
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8793480823696874e-06,
      "loss": 0.5303,
      "step": 4021
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8750530500715836e-06,
      "loss": 0.5008,
      "step": 4022
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8707606856740435e-06,
      "loss": 0.4259,
      "step": 4023
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.866470990784321e-06,
      "loss": 0.5439,
      "step": 4024
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8621839670086616e-06,
      "loss": 0.3737,
      "step": 4025
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.85789961595233e-06,
      "loss": 0.6321,
      "step": 4026
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.853617939219574e-06,
      "loss": 0.4247,
      "step": 4027
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8493389384136494e-06,
      "loss": 0.3738,
      "step": 4028
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8450626151368044e-06,
      "loss": 0.4416,
      "step": 4029
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8407889709902836e-06,
      "loss": 0.4914,
      "step": 4030
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8365180075743413e-06,
      "loss": 0.4859,
      "step": 4031
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.83224972648821e-06,
      "loss": 0.4401,
      "step": 4032
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8279841293301245e-06,
      "loss": 0.437,
      "step": 4033
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.823721217697326e-06,
      "loss": 0.3967,
      "step": 4034
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.819460993186032e-06,
      "loss": 0.3884,
      "step": 4035
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8152034573914613e-06,
      "loss": 0.4695,
      "step": 4036
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.810948611907832e-06,
      "loss": 0.5464,
      "step": 4037
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8066964583283517e-06,
      "loss": 0.567,
      "step": 4038
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.8024469982452072e-06,
      "loss": 0.4235,
      "step": 4039
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7982002332495974e-06,
      "loss": 0.4406,
      "step": 4040
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.793956164931698e-06,
      "loss": 0.4721,
      "step": 4041
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7897147948806803e-06,
      "loss": 0.3203,
      "step": 4042
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.785476124684704e-06,
      "loss": 0.5424,
      "step": 4043
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7812401559309165e-06,
      "loss": 0.4231,
      "step": 4044
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.777006890205456e-06,
      "loss": 0.5175,
      "step": 4045
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.772776329093447e-06,
      "loss": 0.6944,
      "step": 4046
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7685484741790023e-06,
      "loss": 0.4056,
      "step": 4047
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.764323327045221e-06,
      "loss": 0.5775,
      "step": 4048
    },
    {
      "epoch": 0.76,
      "learning_rate": 2.7601008892741877e-06,
      "loss": 0.4805,
      "step": 4049
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.755881162446974e-06,
      "loss": 0.5077,
      "step": 4050
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.751664148143631e-06,
      "loss": 0.4957,
      "step": 4051
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7474498479432087e-06,
      "loss": 0.5301,
      "step": 4052
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.743238263423721e-06,
      "loss": 0.5686,
      "step": 4053
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.739029396162174e-06,
      "loss": 0.5944,
      "step": 4054
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.734823247734567e-06,
      "loss": 0.3937,
      "step": 4055
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7306198197158616e-06,
      "loss": 0.3549,
      "step": 4056
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7264191136800112e-06,
      "loss": 0.5628,
      "step": 4057
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7222211311999536e-06,
      "loss": 0.5884,
      "step": 4058
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7180258738475994e-06,
      "loss": 0.5179,
      "step": 4059
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7138333431938423e-06,
      "loss": 0.3981,
      "step": 4060
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.709643540808554e-06,
      "loss": 0.5186,
      "step": 4061
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7054564682605857e-06,
      "loss": 0.5683,
      "step": 4062
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.7012721271177643e-06,
      "loss": 0.4678,
      "step": 4063
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6970905189468965e-06,
      "loss": 0.5296,
      "step": 4064
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6929116453137637e-06,
      "loss": 0.6048,
      "step": 4065
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6887355077831247e-06,
      "loss": 0.5011,
      "step": 4066
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6845621079187122e-06,
      "loss": 0.6209,
      "step": 4067
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6803914472832328e-06,
      "loss": 0.3989,
      "step": 4068
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6762235274383775e-06,
      "loss": 0.43,
      "step": 4069
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.672058349944795e-06,
      "loss": 0.3818,
      "step": 4070
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.667895916362114e-06,
      "loss": 0.6568,
      "step": 4071
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6637362282489444e-06,
      "loss": 0.4175,
      "step": 4072
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.659579287162859e-06,
      "loss": 0.5332,
      "step": 4073
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6554250946603954e-06,
      "loss": 0.3919,
      "step": 4074
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6512736522970782e-06,
      "loss": 0.4652,
      "step": 4075
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.647124961627394e-06,
      "loss": 0.4125,
      "step": 4076
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6429790242047927e-06,
      "loss": 0.52,
      "step": 4077
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.638835841581705e-06,
      "loss": 0.4526,
      "step": 4078
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.634695415309524e-06,
      "loss": 0.544,
      "step": 4079
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6305577469386113e-06,
      "loss": 0.4686,
      "step": 4080
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6264228380182955e-06,
      "loss": 0.458,
      "step": 4081
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6222906900968724e-06,
      "loss": 0.3937,
      "step": 4082
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6181613047216047e-06,
      "loss": 0.5559,
      "step": 4083
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.61403468343872e-06,
      "loss": 0.462,
      "step": 4084
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6099108277934105e-06,
      "loss": 0.4818,
      "step": 4085
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6057897393298328e-06,
      "loss": 0.566,
      "step": 4086
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.6016714195911085e-06,
      "loss": 0.4388,
      "step": 4087
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5975558701193183e-06,
      "loss": 0.3881,
      "step": 4088
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5934430924555153e-06,
      "loss": 0.5095,
      "step": 4089
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5893330881397084e-06,
      "loss": 0.5064,
      "step": 4090
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5852258587108582e-06,
      "loss": 0.3859,
      "step": 4091
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5811214057069068e-06,
      "loss": 0.555,
      "step": 4092
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5770197306647416e-06,
      "loss": 0.3634,
      "step": 4093
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.572920835120214e-06,
      "loss": 0.6211,
      "step": 4094
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5688247206081364e-06,
      "loss": 0.6196,
      "step": 4095
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.564731388662276e-06,
      "loss": 0.5079,
      "step": 4096
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.560640840815363e-06,
      "loss": 0.4313,
      "step": 4097
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5565530785990798e-06,
      "loss": 0.456,
      "step": 4098
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.55246810354407e-06,
      "loss": 0.463,
      "step": 4099
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.548385917179933e-06,
      "loss": 0.3461,
      "step": 4100
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5443065210352202e-06,
      "loss": 0.4827,
      "step": 4101
    },
    {
      "epoch": 0.77,
      "learning_rate": 2.5402299166374435e-06,
      "loss": 0.323,
      "step": 4102
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5361561055130625e-06,
      "loss": 0.3566,
      "step": 4103
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.532085089187505e-06,
      "loss": 0.5247,
      "step": 4104
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5280168691851325e-06,
      "loss": 0.3731,
      "step": 4105
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.52395144702927e-06,
      "loss": 0.569,
      "step": 4106
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5198888242422014e-06,
      "loss": 0.503,
      "step": 4107
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.515829002345155e-06,
      "loss": 0.3921,
      "step": 4108
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.511771982858301e-06,
      "loss": 0.2602,
      "step": 4109
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.507717767300778e-06,
      "loss": 0.4531,
      "step": 4110
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.5036663571906693e-06,
      "loss": 0.5123,
      "step": 4111
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4996177540449952e-06,
      "loss": 0.477,
      "step": 4112
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.495571959379742e-06,
      "loss": 0.424,
      "step": 4113
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4915289747098357e-06,
      "loss": 0.5817,
      "step": 4114
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4874888015491516e-06,
      "loss": 0.3698,
      "step": 4115
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4834514414105115e-06,
      "loss": 0.4479,
      "step": 4116
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4794168958056854e-06,
      "loss": 0.3534,
      "step": 4117
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.475385166245388e-06,
      "loss": 0.4428,
      "step": 4118
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.47135625423928e-06,
      "loss": 0.5796,
      "step": 4119
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4673301612959653e-06,
      "loss": 0.5936,
      "step": 4120
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4633068889230027e-06,
      "loss": 0.4907,
      "step": 4121
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.459286438626877e-06,
      "loss": 0.4203,
      "step": 4122
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4552688119130254e-06,
      "loss": 0.5458,
      "step": 4123
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4512540102858338e-06,
      "loss": 0.4572,
      "step": 4124
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.447242035248626e-06,
      "loss": 0.5093,
      "step": 4125
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.443232888303655e-06,
      "loss": 0.4557,
      "step": 4126
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.439226570952137e-06,
      "loss": 0.4032,
      "step": 4127
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4352230846942116e-06,
      "loss": 0.4734,
      "step": 4128
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4312224310289654e-06,
      "loss": 0.446,
      "step": 4129
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.427224611454424e-06,
      "loss": 0.4982,
      "step": 4130
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4232296274675503e-06,
      "loss": 0.42,
      "step": 4131
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4192374805642447e-06,
      "loss": 0.3976,
      "step": 4132
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4152481722393482e-06,
      "loss": 0.3924,
      "step": 4133
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.411261703986636e-06,
      "loss": 0.4225,
      "step": 4134
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.4072780772988224e-06,
      "loss": 0.3734,
      "step": 4135
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.403297293667555e-06,
      "loss": 0.3684,
      "step": 4136
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3993193545834182e-06,
      "loss": 0.4606,
      "step": 4137
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.39534426153593e-06,
      "loss": 0.494,
      "step": 4138
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.391372016013551e-06,
      "loss": 0.6671,
      "step": 4139
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3874026195036615e-06,
      "loss": 0.4033,
      "step": 4140
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.38343607349258e-06,
      "loss": 0.4289,
      "step": 4141
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3794723794655705e-06,
      "loss": 0.5231,
      "step": 4142
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3755115389068086e-06,
      "loss": 0.4162,
      "step": 4143
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3715535532994103e-06,
      "loss": 0.3402,
      "step": 4144
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3675984241254314e-06,
      "loss": 0.4007,
      "step": 4145
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.363646152865847e-06,
      "loss": 0.588,
      "step": 4146
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.35969674100056e-06,
      "loss": 0.341,
      "step": 4147
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3557501900084143e-06,
      "loss": 0.3958,
      "step": 4148
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.351806501367173e-06,
      "loss": 0.3805,
      "step": 4149
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3478656765535322e-06,
      "loss": 0.4854,
      "step": 4150
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.343927717043113e-06,
      "loss": 0.4493,
      "step": 4151
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3399926243104642e-06,
      "loss": 0.5169,
      "step": 4152
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.336060399829061e-06,
      "loss": 0.3678,
      "step": 4153
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3321310450713066e-06,
      "loss": 0.5847,
      "step": 4154
    },
    {
      "epoch": 0.78,
      "learning_rate": 2.3282045615085246e-06,
      "loss": 0.4593,
      "step": 4155
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3242809506109755e-06,
      "loss": 0.4679,
      "step": 4156
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3203602138478264e-06,
      "loss": 0.4131,
      "step": 4157
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3164423526871795e-06,
      "loss": 0.4316,
      "step": 4158
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3125273685960614e-06,
      "loss": 0.4641,
      "step": 4159
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.30861526304042e-06,
      "loss": 0.4391,
      "step": 4160
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.304706037485114e-06,
      "loss": 0.4048,
      "step": 4161
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.3007996933939424e-06,
      "loss": 0.4374,
      "step": 4162
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.296896232229616e-06,
      "loss": 0.421,
      "step": 4163
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.292995655453758e-06,
      "loss": 0.4362,
      "step": 4164
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2890979645269274e-06,
      "loss": 0.4025,
      "step": 4165
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.285203160908592e-06,
      "loss": 0.4894,
      "step": 4166
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.281311246057143e-06,
      "loss": 0.4725,
      "step": 4167
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.277422221429887e-06,
      "loss": 0.4159,
      "step": 4168
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2735360884830506e-06,
      "loss": 0.5743,
      "step": 4169
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.269652848671775e-06,
      "loss": 0.4866,
      "step": 4170
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.265772503450122e-06,
      "loss": 0.3624,
      "step": 4171
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.261895054271066e-06,
      "loss": 0.4548,
      "step": 4172
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2580205025864965e-06,
      "loss": 0.4654,
      "step": 4173
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2541488498472276e-06,
      "loss": 0.4578,
      "step": 4174
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.250280097502973e-06,
      "loss": 0.4985,
      "step": 4175
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2464142470023644e-06,
      "loss": 0.4496,
      "step": 4176
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.242551299792962e-06,
      "loss": 0.4837,
      "step": 4177
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.238691257321217e-06,
      "loss": 0.5034,
      "step": 4178
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.234834121032503e-06,
      "loss": 0.4488,
      "step": 4179
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.230979892371111e-06,
      "loss": 0.4991,
      "step": 4180
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2271285727802373e-06,
      "loss": 0.4707,
      "step": 4181
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2232801637019808e-06,
      "loss": 0.2763,
      "step": 4182
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.219434666577368e-06,
      "loss": 0.4725,
      "step": 4183
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2155920828463217e-06,
      "loss": 0.5044,
      "step": 4184
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2117524139476797e-06,
      "loss": 0.4263,
      "step": 4185
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2079156613191866e-06,
      "loss": 0.4112,
      "step": 4186
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.204081826397494e-06,
      "loss": 0.4848,
      "step": 4187
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.2002509106181625e-06,
      "loss": 0.3832,
      "step": 4188
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.196422915415659e-06,
      "loss": 0.3419,
      "step": 4189
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1925978422233562e-06,
      "loss": 0.4167,
      "step": 4190
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1887756924735393e-06,
      "loss": 0.5241,
      "step": 4191
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1849564675973857e-06,
      "loss": 0.5519,
      "step": 4192
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1811401690249846e-06,
      "loss": 0.4463,
      "step": 4193
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1773267981853362e-06,
      "loss": 0.3482,
      "step": 4194
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.173516356506339e-06,
      "loss": 0.4619,
      "step": 4195
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.169708845414782e-06,
      "loss": 0.5241,
      "step": 4196
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1659042663363795e-06,
      "loss": 0.4797,
      "step": 4197
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1621026206957374e-06,
      "loss": 0.3778,
      "step": 4198
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1583039099163527e-06,
      "loss": 0.5337,
      "step": 4199
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.154508135420642e-06,
      "loss": 0.4348,
      "step": 4200
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.150715298629913e-06,
      "loss": 0.4047,
      "step": 4201
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1469254009643724e-06,
      "loss": 0.3876,
      "step": 4202
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1431384438431303e-06,
      "loss": 0.4484,
      "step": 4203
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1393544286841918e-06,
      "loss": 0.587,
      "step": 4204
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1355733569044633e-06,
      "loss": 0.3627,
      "step": 4205
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1317952299197486e-06,
      "loss": 0.4068,
      "step": 4206
    },
    {
      "epoch": 0.79,
      "learning_rate": 2.1280200491447465e-06,
      "loss": 0.4411,
      "step": 4207
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.124247815993057e-06,
      "loss": 0.4256,
      "step": 4208
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1204785318771717e-06,
      "loss": 0.4039,
      "step": 4209
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1167121982084815e-06,
      "loss": 0.4994,
      "step": 4210
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1129488163972667e-06,
      "loss": 0.6369,
      "step": 4211
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1091883878527174e-06,
      "loss": 0.4348,
      "step": 4212
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1054309139828945e-06,
      "loss": 0.4072,
      "step": 4213
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.1016763961947728e-06,
      "loss": 0.4269,
      "step": 4214
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0979248358942117e-06,
      "loss": 0.4938,
      "step": 4215
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.094176234485963e-06,
      "loss": 0.6142,
      "step": 4216
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0904305933736714e-06,
      "loss": 0.3099,
      "step": 4217
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0866879139598738e-06,
      "loss": 0.4814,
      "step": 4218
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.082948197645999e-06,
      "loss": 0.427,
      "step": 4219
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.079211445832362e-06,
      "loss": 0.3811,
      "step": 4220
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.075477659918174e-06,
      "loss": 0.4214,
      "step": 4221
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0717468413015285e-06,
      "loss": 0.3537,
      "step": 4222
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0680189913794156e-06,
      "loss": 0.5867,
      "step": 4223
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0642941115477087e-06,
      "loss": 0.4768,
      "step": 4224
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0605722032011664e-06,
      "loss": 0.3119,
      "step": 4225
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0568532677334485e-06,
      "loss": 0.2748,
      "step": 4226
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.053137306537082e-06,
      "loss": 0.4182,
      "step": 4227
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0494243210034905e-06,
      "loss": 0.4885,
      "step": 4228
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0457143125229907e-06,
      "loss": 0.3746,
      "step": 4229
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0420072824847693e-06,
      "loss": 0.4857,
      "step": 4230
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0383032322769047e-06,
      "loss": 0.5273,
      "step": 4231
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0346021632863643e-06,
      "loss": 0.4093,
      "step": 4232
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.030904076898996e-06,
      "loss": 0.497,
      "step": 4233
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0272089744995216e-06,
      "loss": 0.4741,
      "step": 4234
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0235168574715604e-06,
      "loss": 0.4934,
      "step": 4235
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.019827727197605e-06,
      "loss": 0.5458,
      "step": 4236
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0161415850590327e-06,
      "loss": 0.5033,
      "step": 4237
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0124584324360998e-06,
      "loss": 0.6008,
      "step": 4238
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.008778270707944e-06,
      "loss": 0.4388,
      "step": 4239
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0051011012525845e-06,
      "loss": 0.5125,
      "step": 4240
    },
    {
      "epoch": 0.8,
      "learning_rate": 2.0014269254469176e-06,
      "loss": 0.5884,
      "step": 4241
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.997755744666722e-06,
      "loss": 0.4486,
      "step": 4242
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9940875602866506e-06,
      "loss": 0.5187,
      "step": 4243
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9904223736802375e-06,
      "loss": 0.3926,
      "step": 4244
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9867601862198906e-06,
      "loss": 0.4412,
      "step": 4245
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9831009992769036e-06,
      "loss": 0.4335,
      "step": 4246
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9794448142214396e-06,
      "loss": 0.5506,
      "step": 4247
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9757916324225313e-06,
      "loss": 0.4198,
      "step": 4248
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.972141455248102e-06,
      "loss": 0.4364,
      "step": 4249
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9684942840649423e-06,
      "loss": 0.3555,
      "step": 4250
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9648501202387103e-06,
      "loss": 0.4667,
      "step": 4251
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9612089651339504e-06,
      "loss": 0.5693,
      "step": 4252
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9575708201140743e-06,
      "loss": 0.4127,
      "step": 4253
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.953935686541366e-06,
      "loss": 0.5141,
      "step": 4254
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9503035657769832e-06,
      "loss": 0.4444,
      "step": 4255
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.946674459180955e-06,
      "loss": 0.4253,
      "step": 4256
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9430483681121836e-06,
      "loss": 0.4052,
      "step": 4257
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9394252939284385e-06,
      "loss": 0.3543,
      "step": 4258
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9358052379863622e-06,
      "loss": 0.5268,
      "step": 4259
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.932188201641465e-06,
      "loss": 0.3659,
      "step": 4260
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9285741862481343e-06,
      "loss": 0.3849,
      "step": 4261
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9249631931596136e-06,
      "loss": 0.548,
      "step": 4262
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9213552237280196e-06,
      "loss": 0.5072,
      "step": 4263
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.917750279304347e-06,
      "loss": 0.6105,
      "step": 4264
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9141483612384415e-06,
      "loss": 0.4785,
      "step": 4265
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9105494708790227e-06,
      "loss": 0.4302,
      "step": 4266
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9069536095736817e-06,
      "loss": 0.3906,
      "step": 4267
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9033607786688724e-06,
      "loss": 0.5944,
      "step": 4268
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.899770979509903e-06,
      "loss": 0.548,
      "step": 4269
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8961842134409636e-06,
      "loss": 0.348,
      "step": 4270
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8926004818050991e-06,
      "loss": 0.4113,
      "step": 4271
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8890197859442182e-06,
      "loss": 0.459,
      "step": 4272
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8854421271990964e-06,
      "loss": 0.4815,
      "step": 4273
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8818675069093684e-06,
      "loss": 0.4311,
      "step": 4274
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8782959264135347e-06,
      "loss": 0.5172,
      "step": 4275
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8747273870489536e-06,
      "loss": 0.5479,
      "step": 4276
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8711618901518446e-06,
      "loss": 0.5785,
      "step": 4277
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.867599437057298e-06,
      "loss": 0.5492,
      "step": 4278
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.86404002909925e-06,
      "loss": 0.3489,
      "step": 4279
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8604836676105009e-06,
      "loss": 0.5162,
      "step": 4280
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.856930353922719e-06,
      "loss": 0.6864,
      "step": 4281
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8533800893664245e-06,
      "loss": 0.5591,
      "step": 4282
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.849832875270987e-06,
      "loss": 0.3896,
      "step": 4283
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8462887129646534e-06,
      "loss": 0.3685,
      "step": 4284
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.842747603774515e-06,
      "loss": 0.4057,
      "step": 4285
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.839209549026516e-06,
      "loss": 0.4456,
      "step": 4286
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8356745500454699e-06,
      "loss": 0.2935,
      "step": 4287
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8321426081550375e-06,
      "loss": 0.585,
      "step": 4288
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8286137246777369e-06,
      "loss": 0.6112,
      "step": 4289
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8250879009349398e-06,
      "loss": 0.4596,
      "step": 4290
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8215651382468745e-06,
      "loss": 0.4268,
      "step": 4291
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8180454379326207e-06,
      "loss": 0.5969,
      "step": 4292
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8145288013101124e-06,
      "loss": 0.6214,
      "step": 4293
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.811015229696137e-06,
      "loss": 0.4836,
      "step": 4294
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8075047244063338e-06,
      "loss": 0.4509,
      "step": 4295
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8039972867551935e-06,
      "loss": 0.4505,
      "step": 4296
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8004929180560582e-06,
      "loss": 0.4865,
      "step": 4297
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.79699161962112e-06,
      "loss": 0.4492,
      "step": 4298
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7934933927614284e-06,
      "loss": 0.5445,
      "step": 4299
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.789998238786871e-06,
      "loss": 0.4584,
      "step": 4300
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7865061590061884e-06,
      "loss": 0.4298,
      "step": 4301
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7830171547269792e-06,
      "loss": 0.3315,
      "step": 4302
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.779531227255683e-06,
      "loss": 0.4269,
      "step": 4303
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7760483778975801e-06,
      "loss": 0.493,
      "step": 4304
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7725686079568138e-06,
      "loss": 0.5213,
      "step": 4305
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7690919187363664e-06,
      "loss": 0.4437,
      "step": 4306
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7656183115380577e-06,
      "loss": 0.4102,
      "step": 4307
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7621477876625725e-06,
      "loss": 0.3597,
      "step": 4308
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7586803484094272e-06,
      "loss": 0.4728,
      "step": 4309
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7552159950769864e-06,
      "loss": 0.3907,
      "step": 4310
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7517547289624605e-06,
      "loss": 0.3175,
      "step": 4311
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7482965513619011e-06,
      "loss": 0.2884,
      "step": 4312
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7448414635702126e-06,
      "loss": 0.3995,
      "step": 4313
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7413894668811272e-06,
      "loss": 0.3587,
      "step": 4314
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7379405625872281e-06,
      "loss": 0.5239,
      "step": 4315
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7344947519799483e-06,
      "loss": 0.5055,
      "step": 4316
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7310520363495454e-06,
      "loss": 0.4996,
      "step": 4317
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7276124169851272e-06,
      "loss": 0.3903,
      "step": 4318
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7241758951746479e-06,
      "loss": 0.5545,
      "step": 4319
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.720742472204896e-06,
      "loss": 0.4664,
      "step": 4320
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7173121493614908e-06,
      "loss": 0.5095,
      "step": 4321
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7138849279289072e-06,
      "loss": 0.3635,
      "step": 4322
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7104608091904484e-06,
      "loss": 0.3831,
      "step": 4323
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.707039794428259e-06,
      "loss": 0.4692,
      "step": 4324
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.703621884923321e-06,
      "loss": 0.4907,
      "step": 4325
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7002070819554527e-06,
      "loss": 0.4236,
      "step": 4326
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6967953868033104e-06,
      "loss": 0.6326,
      "step": 4327
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6933868007443854e-06,
      "loss": 0.415,
      "step": 4328
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6899813250550068e-06,
      "loss": 0.4109,
      "step": 4329
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6865789610103367e-06,
      "loss": 0.4963,
      "step": 4330
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6831797098843738e-06,
      "loss": 0.3788,
      "step": 4331
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6797835729499501e-06,
      "loss": 0.4441,
      "step": 4332
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6763905514787304e-06,
      "loss": 0.6014,
      "step": 4333
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.673000646741222e-06,
      "loss": 0.4545,
      "step": 4334
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6696138600067491e-06,
      "loss": 0.4254,
      "step": 4335
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6662301925434786e-06,
      "loss": 0.4602,
      "step": 4336
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6628496456184107e-06,
      "loss": 0.3568,
      "step": 4337
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6594722204973756e-06,
      "loss": 0.6188,
      "step": 4338
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6560979184450254e-06,
      "loss": 0.5029,
      "step": 4339
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6527267407248583e-06,
      "loss": 0.2604,
      "step": 4340
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6493586885991908e-06,
      "loss": 0.5635,
      "step": 4341
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6459937633291756e-06,
      "loss": 0.481,
      "step": 4342
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6426319661747902e-06,
      "loss": 0.4804,
      "step": 4343
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6392732983948434e-06,
      "loss": 0.5107,
      "step": 4344
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6359177612469711e-06,
      "loss": 0.4547,
      "step": 4345
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6325653559876375e-06,
      "loss": 0.4127,
      "step": 4346
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6292160838721316e-06,
      "loss": 0.3686,
      "step": 4347
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.625869946154578e-06,
      "loss": 0.4092,
      "step": 4348
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6225269440879144e-06,
      "loss": 0.5758,
      "step": 4349
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6191870789239116e-06,
      "loss": 0.3886,
      "step": 4350
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.615850351913173e-06,
      "loss": 0.377,
      "step": 4351
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6125167643051098e-06,
      "loss": 0.3583,
      "step": 4352
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.6091863173479694e-06,
      "loss": 0.4211,
      "step": 4353
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.605859012288824e-06,
      "loss": 0.4417,
      "step": 4354
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.602534850373567e-06,
      "loss": 0.4318,
      "step": 4355
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5992138328469077e-06,
      "loss": 0.5362,
      "step": 4356
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5958959609523905e-06,
      "loss": 0.5131,
      "step": 4357
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5925812359323745e-06,
      "loss": 0.4893,
      "step": 4358
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.589269659028041e-06,
      "loss": 0.3718,
      "step": 4359
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5859612314793927e-06,
      "loss": 0.5238,
      "step": 4360
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.582655954525255e-06,
      "loss": 0.3914,
      "step": 4361
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5793538294032717e-06,
      "loss": 0.4338,
      "step": 4362
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5760548573499068e-06,
      "loss": 0.4023,
      "step": 4363
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5727590396004445e-06,
      "loss": 0.4562,
      "step": 4364
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5694663773889851e-06,
      "loss": 0.4707,
      "step": 4365
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.5661768719484526e-06,
      "loss": 0.5468,
      "step": 4366
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.562890524510583e-06,
      "loss": 0.5161,
      "step": 4367
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5596073363059316e-06,
      "loss": 0.4573,
      "step": 4368
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5563273085638785e-06,
      "loss": 0.426,
      "step": 4369
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5530504425126047e-06,
      "loss": 0.4628,
      "step": 4370
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.549776739379122e-06,
      "loss": 0.3811,
      "step": 4371
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.546506200389253e-06,
      "loss": 0.4728,
      "step": 4372
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.543238826767627e-06,
      "loss": 0.529,
      "step": 4373
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.539974619737704e-06,
      "loss": 0.5223,
      "step": 4374
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.536713580521746e-06,
      "loss": 0.4623,
      "step": 4375
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5334557103408332e-06,
      "loss": 0.3491,
      "step": 4376
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.530201010414859e-06,
      "loss": 0.5557,
      "step": 4377
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5269494819625286e-06,
      "loss": 0.4762,
      "step": 4378
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5237011262013624e-06,
      "loss": 0.4682,
      "step": 4379
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5204559443476886e-06,
      "loss": 0.4267,
      "step": 4380
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5172139376166507e-06,
      "loss": 0.3609,
      "step": 4381
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.513975107222201e-06,
      "loss": 0.3598,
      "step": 4382
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5107394543771037e-06,
      "loss": 0.4746,
      "step": 4383
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.507506980292933e-06,
      "loss": 0.3881,
      "step": 4384
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.50427768618007e-06,
      "loss": 0.3472,
      "step": 4385
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.5010515732477138e-06,
      "loss": 0.3608,
      "step": 4386
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4978286427038602e-06,
      "loss": 0.5057,
      "step": 4387
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4946088957553196e-06,
      "loss": 0.4629,
      "step": 4388
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.491392333607713e-06,
      "loss": 0.4646,
      "step": 4389
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4881789574654681e-06,
      "loss": 0.3662,
      "step": 4390
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4849687685318092e-06,
      "loss": 0.3658,
      "step": 4391
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4817617680087826e-06,
      "loss": 0.5085,
      "step": 4392
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.478557957097233e-06,
      "loss": 0.5192,
      "step": 4393
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4753573369968054e-06,
      "loss": 0.4579,
      "step": 4394
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4721599089059624e-06,
      "loss": 0.42,
      "step": 4395
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4689656740219615e-06,
      "loss": 0.3216,
      "step": 4396
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4657746335408695e-06,
      "loss": 0.5495,
      "step": 4397
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4625867886575552e-06,
      "loss": 0.4509,
      "step": 4398
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.459402140565691e-06,
      "loss": 0.4841,
      "step": 4399
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4562206904577514e-06,
      "loss": 0.4082,
      "step": 4400
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4530424395250154e-06,
      "loss": 0.521,
      "step": 4401
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.44986738895756e-06,
      "loss": 0.4269,
      "step": 4402
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.446695539944275e-06,
      "loss": 0.4324,
      "step": 4403
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4435268936728352e-06,
      "loss": 0.5534,
      "step": 4404
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4403614513297248e-06,
      "loss": 0.4441,
      "step": 4405
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4371992141002323e-06,
      "loss": 0.4878,
      "step": 4406
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4340401831684413e-06,
      "loss": 0.3497,
      "step": 4407
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4308843597172284e-06,
      "loss": 0.4469,
      "step": 4408
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4277317449282834e-06,
      "loss": 0.4753,
      "step": 4409
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4245823399820835e-06,
      "loss": 0.441,
      "step": 4410
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4214361460579086e-06,
      "loss": 0.4567,
      "step": 4411
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4182931643338337e-06,
      "loss": 0.3356,
      "step": 4412
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4151533959867348e-06,
      "loss": 0.4192,
      "step": 4413
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4120168421922798e-06,
      "loss": 0.4289,
      "step": 4414
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4088835041249383e-06,
      "loss": 0.492,
      "step": 4415
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4057533829579706e-06,
      "loss": 0.512,
      "step": 4416
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.4026264798634359e-06,
      "loss": 0.4437,
      "step": 4417
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3995027960121866e-06,
      "loss": 0.3996,
      "step": 4418
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3963823325738713e-06,
      "loss": 0.4932,
      "step": 4419
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3932650907169298e-06,
      "loss": 0.3365,
      "step": 4420
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.390151071608603e-06,
      "loss": 0.5138,
      "step": 4421
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3870402764149149e-06,
      "loss": 0.6978,
      "step": 4422
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3839327063006847e-06,
      "loss": 0.4263,
      "step": 4423
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3808283624295337e-06,
      "loss": 0.3676,
      "step": 4424
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3777272459638646e-06,
      "loss": 0.4164,
      "step": 4425
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3746293580648718e-06,
      "loss": 0.3535,
      "step": 4426
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.371534699892547e-06,
      "loss": 0.3616,
      "step": 4427
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3684432726056717e-06,
      "loss": 0.5169,
      "step": 4428
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3653550773618085e-06,
      "loss": 0.4073,
      "step": 4429
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3622701153173222e-06,
      "loss": 0.4144,
      "step": 4430
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3591883876273594e-06,
      "loss": 0.4666,
      "step": 4431
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3561098954458585e-06,
      "loss": 0.4784,
      "step": 4432
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3530346399255444e-06,
      "loss": 0.4376,
      "step": 4433
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3499626222179286e-06,
      "loss": 0.4778,
      "step": 4434
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3468938434733203e-06,
      "loss": 0.4653,
      "step": 4435
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3438283048408019e-06,
      "loss": 0.3715,
      "step": 4436
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3407660074682472e-06,
      "loss": 0.494,
      "step": 4437
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3377069525023267e-06,
      "loss": 0.5235,
      "step": 4438
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.33465114108848e-06,
      "loss": 0.3026,
      "step": 4439
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3315985743709403e-06,
      "loss": 0.5725,
      "step": 4440
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3285492534927313e-06,
      "loss": 0.4422,
      "step": 4441
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3255031795956552e-06,
      "loss": 0.4291,
      "step": 4442
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3224603538202929e-06,
      "loss": 0.3817,
      "step": 4443
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3194207773060219e-06,
      "loss": 0.3826,
      "step": 4444
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3163844511909928e-06,
      "loss": 0.4083,
      "step": 4445
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3133513766121442e-06,
      "loss": 0.4775,
      "step": 4446
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3103215547051962e-06,
      "loss": 0.3876,
      "step": 4447
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.307294986604648e-06,
      "loss": 0.5535,
      "step": 4448
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3042716734437843e-06,
      "loss": 0.4701,
      "step": 4449
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3012516163546685e-06,
      "loss": 0.4474,
      "step": 4450
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.298234816468148e-06,
      "loss": 0.3249,
      "step": 4451
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.295221274913846e-06,
      "loss": 0.6299,
      "step": 4452
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2922109928201699e-06,
      "loss": 0.3859,
      "step": 4453
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2892039713143024e-06,
      "loss": 0.5043,
      "step": 4454
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2862002115222071e-06,
      "loss": 0.6504,
      "step": 4455
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2831997145686326e-06,
      "loss": 0.3192,
      "step": 4456
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2802024815770942e-06,
      "loss": 0.4336,
      "step": 4457
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.277208513669891e-06,
      "loss": 0.3456,
      "step": 4458
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2742178119681037e-06,
      "loss": 0.4172,
      "step": 4459
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2712303775915803e-06,
      "loss": 0.4784,
      "step": 4460
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2682462116589512e-06,
      "loss": 0.2923,
      "step": 4461
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.265265315287625e-06,
      "loss": 0.442,
      "step": 4462
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2622876895937842e-06,
      "loss": 0.5033,
      "step": 4463
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2593133356923803e-06,
      "loss": 0.4039,
      "step": 4464
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2563422546971504e-06,
      "loss": 0.5379,
      "step": 4465
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.253374447720599e-06,
      "loss": 0.404,
      "step": 4466
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.250409915874007e-06,
      "loss": 0.4198,
      "step": 4467
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2474486602674274e-06,
      "loss": 0.5725,
      "step": 4468
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2444906820096881e-06,
      "loss": 0.3667,
      "step": 4469
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2415359822083893e-06,
      "loss": 0.5099,
      "step": 4470
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2385845619699032e-06,
      "loss": 0.553,
      "step": 4471
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2356364223993723e-06,
      "loss": 0.377,
      "step": 4472
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2326915646007187e-06,
      "loss": 0.5357,
      "step": 4473
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2297499896766241e-06,
      "loss": 0.5247,
      "step": 4474
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.226811698728546e-06,
      "loss": 0.4649,
      "step": 4475
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2238766928567158e-06,
      "loss": 0.4344,
      "step": 4476
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.220944973160133e-06,
      "loss": 0.4309,
      "step": 4477
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2180165407365586e-06,
      "loss": 0.5159,
      "step": 4478
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2150913966825361e-06,
      "loss": 0.4978,
      "step": 4479
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.212169542093371e-06,
      "loss": 0.4029,
      "step": 4480
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.209250978063129e-06,
      "loss": 0.3945,
      "step": 4481
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2063357056846604e-06,
      "loss": 0.3731,
      "step": 4482
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2034237260495708e-06,
      "loss": 0.3561,
      "step": 4483
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2005150402482369e-06,
      "loss": 0.6454,
      "step": 4484
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1976096493698008e-06,
      "loss": 0.5069,
      "step": 4485
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1947075545021724e-06,
      "loss": 0.4338,
      "step": 4486
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1918087567320257e-06,
      "loss": 0.4902,
      "step": 4487
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1889132571448025e-06,
      "loss": 0.4461,
      "step": 4488
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1860210568247055e-06,
      "loss": 0.5654,
      "step": 4489
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1831321568547072e-06,
      "loss": 0.475,
      "step": 4490
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.180246558316539e-06,
      "loss": 0.5116,
      "step": 4491
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1773642622907012e-06,
      "loss": 0.4765,
      "step": 4492
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1744852698564523e-06,
      "loss": 0.4673,
      "step": 4493
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1716095820918217e-06,
      "loss": 0.5653,
      "step": 4494
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1687372000735898e-06,
      "loss": 0.4819,
      "step": 4495
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1658681248773096e-06,
      "loss": 0.4153,
      "step": 4496
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1630023575772908e-06,
      "loss": 0.6085,
      "step": 4497
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1601398992466061e-06,
      "loss": 0.3435,
      "step": 4498
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1572807509570883e-06,
      "loss": 0.4097,
      "step": 4499
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1544249137793317e-06,
      "loss": 0.5463,
      "step": 4500
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.151572388782688e-06,
      "loss": 0.4147,
      "step": 4501
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1487231770352736e-06,
      "loss": 0.4242,
      "step": 4502
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1458772796039597e-06,
      "loss": 0.442,
      "step": 4503
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1430346975543793e-06,
      "loss": 0.5083,
      "step": 4504
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1401954319509213e-06,
      "loss": 0.4865,
      "step": 4505
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.137359483856737e-06,
      "loss": 0.6597,
      "step": 4506
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1345268543337283e-06,
      "loss": 0.4086,
      "step": 4507
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.131697544442567e-06,
      "loss": 0.4697,
      "step": 4508
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1288715552426666e-06,
      "loss": 0.7194,
      "step": 4509
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.126048887792206e-06,
      "loss": 0.4776,
      "step": 4510
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1232295431481222e-06,
      "loss": 0.3922,
      "step": 4511
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.120413522366104e-06,
      "loss": 0.547,
      "step": 4512
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.11760082650059e-06,
      "loss": 0.5074,
      "step": 4513
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.114791456604788e-06,
      "loss": 0.3752,
      "step": 4514
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1119854137306507e-06,
      "loss": 0.4769,
      "step": 4515
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1091826989288811e-06,
      "loss": 0.4047,
      "step": 4516
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1063833132489477e-06,
      "loss": 0.4845,
      "step": 4517
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.103587257739064e-06,
      "loss": 0.5022,
      "step": 4518
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.1007945334462e-06,
      "loss": 0.4824,
      "step": 4519
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0980051414160764e-06,
      "loss": 0.3594,
      "step": 4520
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0952190826931674e-06,
      "loss": 0.3615,
      "step": 4521
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0924363583206986e-06,
      "loss": 0.4267,
      "step": 4522
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0896569693406468e-06,
      "loss": 0.5236,
      "step": 4523
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.08688091679374e-06,
      "loss": 0.4887,
      "step": 4524
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.0841082017194583e-06,
      "loss": 0.3948,
      "step": 4525
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.081338825156031e-06,
      "loss": 0.4658,
      "step": 4526
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0785727881404329e-06,
      "loss": 0.3377,
      "step": 4527
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.075810091708399e-06,
      "loss": 0.4933,
      "step": 4528
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0730507368944066e-06,
      "loss": 0.419,
      "step": 4529
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0702947247316765e-06,
      "loss": 0.4352,
      "step": 4530
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0675420562521876e-06,
      "loss": 0.6728,
      "step": 4531
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0647927324866624e-06,
      "loss": 0.4231,
      "step": 4532
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0620467544645719e-06,
      "loss": 0.4546,
      "step": 4533
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0593041232141333e-06,
      "loss": 0.4,
      "step": 4534
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0565648397623108e-06,
      "loss": 0.429,
      "step": 4535
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.053828905134815e-06,
      "loss": 0.4915,
      "step": 4536
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.051096320356103e-06,
      "loss": 0.3471,
      "step": 4537
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0483670864493777e-06,
      "loss": 0.2603,
      "step": 4538
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0456412044365872e-06,
      "loss": 0.5257,
      "step": 4539
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0429186753384234e-06,
      "loss": 0.3231,
      "step": 4540
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0401995001743248e-06,
      "loss": 0.2725,
      "step": 4541
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.037483679962471e-06,
      "loss": 0.3536,
      "step": 4542
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.034771215719792e-06,
      "loss": 0.3998,
      "step": 4543
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0320621084619508e-06,
      "loss": 0.5626,
      "step": 4544
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0293563592033595e-06,
      "loss": 0.5161,
      "step": 4545
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0266539689571775e-06,
      "loss": 0.4284,
      "step": 4546
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0239549387352954e-06,
      "loss": 0.4485,
      "step": 4547
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0212592695483524e-06,
      "loss": 0.4912,
      "step": 4548
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0185669624057327e-06,
      "loss": 0.4101,
      "step": 4549
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0158780183155547e-06,
      "loss": 0.398,
      "step": 4550
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.013192438284676e-06,
      "loss": 0.4233,
      "step": 4551
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0105102233187036e-06,
      "loss": 0.4092,
      "step": 4552
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0078313744219791e-06,
      "loss": 0.5178,
      "step": 4553
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.0051558925975813e-06,
      "loss": 0.3877,
      "step": 4554
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.002483778847333e-06,
      "loss": 0.2311,
      "step": 4555
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.998150341717927e-07,
      "loss": 0.3367,
      "step": 4556
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.97149659570259e-07,
      "loss": 0.288,
      "step": 4557
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.94487656040768e-07,
      "loss": 0.4986,
      "step": 4558
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.91829024580091e-07,
      "loss": 0.5176,
      "step": 4559
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.891737661837463e-07,
      "loss": 0.3319,
      "step": 4560
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.865218818459742e-07,
      "loss": 0.3067,
      "step": 4561
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.838733725597615e-07,
      "loss": 0.4669,
      "step": 4562
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.812282393168315e-07,
      "loss": 0.5565,
      "step": 4563
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.785864831076418e-07,
      "loss": 0.5029,
      "step": 4564
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.759481049213792e-07,
      "loss": 0.5006,
      "step": 4565
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.733131057459767e-07,
      "loss": 0.4564,
      "step": 4566
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.706814865680957e-07,
      "loss": 0.3777,
      "step": 4567
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.680532483731287e-07,
      "loss": 0.4976,
      "step": 4568
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.6542839214521e-07,
      "loss": 0.4417,
      "step": 4569
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.62806918867204e-07,
      "loss": 0.4843,
      "step": 4570
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.601888295207063e-07,
      "loss": 0.3446,
      "step": 4571
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.575741250860482e-07,
      "loss": 0.3729,
      "step": 4572
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.549628065422922e-07,
      "loss": 0.5169,
      "step": 4573
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.52354874867234e-07,
      "loss": 0.4229,
      "step": 4574
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.497503310374012e-07,
      "loss": 0.3014,
      "step": 4575
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.471491760280494e-07,
      "loss": 0.5147,
      "step": 4576
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.445514108131693e-07,
      "loss": 0.5292,
      "step": 4577
    },
    {
      "epoch": 0.86,
      "learning_rate": 9.419570363654851e-07,
      "loss": 0.3667,
      "step": 4578
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.393660536564408e-07,
      "loss": 0.5919,
      "step": 4579
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.367784636562183e-07,
      "loss": 0.5281,
      "step": 4580
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.341942673337335e-07,
      "loss": 0.3432,
      "step": 4581
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.316134656566189e-07,
      "loss": 0.4038,
      "step": 4582
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.290360595912429e-07,
      "loss": 0.3473,
      "step": 4583
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.264620501027066e-07,
      "loss": 0.5126,
      "step": 4584
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.238914381548347e-07,
      "loss": 0.469,
      "step": 4585
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.213242247101739e-07,
      "loss": 0.5398,
      "step": 4586
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.187604107300107e-07,
      "loss": 0.4031,
      "step": 4587
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.161999971743507e-07,
      "loss": 0.375,
      "step": 4588
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.136429850019268e-07,
      "loss": 0.4173,
      "step": 4589
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.110893751701988e-07,
      "loss": 0.3209,
      "step": 4590
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.08539168635355e-07,
      "loss": 0.4489,
      "step": 4591
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.059923663523074e-07,
      "loss": 0.3049,
      "step": 4592
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.034489692746906e-07,
      "loss": 0.4224,
      "step": 4593
    },
    {
      "epoch": 0.87,
      "learning_rate": 9.009089783548686e-07,
      "loss": 0.3851,
      "step": 4594
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.983723945439304e-07,
      "loss": 0.5131,
      "step": 4595
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.958392187916842e-07,
      "loss": 0.5575,
      "step": 4596
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.933094520466634e-07,
      "loss": 0.4109,
      "step": 4597
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.907830952561303e-07,
      "loss": 0.5893,
      "step": 4598
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.88260149366067e-07,
      "loss": 0.3955,
      "step": 4599
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.857406153211722e-07,
      "loss": 0.492,
      "step": 4600
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.832244940648782e-07,
      "loss": 0.5716,
      "step": 4601
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.807117865393344e-07,
      "loss": 0.3964,
      "step": 4602
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.78202493685405e-07,
      "loss": 0.5571,
      "step": 4603
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.756966164426884e-07,
      "loss": 0.5159,
      "step": 4604
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.731941557494949e-07,
      "loss": 0.4399,
      "step": 4605
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.706951125428597e-07,
      "loss": 0.437,
      "step": 4606
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.681994877585365e-07,
      "loss": 0.338,
      "step": 4607
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.657072823309986e-07,
      "loss": 0.4614,
      "step": 4608
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.63218497193441e-07,
      "loss": 0.504,
      "step": 4609
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.607331332777758e-07,
      "loss": 0.3848,
      "step": 4610
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.582511915146363e-07,
      "loss": 0.5553,
      "step": 4611
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.557726728333715e-07,
      "loss": 0.433,
      "step": 4612
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.532975781620511e-07,
      "loss": 0.3816,
      "step": 4613
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.508259084274628e-07,
      "loss": 0.4625,
      "step": 4614
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.483576645551072e-07,
      "loss": 0.3616,
      "step": 4615
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.458928474692118e-07,
      "loss": 0.5007,
      "step": 4616
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.434314580927105e-07,
      "loss": 0.3055,
      "step": 4617
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.409734973472572e-07,
      "loss": 0.312,
      "step": 4618
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.385189661532278e-07,
      "loss": 0.3812,
      "step": 4619
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.360678654297049e-07,
      "loss": 0.5736,
      "step": 4620
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.336201960944935e-07,
      "loss": 0.4656,
      "step": 4621
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.311759590641099e-07,
      "loss": 0.3843,
      "step": 4622
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.287351552537859e-07,
      "loss": 0.568,
      "step": 4623
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.26297785577469e-07,
      "loss": 0.3242,
      "step": 4624
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.238638509478202e-07,
      "loss": 0.445,
      "step": 4625
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.21433352276213e-07,
      "loss": 0.3593,
      "step": 4626
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.19006290472737e-07,
      "loss": 0.4418,
      "step": 4627
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.165826664461918e-07,
      "loss": 0.5307,
      "step": 4628
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.141624811040894e-07,
      "loss": 0.3135,
      "step": 4629
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.117457353526626e-07,
      "loss": 0.4371,
      "step": 4630
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.093324300968442e-07,
      "loss": 0.4245,
      "step": 4631
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.069225662402835e-07,
      "loss": 0.4166,
      "step": 4632
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.045161446853466e-07,
      "loss": 0.414,
      "step": 4633
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.021131663331027e-07,
      "loss": 0.4751,
      "step": 4634
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.997136320833332e-07,
      "loss": 0.3984,
      "step": 4635
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.973175428345359e-07,
      "loss": 0.4691,
      "step": 4636
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.949248994839131e-07,
      "loss": 0.4196,
      "step": 4637
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.92535702927375e-07,
      "loss": 0.4222,
      "step": 4638
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.901499540595492e-07,
      "loss": 0.4218,
      "step": 4639
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.877676537737633e-07,
      "loss": 0.4869,
      "step": 4640
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.853888029620604e-07,
      "loss": 0.3958,
      "step": 4641
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.83013402515187e-07,
      "loss": 0.363,
      "step": 4642
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.806414533226014e-07,
      "loss": 0.461,
      "step": 4643
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.782729562724678e-07,
      "loss": 0.409,
      "step": 4644
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.759079122516566e-07,
      "loss": 0.3841,
      "step": 4645
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.735463221457496e-07,
      "loss": 0.3589,
      "step": 4646
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.711881868390292e-07,
      "loss": 0.5451,
      "step": 4647
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.688335072144882e-07,
      "loss": 0.5478,
      "step": 4648
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.664822841538267e-07,
      "loss": 0.3569,
      "step": 4649
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.641345185374438e-07,
      "loss": 0.3881,
      "step": 4650
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.617902112444542e-07,
      "loss": 0.6265,
      "step": 4651
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.594493631526667e-07,
      "loss": 0.4872,
      "step": 4652
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.571119751386035e-07,
      "loss": 0.4549,
      "step": 4653
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.54778048077488e-07,
      "loss": 0.5077,
      "step": 4654
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.524475828432453e-07,
      "loss": 0.604,
      "step": 4655
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.501205803085076e-07,
      "loss": 0.2191,
      "step": 4656
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.477970413446089e-07,
      "loss": 0.4496,
      "step": 4657
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.454769668215867e-07,
      "loss": 0.3789,
      "step": 4658
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.431603576081814e-07,
      "loss": 0.3441,
      "step": 4659
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.40847214571836e-07,
      "loss": 0.301,
      "step": 4660
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.385375385786952e-07,
      "loss": 0.3613,
      "step": 4661
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.362313304936052e-07,
      "loss": 0.4861,
      "step": 4662
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.33928591180113e-07,
      "loss": 0.5214,
      "step": 4663
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.316293215004689e-07,
      "loss": 0.636,
      "step": 4664
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.29333522315625e-07,
      "loss": 0.4496,
      "step": 4665
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.270411944852285e-07,
      "loss": 0.5043,
      "step": 4666
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.247523388676292e-07,
      "loss": 0.3628,
      "step": 4667
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.224669563198827e-07,
      "loss": 0.4365,
      "step": 4668
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.201850476977346e-07,
      "loss": 0.3645,
      "step": 4669
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.179066138556334e-07,
      "loss": 0.4588,
      "step": 4670
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.156316556467302e-07,
      "loss": 0.3884,
      "step": 4671
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.133601739228735e-07,
      "loss": 0.5064,
      "step": 4672
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.110921695346007e-07,
      "loss": 0.5612,
      "step": 4673
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.088276433311614e-07,
      "loss": 0.3232,
      "step": 4674
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.06566596160494e-07,
      "loss": 0.4567,
      "step": 4675
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.043090288692367e-07,
      "loss": 0.3832,
      "step": 4676
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.020549423027223e-07,
      "loss": 0.4134,
      "step": 4677
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.998043373049845e-07,
      "loss": 0.4939,
      "step": 4678
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.975572147187503e-07,
      "loss": 0.3565,
      "step": 4679
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.953135753854423e-07,
      "loss": 0.3928,
      "step": 4680
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.930734201451817e-07,
      "loss": 0.4446,
      "step": 4681
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.908367498367819e-07,
      "loss": 0.5913,
      "step": 4682
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.886035652977541e-07,
      "loss": 0.4237,
      "step": 4683
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.863738673643006e-07,
      "loss": 0.4456,
      "step": 4684
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.841476568713234e-07,
      "loss": 0.4087,
      "step": 4685
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.819249346524171e-07,
      "loss": 0.4661,
      "step": 4686
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.797057015398634e-07,
      "loss": 0.443,
      "step": 4687
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.774899583646478e-07,
      "loss": 0.3886,
      "step": 4688
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.752777059564431e-07,
      "loss": 0.3969,
      "step": 4689
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.730689451436134e-07,
      "loss": 0.6363,
      "step": 4690
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.708636767532218e-07,
      "loss": 0.3361,
      "step": 4691
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.686619016110186e-07,
      "loss": 0.4692,
      "step": 4692
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.664636205414465e-07,
      "loss": 0.5039,
      "step": 4693
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.642688343676418e-07,
      "loss": 0.4603,
      "step": 4694
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.620775439114302e-07,
      "loss": 0.4471,
      "step": 4695
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.59889749993331e-07,
      "loss": 0.6286,
      "step": 4696
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.577054534325511e-07,
      "loss": 0.4438,
      "step": 4697
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.555246550469907e-07,
      "loss": 0.4633,
      "step": 4698
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.533473556532366e-07,
      "loss": 0.3515,
      "step": 4699
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.511735560665699e-07,
      "loss": 0.5262,
      "step": 4700
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.490032571009586e-07,
      "loss": 0.4825,
      "step": 4701
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.468364595690579e-07,
      "loss": 0.4828,
      "step": 4702
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.44673164282219e-07,
      "loss": 0.4265,
      "step": 4703
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.425133720504717e-07,
      "loss": 0.4584,
      "step": 4704
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.403570836825413e-07,
      "loss": 0.3996,
      "step": 4705
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.382042999858407e-07,
      "loss": 0.4865,
      "step": 4706
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.360550217664685e-07,
      "loss": 0.6954,
      "step": 4707
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.339092498292076e-07,
      "loss": 0.4397,
      "step": 4708
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.317669849775365e-07,
      "loss": 0.3926,
      "step": 4709
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.296282280136135e-07,
      "loss": 0.3287,
      "step": 4710
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.274929797382834e-07,
      "loss": 0.3848,
      "step": 4711
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.253612409510812e-07,
      "loss": 0.4506,
      "step": 4712
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.232330124502262e-07,
      "loss": 0.5207,
      "step": 4713
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.21108295032622e-07,
      "loss": 0.5029,
      "step": 4714
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.189870894938587e-07,
      "loss": 0.4603,
      "step": 4715
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.168693966282091e-07,
      "loss": 0.4384,
      "step": 4716
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.147552172286375e-07,
      "loss": 0.5291,
      "step": 4717
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.12644552086783e-07,
      "loss": 0.5228,
      "step": 4718
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.105374019929733e-07,
      "loss": 0.5607,
      "step": 4719
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.084337677362262e-07,
      "loss": 0.4597,
      "step": 4720
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.063336501042305e-07,
      "loss": 0.4981,
      "step": 4721
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.04237049883365e-07,
      "loss": 0.6502,
      "step": 4722
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.02143967858696e-07,
      "loss": 0.4657,
      "step": 4723
    },
    {
      "epoch": 0.89,
      "learning_rate": 6.000544048139645e-07,
      "loss": 0.4357,
      "step": 4724
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.979683615315957e-07,
      "loss": 0.4124,
      "step": 4725
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.958858387926991e-07,
      "loss": 0.5485,
      "step": 4726
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.938068373770667e-07,
      "loss": 0.657,
      "step": 4727
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.917313580631678e-07,
      "loss": 0.5689,
      "step": 4728
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.896594016281553e-07,
      "loss": 0.3134,
      "step": 4729
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.875909688478643e-07,
      "loss": 0.2703,
      "step": 4730
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.855260604968083e-07,
      "loss": 0.4522,
      "step": 4731
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.834646773481811e-07,
      "loss": 0.345,
      "step": 4732
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.814068201738587e-07,
      "loss": 0.5418,
      "step": 4733
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.793524897443947e-07,
      "loss": 0.412,
      "step": 4734
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.773016868290226e-07,
      "loss": 0.427,
      "step": 4735
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.752544121956549e-07,
      "loss": 0.3451,
      "step": 4736
    },
    {
      "epoch": 0.89,
      "learning_rate": 5.732106666108827e-07,
      "loss": 0.518,
      "step": 4737
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.711704508399806e-07,
      "loss": 0.5816,
      "step": 4738
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.69133765646891e-07,
      "loss": 0.7014,
      "step": 4739
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.671006117942435e-07,
      "loss": 0.3864,
      "step": 4740
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.650709900433427e-07,
      "loss": 0.5267,
      "step": 4741
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.630449011541716e-07,
      "loss": 0.5206,
      "step": 4742
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.610223458853836e-07,
      "loss": 0.4754,
      "step": 4743
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.590033249943194e-07,
      "loss": 0.416,
      "step": 4744
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.56987839236991e-07,
      "loss": 0.5505,
      "step": 4745
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.549758893680834e-07,
      "loss": 0.4715,
      "step": 4746
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.529674761409643e-07,
      "loss": 0.4859,
      "step": 4747
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.509626003076718e-07,
      "loss": 0.5902,
      "step": 4748
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.489612626189245e-07,
      "loss": 0.527,
      "step": 4749
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.469634638241106e-07,
      "loss": 0.4438,
      "step": 4750
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.449692046712951e-07,
      "loss": 0.5534,
      "step": 4751
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.429784859072229e-07,
      "loss": 0.4029,
      "step": 4752
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.409913082773032e-07,
      "loss": 0.3151,
      "step": 4753
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.390076725256265e-07,
      "loss": 0.5193,
      "step": 4754
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.370275793949586e-07,
      "loss": 0.3671,
      "step": 4755
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.35051029626732e-07,
      "loss": 0.3798,
      "step": 4756
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.330780239610534e-07,
      "loss": 0.3386,
      "step": 4757
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.311085631367108e-07,
      "loss": 0.3674,
      "step": 4758
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.291426478911565e-07,
      "loss": 0.2639,
      "step": 4759
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.27180278960514e-07,
      "loss": 0.4063,
      "step": 4760
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.252214570795877e-07,
      "loss": 0.5063,
      "step": 4761
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.232661829818464e-07,
      "loss": 0.4754,
      "step": 4762
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.213144573994322e-07,
      "loss": 0.4331,
      "step": 4763
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.193662810631595e-07,
      "loss": 0.6323,
      "step": 4764
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.174216547025124e-07,
      "loss": 0.6156,
      "step": 4765
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.154805790456486e-07,
      "loss": 0.4101,
      "step": 4766
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.135430548193909e-07,
      "loss": 0.5782,
      "step": 4767
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.116090827492381e-07,
      "loss": 0.5882,
      "step": 4768
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.096786635593565e-07,
      "loss": 0.406,
      "step": 4769
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.077517979725788e-07,
      "loss": 0.5302,
      "step": 4770
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.058284867104135e-07,
      "loss": 0.4328,
      "step": 4771
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.039087304930323e-07,
      "loss": 0.4481,
      "step": 4772
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.019925300392814e-07,
      "loss": 0.4652,
      "step": 4773
    },
    {
      "epoch": 0.9,
      "learning_rate": 5.000798860666689e-07,
      "loss": 0.533,
      "step": 4774
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.981707992913753e-07,
      "loss": 0.3416,
      "step": 4775
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.962652704282533e-07,
      "loss": 0.4372,
      "step": 4776
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.943633001908111e-07,
      "loss": 0.5058,
      "step": 4777
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.924648892912365e-07,
      "loss": 0.4588,
      "step": 4778
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.9057003844038e-07,
      "loss": 0.4679,
      "step": 4779
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.886787483477573e-07,
      "loss": 0.5046,
      "step": 4780
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.86791019721553e-07,
      "loss": 0.4106,
      "step": 4781
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.849068532686185e-07,
      "loss": 0.361,
      "step": 4782
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.830262496944693e-07,
      "loss": 0.3534,
      "step": 4783
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.811492097032888e-07,
      "loss": 0.5078,
      "step": 4784
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.792757339979248e-07,
      "loss": 0.4319,
      "step": 4785
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.774058232798906e-07,
      "loss": 0.5215,
      "step": 4786
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.7553947824936496e-07,
      "loss": 0.4092,
      "step": 4787
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.7367669960519115e-07,
      "loss": 0.3416,
      "step": 4788
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.718174880448767e-07,
      "loss": 0.4719,
      "step": 4789
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.69961844264597e-07,
      "loss": 0.468,
      "step": 4790
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.6810976895918623e-07,
      "loss": 0.6524,
      "step": 4791
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.66261262822143e-07,
      "loss": 0.4645,
      "step": 4792
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.6441632654563473e-07,
      "loss": 0.5071,
      "step": 4793
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.625749608204877e-07,
      "loss": 0.2939,
      "step": 4794
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.6073716633618925e-07,
      "loss": 0.5187,
      "step": 4795
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.589029437808956e-07,
      "loss": 0.5587,
      "step": 4796
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5707229384142184e-07,
      "loss": 0.3857,
      "step": 4797
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5524521720324176e-07,
      "loss": 0.3955,
      "step": 4798
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5342171455049933e-07,
      "loss": 0.507,
      "step": 4799
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.5160178656599495e-07,
      "loss": 0.4756,
      "step": 4800
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4978543393119023e-07,
      "loss": 0.4073,
      "step": 4801
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.479726573262111e-07,
      "loss": 0.4203,
      "step": 4802
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4616345742984123e-07,
      "loss": 0.4066,
      "step": 4803
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4435783491952655e-07,
      "loss": 0.4113,
      "step": 4804
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.42555790471374e-07,
      "loss": 0.4009,
      "step": 4805
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.407573247601493e-07,
      "loss": 0.5109,
      "step": 4806
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.3896243845927943e-07,
      "loss": 0.4121,
      "step": 4807
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.371711322408501e-07,
      "loss": 0.5554,
      "step": 4808
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.353834067756069e-07,
      "loss": 0.5243,
      "step": 4809
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.335992627329555e-07,
      "loss": 0.4669,
      "step": 4810
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.3181870078096047e-07,
      "loss": 0.489,
      "step": 4811
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.300417215863406e-07,
      "loss": 0.4242,
      "step": 4812
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.282683258144793e-07,
      "loss": 0.4773,
      "step": 4813
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.2649851412941644e-07,
      "loss": 0.3248,
      "step": 4814
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.2473228719384863e-07,
      "loss": 0.3414,
      "step": 4815
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.229696456691301e-07,
      "loss": 0.5359,
      "step": 4816
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.21210590215273e-07,
      "loss": 0.4933,
      "step": 4817
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.1945512149094703e-07,
      "loss": 0.4705,
      "step": 4818
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.1770324015347865e-07,
      "loss": 0.5122,
      "step": 4819
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.1595494685885087e-07,
      "loss": 0.3634,
      "step": 4820
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.1421024226170335e-07,
      "loss": 0.4787,
      "step": 4821
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.1246912701533246e-07,
      "loss": 0.4548,
      "step": 4822
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.1073160177169e-07,
      "loss": 0.4518,
      "step": 4823
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.089976671813811e-07,
      "loss": 0.4998,
      "step": 4824
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.072673238936742e-07,
      "loss": 0.4892,
      "step": 4825
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.0554057255648315e-07,
      "loss": 0.4068,
      "step": 4826
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.0381741381638085e-07,
      "loss": 0.4294,
      "step": 4827
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.020978483185989e-07,
      "loss": 0.3783,
      "step": 4828
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.003818767070189e-07,
      "loss": 0.3962,
      "step": 4829
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.9866949962417465e-07,
      "loss": 0.3698,
      "step": 4830
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.969607177112611e-07,
      "loss": 0.4191,
      "step": 4831
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.95255531608123e-07,
      "loss": 0.4231,
      "step": 4832
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.9355394195325414e-07,
      "loss": 0.5547,
      "step": 4833
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.918559493838114e-07,
      "loss": 0.4448,
      "step": 4834
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.901615545355963e-07,
      "loss": 0.4406,
      "step": 4835
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.88470758043068e-07,
      "loss": 0.4828,
      "step": 4836
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8678356053933666e-07,
      "loss": 0.4338,
      "step": 4837
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8509996265616466e-07,
      "loss": 0.4797,
      "step": 4838
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.834199650239667e-07,
      "loss": 0.3929,
      "step": 4839
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.817435682718096e-07,
      "loss": 0.3972,
      "step": 4840
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8007077302741e-07,
      "loss": 0.3205,
      "step": 4841
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.7840157991714256e-07,
      "loss": 0.4332,
      "step": 4842
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.7673598956602276e-07,
      "loss": 0.5134,
      "step": 4843
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.75074002597724e-07,
      "loss": 0.3759,
      "step": 4844
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.734156196345706e-07,
      "loss": 0.2775,
      "step": 4845
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.7176084129753596e-07,
      "loss": 0.5539,
      "step": 4846
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.7010966820623996e-07,
      "loss": 0.4004,
      "step": 4847
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.684621009789602e-07,
      "loss": 0.4751,
      "step": 4848
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.668181402326165e-07,
      "loss": 0.3445,
      "step": 4849
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.6517778658278546e-07,
      "loss": 0.5193,
      "step": 4850
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.635410406436857e-07,
      "loss": 0.4555,
      "step": 4851
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.619079030281902e-07,
      "loss": 0.4478,
      "step": 4852
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.6027837434781886e-07,
      "loss": 0.3711,
      "step": 4853
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.586524552127413e-07,
      "loss": 0.4055,
      "step": 4854
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.570301462317727e-07,
      "loss": 0.468,
      "step": 4855
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.554114480123805e-07,
      "loss": 0.5527,
      "step": 4856
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.5379636116067764e-07,
      "loss": 0.4733,
      "step": 4857
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.521848862814248e-07,
      "loss": 0.3929,
      "step": 4858
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.505770239780304e-07,
      "loss": 0.5241,
      "step": 4859
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.489727748525518e-07,
      "loss": 0.3505,
      "step": 4860
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.473721395056906e-07,
      "loss": 0.3301,
      "step": 4861
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.4577511853679637e-07,
      "loss": 0.4449,
      "step": 4862
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.4418171254386844e-07,
      "loss": 0.5318,
      "step": 4863
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.425919221235463e-07,
      "loss": 0.4173,
      "step": 4864
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.4100574787111927e-07,
      "loss": 0.4953,
      "step": 4865
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.394231903805245e-07,
      "loss": 0.4146,
      "step": 4866
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.378442502443424e-07,
      "loss": 0.41,
      "step": 4867
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3626892805379565e-07,
      "loss": 0.3671,
      "step": 4868
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3469722439876007e-07,
      "loss": 0.4766,
      "step": 4869
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3312913986775055e-07,
      "loss": 0.5288,
      "step": 4870
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3156467504792846e-07,
      "loss": 0.5301,
      "step": 4871
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.3000383052509853e-07,
      "loss": 0.3755,
      "step": 4872
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.284466068837133e-07,
      "loss": 0.3879,
      "step": 4873
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.2689300470686633e-07,
      "loss": 0.4814,
      "step": 4874
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.2534302457629674e-07,
      "loss": 0.5066,
      "step": 4875
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.2379666707238464e-07,
      "loss": 0.3163,
      "step": 4876
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.222539327741592e-07,
      "loss": 0.5546,
      "step": 4877
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.207148222592871e-07,
      "loss": 0.4886,
      "step": 4878
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.191793361040796e-07,
      "loss": 0.5329,
      "step": 4879
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.176474748834946e-07,
      "loss": 0.4125,
      "step": 4880
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.1611923917112987e-07,
      "loss": 0.5257,
      "step": 4881
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.1459462953922104e-07,
      "loss": 0.3612,
      "step": 4882
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.130736465586559e-07,
      "loss": 0.489,
      "step": 4883
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.1155629079895775e-07,
      "loss": 0.5645,
      "step": 4884
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.100425628282899e-07,
      "loss": 0.4006,
      "step": 4885
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.085324632134623e-07,
      "loss": 0.5257,
      "step": 4886
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.070259925199248e-07,
      "loss": 0.7462,
      "step": 4887
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.0552315131176625e-07,
      "loss": 0.3979,
      "step": 4888
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.040239401517187e-07,
      "loss": 0.3904,
      "step": 4889
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.0252835960115524e-07,
      "loss": 0.4184,
      "step": 4890
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.010364102200869e-07,
      "loss": 0.4603,
      "step": 4891
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.9954809256716677e-07,
      "loss": 0.3515,
      "step": 4892
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.9806340719968795e-07,
      "loss": 0.3631,
      "step": 4893
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.9658235467358356e-07,
      "loss": 0.3496,
      "step": 4894
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.951049355434277e-07,
      "loss": 0.342,
      "step": 4895
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.9363115036243116e-07,
      "loss": 0.4748,
      "step": 4896
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.921609996824437e-07,
      "loss": 0.4662,
      "step": 4897
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.906944840539605e-07,
      "loss": 0.4973,
      "step": 4898
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.892316040261067e-07,
      "loss": 0.3884,
      "step": 4899
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8777236014665086e-07,
      "loss": 0.4985,
      "step": 4900
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.863167529620015e-07,
      "loss": 0.4041,
      "step": 4901
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8486478301720246e-07,
      "loss": 0.3986,
      "step": 4902
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.834164508559367e-07,
      "loss": 0.4765,
      "step": 4903
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.819717570205238e-07,
      "loss": 0.3913,
      "step": 4904
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.8053070205192433e-07,
      "loss": 0.4888,
      "step": 4905
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.790932864897322e-07,
      "loss": 0.5525,
      "step": 4906
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7765951087218134e-07,
      "loss": 0.4809,
      "step": 4907
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.762293757361412e-07,
      "loss": 0.3934,
      "step": 4908
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.748028816171189e-07,
      "loss": 0.3164,
      "step": 4909
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.733800290492572e-07,
      "loss": 0.5103,
      "step": 4910
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7196081856533775e-07,
      "loss": 0.3697,
      "step": 4911
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7054525069677653e-07,
      "loss": 0.4568,
      "step": 4912
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6913332597362506e-07,
      "loss": 0.3642,
      "step": 4913
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.677250449245705e-07,
      "loss": 0.4797,
      "step": 4914
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.663204080769388e-07,
      "loss": 0.4867,
      "step": 4915
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6491941595668926e-07,
      "loss": 0.4492,
      "step": 4916
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6352206908841325e-07,
      "loss": 0.3513,
      "step": 4917
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6212836799534237e-07,
      "loss": 0.4956,
      "step": 4918
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.607383131993424e-07,
      "loss": 0.4402,
      "step": 4919
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5935190522090924e-07,
      "loss": 0.4343,
      "step": 4920
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.579691445791788e-07,
      "loss": 0.4526,
      "step": 4921
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.565900317919179e-07,
      "loss": 0.3998,
      "step": 4922
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5521456737552906e-07,
      "loss": 0.5346,
      "step": 4923
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5384275184504705e-07,
      "loss": 0.4271,
      "step": 4924
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.5247458571414083e-07,
      "loss": 0.4155,
      "step": 4925
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.511100694951152e-07,
      "loss": 0.4444,
      "step": 4926
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.497492036989058e-07,
      "loss": 0.3934,
      "step": 4927
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4839198883508055e-07,
      "loss": 0.4071,
      "step": 4928
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4703842541184407e-07,
      "loss": 0.4191,
      "step": 4929
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4568851393602877e-07,
      "loss": 0.502,
      "step": 4930
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.443422549131047e-07,
      "loss": 0.5576,
      "step": 4931
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4299964884716867e-07,
      "loss": 0.4374,
      "step": 4932
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.4166069624095736e-07,
      "loss": 0.5818,
      "step": 4933
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.403253975958297e-07,
      "loss": 0.4857,
      "step": 4934
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3899375341178476e-07,
      "loss": 0.402,
      "step": 4935
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3766576418745024e-07,
      "loss": 0.4552,
      "step": 4936
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3634143042008396e-07,
      "loss": 0.525,
      "step": 4937
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3502075260557588e-07,
      "loss": 0.3957,
      "step": 4938
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3370373123844715e-07,
      "loss": 0.4856,
      "step": 4939
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.323903668118499e-07,
      "loss": 0.3601,
      "step": 4940
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.3108065981756743e-07,
      "loss": 0.3488,
      "step": 4941
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2977461074601304e-07,
      "loss": 0.383,
      "step": 4942
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2847222008622993e-07,
      "loss": 0.416,
      "step": 4943
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2717348832589025e-07,
      "loss": 0.6218,
      "step": 4944
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2587841595129945e-07,
      "loss": 0.4051,
      "step": 4945
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.245870034473896e-07,
      "loss": 0.414,
      "step": 4946
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2329925129772613e-07,
      "loss": 0.5589,
      "step": 4947
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.2201515998449884e-07,
      "loss": 0.4458,
      "step": 4948
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.2073472998852874e-07,
      "loss": 0.3412,
      "step": 4949
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1945796178927003e-07,
      "loss": 0.4456,
      "step": 4950
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.181848558647981e-07,
      "loss": 0.4796,
      "step": 4951
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.169154126918216e-07,
      "loss": 0.4019,
      "step": 4952
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1564963274568028e-07,
      "loss": 0.4488,
      "step": 4953
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.143875165003373e-07,
      "loss": 0.3775,
      "step": 4954
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1312906442838344e-07,
      "loss": 0.4453,
      "step": 4955
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1187427700104292e-07,
      "loss": 0.5731,
      "step": 4956
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.1062315468816318e-07,
      "loss": 0.5096,
      "step": 4957
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0937569795822177e-07,
      "loss": 0.4922,
      "step": 4958
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0813190727832055e-07,
      "loss": 0.4735,
      "step": 4959
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0689178311419257e-07,
      "loss": 0.5767,
      "step": 4960
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0565532593019632e-07,
      "loss": 0.3884,
      "step": 4961
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0442253618931484e-07,
      "loss": 0.4009,
      "step": 4962
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.031934143531611e-07,
      "loss": 0.4019,
      "step": 4963
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0196796088197468e-07,
      "loss": 0.3898,
      "step": 4964
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.0074617623461968e-07,
      "loss": 0.5633,
      "step": 4965
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9952806086858567e-07,
      "loss": 0.3963,
      "step": 4966
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9831361523999227e-07,
      "loss": 0.5418,
      "step": 4967
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9710283980358237e-07,
      "loss": 0.4345,
      "step": 4968
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9589573501272218e-07,
      "loss": 0.4501,
      "step": 4969
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9469230131940907e-07,
      "loss": 0.2699,
      "step": 4970
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.934925391742626e-07,
      "loss": 0.3776,
      "step": 4971
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9229644902652333e-07,
      "loss": 0.5306,
      "step": 4972
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.911040313240664e-07,
      "loss": 0.3894,
      "step": 4973
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8991528651338465e-07,
      "loss": 0.5127,
      "step": 4974
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8873021503959644e-07,
      "loss": 0.5183,
      "step": 4975
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8754881734644792e-07,
      "loss": 0.4378,
      "step": 4976
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8637109387630637e-07,
      "loss": 0.3916,
      "step": 4977
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8519704507016455e-07,
      "loss": 0.4314,
      "step": 4978
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8402667136763864e-07,
      "loss": 0.3611,
      "step": 4979
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8285997320697024e-07,
      "loss": 0.5461,
      "step": 4980
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8169695102502217e-07,
      "loss": 0.4065,
      "step": 4981
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8053760525728493e-07,
      "loss": 0.5932,
      "step": 4982
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7938193633786682e-07,
      "loss": 0.3875,
      "step": 4983
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7822994469950282e-07,
      "loss": 0.4371,
      "step": 4984
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7708163077355234e-07,
      "loss": 0.4773,
      "step": 4985
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.759369949899936e-07,
      "loss": 0.4485,
      "step": 4986
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7479603777742937e-07,
      "loss": 0.3606,
      "step": 4987
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7365875956308787e-07,
      "loss": 0.4681,
      "step": 4988
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7252516077281623e-07,
      "loss": 0.3942,
      "step": 4989
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.713952418310827e-07,
      "loss": 0.408,
      "step": 4990
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.7026900316098217e-07,
      "loss": 0.2346,
      "step": 4991
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.691464451842284e-07,
      "loss": 0.4081,
      "step": 4992
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.680275683211574e-07,
      "loss": 0.4445,
      "step": 4993
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6691237299072738e-07,
      "loss": 0.5196,
      "step": 4994
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.658008596105176e-07,
      "loss": 0.4522,
      "step": 4995
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6469302859672743e-07,
      "loss": 0.3698,
      "step": 4996
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6358888036418053e-07,
      "loss": 0.3549,
      "step": 4997
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6248841532631731e-07,
      "loss": 0.4717,
      "step": 4998
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6139163389520375e-07,
      "loss": 0.4656,
      "step": 4999
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.602985364815235e-07,
      "loss": 0.4128,
      "step": 5000
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.5920912349457918e-07,
      "loss": 0.455,
      "step": 5001
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5812339534229893e-07,
      "loss": 0.537,
      "step": 5002
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5704135243122865e-07,
      "loss": 0.4549,
      "step": 5003
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.559629951665298e-07,
      "loss": 0.6598,
      "step": 5004
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5488832395199162e-07,
      "loss": 0.4226,
      "step": 5005
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5381733919001884e-07,
      "loss": 0.4363,
      "step": 5006
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5275004128163407e-07,
      "loss": 0.4881,
      "step": 5007
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5168643062648425e-07,
      "loss": 0.3358,
      "step": 5008
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5062650762283193e-07,
      "loss": 0.4808,
      "step": 5009
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4957027266755963e-07,
      "loss": 0.4883,
      "step": 5010
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4851772615616878e-07,
      "loss": 0.3586,
      "step": 5011
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4746886848278187e-07,
      "loss": 0.3737,
      "step": 5012
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4642370004013696e-07,
      "loss": 0.5082,
      "step": 5013
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4538222121959322e-07,
      "loss": 0.3786,
      "step": 5014
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4434443241112538e-07,
      "loss": 0.5012,
      "step": 5015
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.433103340033304e-07,
      "loss": 0.453,
      "step": 5016
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.422799263834196e-07,
      "loss": 0.372,
      "step": 5017
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4125320993722435e-07,
      "loss": 0.4712,
      "step": 5018
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.4023018504919384e-07,
      "loss": 0.3681,
      "step": 5019
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3921085210239494e-07,
      "loss": 0.4414,
      "step": 5020
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3819521147851122e-07,
      "loss": 0.4931,
      "step": 5021
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3718326355784295e-07,
      "loss": 0.5331,
      "step": 5022
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3617500871931143e-07,
      "loss": 0.4664,
      "step": 5023
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3517044734045136e-07,
      "loss": 0.4558,
      "step": 5024
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3416957979741407e-07,
      "loss": 0.4248,
      "step": 5025
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3317240646497197e-07,
      "loss": 0.548,
      "step": 5026
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3217892771651087e-07,
      "loss": 0.5644,
      "step": 5027
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3118914392403426e-07,
      "loss": 0.4784,
      "step": 5028
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.3020305545816015e-07,
      "loss": 0.4526,
      "step": 5029
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.292206626881254e-07,
      "loss": 0.4054,
      "step": 5030
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.282419659817824e-07,
      "loss": 0.4217,
      "step": 5031
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2726696570559915e-07,
      "loss": 0.5513,
      "step": 5032
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2629566222465805e-07,
      "loss": 0.4128,
      "step": 5033
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.253280559026615e-07,
      "loss": 0.3836,
      "step": 5034
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2436414710192302e-07,
      "loss": 0.4227,
      "step": 5035
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2340393618337167e-07,
      "loss": 0.5342,
      "step": 5036
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.224474235065587e-07,
      "loss": 0.3971,
      "step": 5037
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2149460942964097e-07,
      "loss": 0.4964,
      "step": 5038
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.2054549430939532e-07,
      "loss": 0.4371,
      "step": 5039
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1960007850121523e-07,
      "loss": 0.4003,
      "step": 5040
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1865836235910755e-07,
      "loss": 0.4778,
      "step": 5041
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1772034623568907e-07,
      "loss": 0.451,
      "step": 5042
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1678603048219773e-07,
      "loss": 0.4424,
      "step": 5043
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1585541544848478e-07,
      "loss": 0.5064,
      "step": 5044
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1492850148301148e-07,
      "loss": 0.4037,
      "step": 5045
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1400528893285801e-07,
      "loss": 0.4784,
      "step": 5046
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1308577814371669e-07,
      "loss": 0.3321,
      "step": 5047
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1216996945989212e-07,
      "loss": 0.4276,
      "step": 5048
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1125786322430554e-07,
      "loss": 0.3975,
      "step": 5049
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1034945977849043e-07,
      "loss": 0.4559,
      "step": 5050
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0944475946259358e-07,
      "loss": 0.4288,
      "step": 5051
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0854376261537513e-07,
      "loss": 0.4418,
      "step": 5052
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0764646957420966e-07,
      "loss": 0.4648,
      "step": 5053
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0675288067508394e-07,
      "loss": 0.4224,
      "step": 5054
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0586299625259699e-07,
      "loss": 0.4774,
      "step": 5055
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.049768166399634e-07,
      "loss": 0.4133,
      "step": 5056
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.040943421690055e-07,
      "loss": 0.3888,
      "step": 5057
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.032155731701645e-07,
      "loss": 0.4277,
      "step": 5058
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0234050997248945e-07,
      "loss": 0.3141,
      "step": 5059
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.014691529036449e-07,
      "loss": 0.6237,
      "step": 5060
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.0060150228990429e-07,
      "loss": 0.5081,
      "step": 5061
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.973755845615551e-08,
      "loss": 0.5245,
      "step": 5062
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.887732172589759e-08,
      "loss": 0.4649,
      "step": 5063
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.802079242124396e-08,
      "loss": 0.3948,
      "step": 5064
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.716797086291475e-08,
      "loss": 0.3736,
      "step": 5065
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.631885737024782e-08,
      "loss": 0.4515,
      "step": 5066
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.547345226118666e-08,
      "loss": 0.5459,
      "step": 5067
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.463175585228913e-08,
      "loss": 0.3576,
      "step": 5068
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.379376845872756e-08,
      "loss": 0.4435,
      "step": 5069
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.295949039427876e-08,
      "loss": 0.4364,
      "step": 5070
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.212892197133394e-08,
      "loss": 0.4831,
      "step": 5071
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.130206350089765e-08,
      "loss": 0.4581,
      "step": 5072
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.047891529258112e-08,
      "loss": 0.5572,
      "step": 5073
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.96594776546078e-08,
      "loss": 0.4738,
      "step": 5074
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.884375089381336e-08,
      "loss": 0.3899,
      "step": 5075
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.803173531564235e-08,
      "loss": 0.512,
      "step": 5076
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.722343122414823e-08,
      "loss": 0.5958,
      "step": 5077
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.641883892199775e-08,
      "loss": 0.3792,
      "step": 5078
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.561795871046774e-08,
      "loss": 0.4832,
      "step": 5079
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.482079088944051e-08,
      "loss": 0.4642,
      "step": 5080
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.402733575741506e-08,
      "loss": 0.4044,
      "step": 5081
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.323759361149374e-08,
      "loss": 0.426,
      "step": 5082
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.245156474739335e-08,
      "loss": 0.5061,
      "step": 5083
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.166924945943733e-08,
      "loss": 0.4776,
      "step": 5084
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.089064804056135e-08,
      "loss": 0.4444,
      "step": 5085
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.011576078230665e-08,
      "loss": 0.501,
      "step": 5086
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.93445879748267e-08,
      "loss": 0.4391,
      "step": 5087
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.857712990688492e-08,
      "loss": 0.4102,
      "step": 5088
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.781338686584928e-08,
      "loss": 0.3935,
      "step": 5089
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.705335913770207e-08,
      "loss": 0.4654,
      "step": 5090
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.629704700702901e-08,
      "loss": 0.4139,
      "step": 5091
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.554445075703021e-08,
      "loss": 0.3171,
      "step": 5092
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.479557066951138e-08,
      "loss": 0.4253,
      "step": 5093
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.405040702488486e-08,
      "loss": 0.4391,
      "step": 5094
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.330896010217414e-08,
      "loss": 0.5141,
      "step": 5095
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.257123017901158e-08,
      "loss": 0.4487,
      "step": 5096
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.183721753163508e-08,
      "loss": 0.4113,
      "step": 5097
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.110692243489259e-08,
      "loss": 0.4651,
      "step": 5098
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.038034516223869e-08,
      "loss": 0.4359,
      "step": 5099
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.965748598573797e-08,
      "loss": 0.4132,
      "step": 5100
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.893834517606057e-08,
      "loss": 0.4952,
      "step": 5101
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.822292300248445e-08,
      "loss": 0.4818,
      "step": 5102
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.751121973289754e-08,
      "loss": 0.62,
      "step": 5103
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.68032356337922e-08,
      "loss": 0.489,
      "step": 5104
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.609897097026974e-08,
      "loss": 0.6902,
      "step": 5105
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.539842600603918e-08,
      "loss": 0.3033,
      "step": 5106
    },
    {
      "epoch": 0.96,
      "learning_rate": 6.470160100341516e-08,
      "loss": 0.3292,
      "step": 5107
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.400849622332118e-08,
      "loss": 0.3272,
      "step": 5108
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.331911192528516e-08,
      "loss": 0.5088,
      "step": 5109
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.263344836744623e-08,
      "loss": 0.3662,
      "step": 5110
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.195150580654563e-08,
      "loss": 0.4956,
      "step": 5111
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.127328449793357e-08,
      "loss": 0.3773,
      "step": 5112
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.059878469556801e-08,
      "loss": 0.4356,
      "step": 5113
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.992800665201137e-08,
      "loss": 0.3229,
      "step": 5114
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.9260950618430515e-08,
      "loss": 0.3992,
      "step": 5115
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.859761684460563e-08,
      "loss": 0.3662,
      "step": 5116
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.793800557891471e-08,
      "loss": 0.4117,
      "step": 5117
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.728211706834908e-08,
      "loss": 0.5436,
      "step": 5118
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.662995155850115e-08,
      "loss": 0.367,
      "step": 5119
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.5981509293571154e-08,
      "loss": 0.2543,
      "step": 5120
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.533679051636487e-08,
      "loss": 0.4326,
      "step": 5121
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.469579546829473e-08,
      "loss": 0.3761,
      "step": 5122
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.405852438937764e-08,
      "loss": 0.4994,
      "step": 5123
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.3424977518237163e-08,
      "loss": 0.3981,
      "step": 5124
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.279515509210131e-08,
      "loss": 0.4525,
      "step": 5125
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.2169057346802556e-08,
      "loss": 0.6014,
      "step": 5126
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.154668451678224e-08,
      "loss": 0.4561,
      "step": 5127
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.092803683508396e-08,
      "loss": 0.4997,
      "step": 5128
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.031311453335685e-08,
      "loss": 0.3828,
      "step": 5129
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.970191784185563e-08,
      "loss": 0.5553,
      "step": 5130
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.9094446989440546e-08,
      "loss": 0.4458,
      "step": 5131
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.8490702203575215e-08,
      "loss": 0.3551,
      "step": 5132
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.789068371032879e-08,
      "loss": 0.5681,
      "step": 5133
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.7294391734376e-08,
      "loss": 0.3037,
      "step": 5134
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.6701826498993794e-08,
      "loss": 0.4467,
      "step": 5135
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.611298822606691e-08,
      "loss": 0.5709,
      "step": 5136
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.552787713608231e-08,
      "loss": 0.4378,
      "step": 5137
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.49464934481314e-08,
      "loss": 0.4745,
      "step": 5138
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.436883737991005e-08,
      "loss": 0.447,
      "step": 5139
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3794909147720776e-08,
      "loss": 0.4863,
      "step": 5140
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.322470896646502e-08,
      "loss": 0.3981,
      "step": 5141
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.2658237049655325e-08,
      "loss": 0.5081,
      "step": 5142
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.209549360940091e-08,
      "loss": 0.5154,
      "step": 5143
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.153647885641876e-08,
      "loss": 0.3915,
      "step": 5144
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.098119300003034e-08,
      "loss": 0.4725,
      "step": 5145
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.042963624815932e-08,
      "loss": 0.4194,
      "step": 5146
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.988180880733161e-08,
      "loss": 0.4245,
      "step": 5147
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.9337710882682014e-08,
      "loss": 0.4158,
      "step": 5148
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.879734267794199e-08,
      "loss": 0.4197,
      "step": 5149
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.8260704395450824e-08,
      "loss": 0.4438,
      "step": 5150
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.772779623614997e-08,
      "loss": 0.5332,
      "step": 5151
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.719861839958428e-08,
      "loss": 0.5063,
      "step": 5152
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.6673171083901894e-08,
      "loss": 0.3063,
      "step": 5153
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.6151454485853225e-08,
      "loss": 0.448,
      "step": 5154
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.5633468800793103e-08,
      "loss": 0.4914,
      "step": 5155
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.511921422267972e-08,
      "loss": 0.4349,
      "step": 5156
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.460869094407127e-08,
      "loss": 0.3723,
      "step": 5157
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.410189915613038e-08,
      "loss": 0.4262,
      "step": 5158
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.359883904862415e-08,
      "loss": 0.3784,
      "step": 5159
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.3099510809920795e-08,
      "loss": 0.415,
      "step": 5160
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.260391462698964e-08,
      "loss": 0.5329,
      "step": 5161
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.211205068540668e-08,
      "loss": 0.3383,
      "step": 5162
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.162391916934682e-08,
      "loss": 0.4294,
      "step": 5163
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.113952026158607e-08,
      "loss": 0.3861,
      "step": 5164
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.065885414350933e-08,
      "loss": 0.4187,
      "step": 5165
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.018192099509709e-08,
      "loss": 0.3642,
      "step": 5166
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.9708720994934272e-08,
      "loss": 0.5562,
      "step": 5167
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.923925432021024e-08,
      "loss": 0.5825,
      "step": 5168
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.8773521146712168e-08,
      "loss": 0.436,
      "step": 5169
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.8311521648832775e-08,
      "loss": 0.403,
      "step": 5170
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.7853255999564787e-08,
      "loss": 0.4805,
      "step": 5171
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.739872437050539e-08,
      "loss": 0.3858,
      "step": 5172
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.6947926931848444e-08,
      "loss": 0.5647,
      "step": 5173
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.6500863852395585e-08,
      "loss": 0.5556,
      "step": 5174
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.6057535299546245e-08,
      "loss": 0.3799,
      "step": 5175
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5617941439302073e-08,
      "loss": 0.4397,
      "step": 5176
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5182082436266963e-08,
      "loss": 0.5317,
      "step": 5177
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.4749958453647027e-08,
      "loss": 0.4814,
      "step": 5178
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.4321569653248388e-08,
      "loss": 0.4527,
      "step": 5179
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.3896916195479404e-08,
      "loss": 0.3527,
      "step": 5180
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.347599823934843e-08,
      "loss": 0.5549,
      "step": 5181
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.3058815942467172e-08,
      "loss": 0.3972,
      "step": 5182
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.264536946104623e-08,
      "loss": 0.3983,
      "step": 5183
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.223565894990065e-08,
      "loss": 0.4789,
      "step": 5184
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.1829684562442165e-08,
      "loss": 0.4785,
      "step": 5185
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.1427446450686952e-08,
      "loss": 0.3398,
      "step": 5186
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.1028944765251193e-08,
      "loss": 0.3493,
      "step": 5187
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.0634179655351084e-08,
      "loss": 0.4578,
      "step": 5188
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.024315126880505e-08,
      "loss": 0.5201,
      "step": 5189
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9855859752031526e-08,
      "loss": 0.4861,
      "step": 5190
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.947230525005006e-08,
      "loss": 0.3994,
      "step": 5191
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9092487906480217e-08,
      "loss": 0.6497,
      "step": 5192
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8716407863543784e-08,
      "loss": 0.3898,
      "step": 5193
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.834406526206034e-08,
      "loss": 0.4595,
      "step": 5194
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.7975460241453913e-08,
      "loss": 0.4548,
      "step": 5195
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.76105929397441e-08,
      "loss": 0.5892,
      "step": 5196
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.724946349355605e-08,
      "loss": 0.4218,
      "step": 5197
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.6892072038111605e-08,
      "loss": 0.3186,
      "step": 5198
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.6538418707233716e-08,
      "loss": 0.4447,
      "step": 5199
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.6188503633347562e-08,
      "loss": 0.3255,
      "step": 5200
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5842326947477226e-08,
      "loss": 0.5474,
      "step": 5201
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5499888779244577e-08,
      "loss": 0.4452,
      "step": 5202
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.516118925687593e-08,
      "loss": 0.339,
      "step": 5203
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.4826228507195394e-08,
      "loss": 0.4639,
      "step": 5204
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.4495006655625976e-08,
      "loss": 0.3752,
      "step": 5205
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.4167523826194018e-08,
      "loss": 0.5313,
      "step": 5206
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.3843780141521435e-08,
      "loss": 0.4551,
      "step": 5207
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.3523775722834586e-08,
      "loss": 0.5271,
      "step": 5208
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.3207510689956515e-08,
      "loss": 0.4961,
      "step": 5209
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2894985161311379e-08,
      "loss": 0.521,
      "step": 5210
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2586199253922238e-08,
      "loss": 0.4206,
      "step": 5211
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.2281153083413267e-08,
      "loss": 0.5598,
      "step": 5212
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.1979846764006431e-08,
      "loss": 0.402,
      "step": 5213
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1682280408525926e-08,
      "loss": 0.4417,
      "step": 5214
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1388454128392624e-08,
      "loss": 0.6069,
      "step": 5215
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1098368033628514e-08,
      "loss": 0.4692,
      "step": 5216
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.081202223285449e-08,
      "loss": 0.4391,
      "step": 5217
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.052941683329256e-08,
      "loss": 0.5085,
      "step": 5218
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.025055194076252e-08,
      "loss": 0.5021,
      "step": 5219
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.975427659683069e-09,
      "loss": 0.5962,
      "step": 5220
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.704044093072906e-09,
      "loss": 0.5479,
      "step": 5221
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.43640134255186e-09,
      "loss": 0.4335,
      "step": 5222
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.17249950833754e-09,
      "loss": 0.4472,
      "step": 5223
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.912338689244238e-09,
      "loss": 0.4401,
      "step": 5224
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.655918982689582e-09,
      "loss": 0.4624,
      "step": 5225
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.40324048468788e-09,
      "loss": 0.3957,
      "step": 5226
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.154303289854559e-09,
      "loss": 0.3794,
      "step": 5227
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.909107491401723e-09,
      "loss": 0.3555,
      "step": 5228
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.667653181142598e-09,
      "loss": 0.4709,
      "step": 5229
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.429940449488193e-09,
      "loss": 0.4103,
      "step": 5230
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.19596938544731e-09,
      "loss": 0.4513,
      "step": 5231
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.965740076632088e-09,
      "loss": 0.5765,
      "step": 5232
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.739252609249125e-09,
      "loss": 0.37,
      "step": 5233
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.516507068105027e-09,
      "loss": 0.4782,
      "step": 5234
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.297503536606408e-09,
      "loss": 0.4788,
      "step": 5235
    },
    {
      "epoch": 0.99,
      "learning_rate": 6.0822420967587835e-09,
      "loss": 0.3519,
      "step": 5236
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.870722829164344e-09,
      "loss": 0.5393,
      "step": 5237
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.662945813025289e-09,
      "loss": 0.5254,
      "step": 5238
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.458911126144939e-09,
      "loss": 0.4272,
      "step": 5239
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.258618844919961e-09,
      "loss": 0.4159,
      "step": 5240
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.0620690443514696e-09,
      "loss": 0.4348,
      "step": 5241
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.869261798035041e-09,
      "loss": 0.4283,
      "step": 5242
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.680197178167367e-09,
      "loss": 0.5041,
      "step": 5243
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.4948752555418196e-09,
      "loss": 0.4753,
      "step": 5244
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.313296099550668e-09,
      "loss": 0.4027,
      "step": 5245
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.135459778187301e-09,
      "loss": 0.4504,
      "step": 5246
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.9613663580406745e-09,
      "loss": 0.4456,
      "step": 5247
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.791015904298645e-09,
      "loss": 0.4626,
      "step": 5248
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.624408480747965e-09,
      "loss": 0.3714,
      "step": 5249
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.461544149775398e-09,
      "loss": 0.3102,
      "step": 5250
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.302422972363273e-09,
      "loss": 0.4041,
      "step": 5251
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.14704500809504e-09,
      "loss": 0.4435,
      "step": 5252
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.995410315149716e-09,
      "loss": 0.4388,
      "step": 5253
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.847518950306327e-09,
      "loss": 0.5492,
      "step": 5254
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.7033709689439082e-09,
      "loss": 0.4913,
      "step": 5255
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.562966425034841e-09,
      "loss": 0.4523,
      "step": 5256
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.426305371155957e-09,
      "loss": 0.5,
      "step": 5257
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.293387858477436e-09,
      "loss": 0.414,
      "step": 5258
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.164213936770576e-09,
      "loss": 0.7016,
      "step": 5259
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.0387836544033535e-09,
      "loss": 0.4377,
      "step": 5260
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9170970583426428e-09,
      "loss": 0.5295,
      "step": 5261
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.7991541941531076e-09,
      "loss": 0.5015,
      "step": 5262
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.684955105997199e-09,
      "loss": 0.4784,
      "step": 5263
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.574499836638488e-09,
      "loss": 0.4029,
      "step": 5264
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.4677884274338915e-09,
      "loss": 0.4796,
      "step": 5265
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.3648209183425575e-09,
      "loss": 0.4857,
      "step": 5266
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.265597347920311e-09,
      "loss": 0.38,
      "step": 5267
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.170117753319655e-09,
      "loss": 0.4451,
      "step": 5268
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0783821702931018e-09,
      "loss": 0.4205,
      "step": 5269
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.903906331909518e-10,
      "loss": 0.2654,
      "step": 5270
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.061431749601835e-10,
      "loss": 0.4334,
      "step": 5271
    },
    {
      "epoch": 1.0,
      "learning_rate": 8.256398271466737e-10,
      "loss": 0.4729,
      "step": 5272
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.488806198963083e-10,
      "loss": 0.4737,
      "step": 5273
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.758655819505411e-10,
      "loss": 0.4109,
      "step": 5274
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.065947406475036e-10,
      "loss": 0.4104,
      "step": 5275
    },
    {
      "epoch": 1.0,
      "learning_rate": 5.410681219286673e-10,
      "loss": 0.396,
      "step": 5276
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.792857503266301e-10,
      "loss": 0.46,
      "step": 5277
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.2124764897844005e-10,
      "loss": 0.4403,
      "step": 5278
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.6695383961449227e-10,
      "loss": 0.3591,
      "step": 5279
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.164043425651908e-10,
      "loss": 0.3992,
      "step": 5280
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.69599176758728e-10,
      "loss": 0.3438,
      "step": 5281
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2653835972108463e-10,
      "loss": 0.3423,
      "step": 5282
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8722190757602955e-10,
      "loss": 0.4566,
      "step": 5283
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.516498350451201e-10,
      "loss": 0.4102,
      "step": 5284
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1982215544881214e-10,
      "loss": 0.3734,
      "step": 5285
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.173888070423964e-11,
      "loss": 0.4433,
      "step": 5286
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.740002132743506e-11,
      "loss": 0.4371,
      "step": 5287
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.680558643221922e-11,
      "loss": 0.4685,
      "step": 5288
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.995558372909102e-11,
      "loss": 0.3886,
      "step": 5289
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.6850019528558136e-11,
      "loss": 0.5164,
      "step": 5290
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.488898736696116e-12,
      "loss": 0.4592,
      "step": 5291
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8722248595892668e-12,
      "loss": 0.4171,
      "step": 5292
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0,
      "loss": 0.52,
      "step": 5293
    },
    {
      "epoch": 1.0,
      "step": 5293,
      "total_flos": 241756536569856.0,
      "train_loss": 0.5514708950724287,
      "train_runtime": 24636.6893,
      "train_samples_per_second": 13.749,
      "train_steps_per_second": 0.215
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 5293,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 50000,
  "total_flos": 241756536569856.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}