{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.999069623197395,
  "global_step": 2418,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 2.061855670103093e-07,
      "loss": 2.6931,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.123711340206186e-07,
      "loss": 2.6928,
      "step": 2
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.185567010309279e-07,
      "loss": 2.3858,
      "step": 3
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.247422680412372e-07,
      "loss": 2.2344,
      "step": 4
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.0309278350515464e-06,
      "loss": 2.2382,
      "step": 5
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2371134020618557e-06,
      "loss": 2.2291,
      "step": 6
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.4432989690721649e-06,
      "loss": 2.0878,
      "step": 7
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.6494845360824744e-06,
      "loss": 2.1646,
      "step": 8
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.8556701030927837e-06,
      "loss": 2.0889,
      "step": 9
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.061855670103093e-06,
      "loss": 2.0632,
      "step": 10
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.268041237113402e-06,
      "loss": 2.0567,
      "step": 11
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.4742268041237115e-06,
      "loss": 2.0144,
      "step": 12
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.680412371134021e-06,
      "loss": 1.9909,
      "step": 13
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.8865979381443297e-06,
      "loss": 1.959,
      "step": 14
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.0927835051546395e-06,
      "loss": 1.961,
      "step": 15
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.298969072164949e-06,
      "loss": 1.9459,
      "step": 16
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.5051546391752577e-06,
      "loss": 1.9217,
      "step": 17
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.7113402061855674e-06,
      "loss": 1.897,
      "step": 18
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.917525773195877e-06,
      "loss": 1.8681,
      "step": 19
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.123711340206186e-06,
      "loss": 1.848,
      "step": 20
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.329896907216495e-06,
      "loss": 1.8664,
      "step": 21
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.536082474226804e-06,
      "loss": 1.8656,
      "step": 22
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.742268041237113e-06,
      "loss": 1.8798,
      "step": 23
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.948453608247423e-06,
      "loss": 1.8239,
      "step": 24
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.154639175257732e-06,
      "loss": 1.8775,
      "step": 25
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.360824742268042e-06,
      "loss": 1.8492,
      "step": 26
    },
    {
      "epoch": 0.07,
      "learning_rate": 5.567010309278351e-06,
      "loss": 1.8243,
      "step": 27
    },
    {
      "epoch": 0.07,
      "learning_rate": 5.7731958762886594e-06,
      "loss": 1.7578,
      "step": 28
    },
    {
      "epoch": 0.07,
      "learning_rate": 5.979381443298969e-06,
      "loss": 1.8082,
      "step": 29
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.185567010309279e-06,
      "loss": 1.8139,
      "step": 30
    },
    {
      "epoch": 0.08,
      "learning_rate": 6.391752577319588e-06,
      "loss": 1.7897,
      "step": 31
    },
    {
      "epoch": 0.08,
      "learning_rate": 6.597938144329898e-06,
      "loss": 1.8071,
      "step": 32
    },
    {
      "epoch": 0.08,
      "learning_rate": 6.804123711340207e-06,
      "loss": 1.815,
      "step": 33
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.010309278350515e-06,
      "loss": 1.8062,
      "step": 34
    },
    {
      "epoch": 0.09,
      "learning_rate": 7.216494845360825e-06,
      "loss": 1.7057,
      "step": 35
    },
    {
      "epoch": 0.09,
      "learning_rate": 7.422680412371135e-06,
      "loss": 1.8079,
      "step": 36
    },
    {
      "epoch": 0.09,
      "learning_rate": 7.628865979381444e-06,
      "loss": 1.8125,
      "step": 37
    },
    {
      "epoch": 0.09,
      "learning_rate": 7.835051546391754e-06,
      "loss": 1.79,
      "step": 38
    },
    {
      "epoch": 0.1,
      "learning_rate": 8.041237113402063e-06,
      "loss": 1.7811,
      "step": 39
    },
    {
      "epoch": 0.1,
      "learning_rate": 8.247422680412371e-06,
      "loss": 1.7748,
      "step": 40
    },
    {
      "epoch": 0.1,
      "learning_rate": 8.453608247422681e-06,
      "loss": 1.8435,
      "step": 41
    },
    {
      "epoch": 0.1,
      "learning_rate": 8.65979381443299e-06,
      "loss": 1.8284,
      "step": 42
    },
    {
      "epoch": 0.11,
      "learning_rate": 8.865979381443299e-06,
      "loss": 1.79,
      "step": 43
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.072164948453609e-06,
      "loss": 1.7456,
      "step": 44
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.278350515463918e-06,
      "loss": 1.7697,
      "step": 45
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.484536082474226e-06,
      "loss": 1.7731,
      "step": 46
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.690721649484536e-06,
      "loss": 1.7674,
      "step": 47
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.896907216494846e-06,
      "loss": 1.7783,
      "step": 48
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.0103092783505156e-05,
      "loss": 1.8014,
      "step": 49
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.0309278350515464e-05,
      "loss": 1.7922,
      "step": 50
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.0515463917525775e-05,
      "loss": 1.7722,
      "step": 51
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.0721649484536083e-05,
      "loss": 1.7526,
      "step": 52
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.0927835051546391e-05,
      "loss": 1.7501,
      "step": 53
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.1134020618556703e-05,
      "loss": 1.7925,
      "step": 54
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.134020618556701e-05,
      "loss": 1.753,
      "step": 55
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.1546391752577319e-05,
      "loss": 1.7688,
      "step": 56
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.175257731958763e-05,
      "loss": 1.7584,
      "step": 57
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.1958762886597938e-05,
      "loss": 1.7504,
      "step": 58
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.2164948453608248e-05,
      "loss": 1.789,
      "step": 59
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.2371134020618558e-05,
      "loss": 1.7612,
      "step": 60
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.2577319587628866e-05,
      "loss": 1.7735,
      "step": 61
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.2783505154639176e-05,
      "loss": 1.7335,
      "step": 62
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.2989690721649485e-05,
      "loss": 1.6972,
      "step": 63
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.3195876288659795e-05,
      "loss": 1.7145,
      "step": 64
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.3402061855670103e-05,
      "loss": 1.7023,
      "step": 65
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.3608247422680415e-05,
      "loss": 1.7368,
      "step": 66
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.3814432989690723e-05,
      "loss": 1.8007,
      "step": 67
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.402061855670103e-05,
      "loss": 1.7293,
      "step": 68
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.4226804123711342e-05,
      "loss": 1.8294,
      "step": 69
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.443298969072165e-05,
      "loss": 1.7392,
      "step": 70
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.4639175257731958e-05,
      "loss": 1.7989,
      "step": 71
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.484536082474227e-05,
      "loss": 1.7857,
      "step": 72
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.5051546391752578e-05,
      "loss": 1.7537,
      "step": 73
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.5257731958762888e-05,
      "loss": 1.7359,
      "step": 74
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.5463917525773197e-05,
      "loss": 1.7415,
      "step": 75
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.5670103092783507e-05,
      "loss": 1.7113,
      "step": 76
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.5876288659793813e-05,
      "loss": 1.7073,
      "step": 77
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.6082474226804127e-05,
      "loss": 1.7782,
      "step": 78
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.6288659793814433e-05,
      "loss": 1.7182,
      "step": 79
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.6494845360824743e-05,
      "loss": 1.7317,
      "step": 80
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.6701030927835052e-05,
      "loss": 1.7389,
      "step": 81
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.6907216494845362e-05,
      "loss": 1.8106,
      "step": 82
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.7113402061855672e-05,
      "loss": 1.7105,
      "step": 83
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.731958762886598e-05,
      "loss": 1.8041,
      "step": 84
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.752577319587629e-05,
      "loss": 1.7307,
      "step": 85
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.7731958762886598e-05,
      "loss": 1.7587,
      "step": 86
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.793814432989691e-05,
      "loss": 1.759,
      "step": 87
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8144329896907217e-05,
      "loss": 1.7615,
      "step": 88
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8350515463917527e-05,
      "loss": 1.6968,
      "step": 89
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.8556701030927837e-05,
      "loss": 1.7071,
      "step": 90
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8762886597938147e-05,
      "loss": 1.7431,
      "step": 91
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.8969072164948453e-05,
      "loss": 1.7429,
      "step": 92
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9175257731958766e-05,
      "loss": 1.7768,
      "step": 93
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9381443298969072e-05,
      "loss": 1.7858,
      "step": 94
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9587628865979382e-05,
      "loss": 1.7121,
      "step": 95
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9793814432989692e-05,
      "loss": 1.6771,
      "step": 96
    },
    {
      "epoch": 0.24,
      "learning_rate": 2e-05,
      "loss": 1.7491,
      "step": 97
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9999990839495287e-05,
      "loss": 1.7543,
      "step": 98
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.999996335799793e-05,
      "loss": 1.7928,
      "step": 99
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.999991755555828e-05,
      "loss": 1.737,
      "step": 100
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.999985343226025e-05,
      "loss": 1.705,
      "step": 101
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9999770988221318e-05,
      "loss": 1.7246,
      "step": 102
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.999967022359253e-05,
      "loss": 1.7468,
      "step": 103
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9999551138558503e-05,
      "loss": 1.7227,
      "step": 104
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9999413733337408e-05,
      "loss": 1.7342,
      "step": 105
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9999258008180985e-05,
      "loss": 1.707,
      "step": 106
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9999083963374542e-05,
      "loss": 1.7221,
      "step": 107
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.999889159923694e-05,
      "loss": 1.7603,
      "step": 108
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9998680916120614e-05,
      "loss": 1.7246,
      "step": 109
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.999845191441156e-05,
      "loss": 1.765,
      "step": 110
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9998204594529324e-05,
      "loss": 1.7194,
      "step": 111
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9997938956927027e-05,
      "loss": 1.7338,
      "step": 112
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9997655002091345e-05,
      "loss": 1.7232,
      "step": 113
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9997352730542506e-05,
      "loss": 1.7854,
      "step": 114
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9997032142834308e-05,
      "loss": 1.6751,
      "step": 115
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9996693239554096e-05,
      "loss": 1.7488,
      "step": 116
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9996336021322776e-05,
      "loss": 1.7532,
      "step": 117
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.999596048879481e-05,
      "loss": 1.7223,
      "step": 118
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9995566642658208e-05,
      "loss": 1.7123,
      "step": 119
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9995154483634534e-05,
      "loss": 1.749,
      "step": 120
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9994724012478915e-05,
      "loss": 1.7041,
      "step": 121
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.999427522998001e-05,
      "loss": 1.7112,
      "step": 122
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.999380813696003e-05,
      "loss": 1.7121,
      "step": 123
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9993322734274746e-05,
      "loss": 1.7078,
      "step": 124
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9992819022813458e-05,
      "loss": 1.7655,
      "step": 125
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.999229700349902e-05,
      "loss": 1.7325,
      "step": 126
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9991756677287823e-05,
      "loss": 1.7242,
      "step": 127
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9991198045169798e-05,
      "loss": 1.7089,
      "step": 128
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9990621108168417e-05,
      "loss": 1.7334,
      "step": 129
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9990025867340683e-05,
      "loss": 1.7983,
      "step": 130
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.998941232377714e-05,
      "loss": 1.7242,
      "step": 131
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9988780478601865e-05,
      "loss": 1.7401,
      "step": 132
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9988130332972458e-05,
      "loss": 1.7167,
      "step": 133
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9987461888080053e-05,
      "loss": 1.6843,
      "step": 134
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9986775145149305e-05,
      "loss": 1.7499,
      "step": 135
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9986070105438403e-05,
      "loss": 1.719,
      "step": 136
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.998534677023905e-05,
      "loss": 1.6586,
      "step": 137
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9984605140876462e-05,
      "loss": 1.7229,
      "step": 138
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9983845218709382e-05,
      "loss": 1.6969,
      "step": 139
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.998306700513007e-05,
      "loss": 1.694,
      "step": 140
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9982270501564286e-05,
      "loss": 1.6947,
      "step": 141
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9981455709471305e-05,
      "loss": 1.7095,
      "step": 142
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9980622630343908e-05,
      "loss": 1.6833,
      "step": 143
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9979771265708386e-05,
      "loss": 1.6749,
      "step": 144
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9978901617124514e-05,
      "loss": 1.7418,
      "step": 145
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9978013686185588e-05,
      "loss": 1.7273,
      "step": 146
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.997710747451838e-05,
      "loss": 1.7733,
      "step": 147
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9976182983783165e-05,
      "loss": 1.6835,
      "step": 148
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.99752402156737e-05,
      "loss": 1.6519,
      "step": 149
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.997427917191723e-05,
      "loss": 1.7668,
      "step": 150
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.997329985427449e-05,
      "loss": 1.7449,
      "step": 151
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9972302264539686e-05,
      "loss": 1.6546,
      "step": 152
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.99712864045405e-05,
      "loss": 1.7126,
      "step": 153
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9970252276138094e-05,
      "loss": 1.7632,
      "step": 154
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9969199881227094e-05,
      "loss": 1.715,
      "step": 155
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9968129221735593e-05,
      "loss": 1.6783,
      "step": 156
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9967040299625152e-05,
      "loss": 1.6405,
      "step": 157
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9965933116890775e-05,
      "loss": 1.6665,
      "step": 158
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9964807675560947e-05,
      "loss": 1.7266,
      "step": 159
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9963663977697583e-05,
      "loss": 1.6817,
      "step": 160
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.996250202539605e-05,
      "loss": 1.7325,
      "step": 161
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9961321820785167e-05,
      "loss": 1.7291,
      "step": 162
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9960123366027187e-05,
      "loss": 1.6795,
      "step": 163
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9958906663317795e-05,
      "loss": 1.6884,
      "step": 164
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.995767171488612e-05,
      "loss": 1.6681,
      "step": 165
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.995641852299471e-05,
      "loss": 1.6567,
      "step": 166
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9955147089939537e-05,
      "loss": 1.7134,
      "step": 167
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9953857418049995e-05,
      "loss": 1.6897,
      "step": 168
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9952549509688896e-05,
      "loss": 1.7775,
      "step": 169
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.995122336725246e-05,
      "loss": 1.6888,
      "step": 170
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9949878993170312e-05,
      "loss": 1.7392,
      "step": 171
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.994851638990548e-05,
      "loss": 1.6413,
      "step": 172
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9947135559954393e-05,
      "loss": 1.7008,
      "step": 173
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9945736505846866e-05,
      "loss": 1.71,
      "step": 174
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9944319230146116e-05,
      "loss": 1.6869,
      "step": 175
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.994288373544873e-05,
      "loss": 1.6317,
      "step": 176
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9941430024384676e-05,
      "loss": 1.691,
      "step": 177
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9939958099617306e-05,
      "loss": 1.7102,
      "step": 178
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.993846796384333e-05,
      "loss": 1.7316,
      "step": 179
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.993695961979283e-05,
      "loss": 1.6549,
      "step": 180
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9935433070229242e-05,
      "loss": 1.7298,
      "step": 181
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.993388831794936e-05,
      "loss": 1.6484,
      "step": 182
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.993232536578333e-05,
      "loss": 1.6755,
      "step": 183
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.993074421659463e-05,
      "loss": 1.6736,
      "step": 184
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9929144873280092e-05,
      "loss": 1.6565,
      "step": 185
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.992752733876987e-05,
      "loss": 1.705,
      "step": 186
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9925891616027456e-05,
      "loss": 1.7221,
      "step": 187
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9924237708049653e-05,
      "loss": 1.6484,
      "step": 188
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.992256561786659e-05,
      "loss": 1.6872,
      "step": 189
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9920875348541707e-05,
      "loss": 1.5837,
      "step": 190
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9919166903171744e-05,
      "loss": 1.6485,
      "step": 191
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9917440284886744e-05,
      "loss": 1.6711,
      "step": 192
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9915695496850055e-05,
      "loss": 1.6767,
      "step": 193
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9913932542258296e-05,
      "loss": 1.7325,
      "step": 194
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.991215142434138e-05,
      "loss": 1.657,
      "step": 195
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.99103521463625e-05,
      "loss": 1.6882,
      "step": 196
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9908534711618106e-05,
      "loss": 1.6532,
      "step": 197
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.990669912343793e-05,
      "loss": 1.7259,
      "step": 198
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.990484538518495e-05,
      "loss": 1.7128,
      "step": 199
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9902973500255404e-05,
      "loss": 1.7065,
      "step": 200
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.990108347207877e-05,
      "loss": 1.6833,
      "step": 201
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9899175304117773e-05,
      "loss": 1.7199,
      "step": 202
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.989724899986837e-05,
      "loss": 1.7002,
      "step": 203
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.989530456285975e-05,
      "loss": 1.6644,
      "step": 204
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.989334199665431e-05,
      "loss": 1.6741,
      "step": 205
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.989136130484767e-05,
      "loss": 1.6516,
      "step": 206
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9889362491068658e-05,
      "loss": 1.7025,
      "step": 207
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9887345558979304e-05,
      "loss": 1.6879,
      "step": 208
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9885310512274835e-05,
      "loss": 1.677,
      "step": 209
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9883257354683655e-05,
      "loss": 1.7029,
      "step": 210
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9881186089967358e-05,
      "loss": 1.7084,
      "step": 211
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9879096721920716e-05,
      "loss": 1.677,
      "step": 212
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9876989254371654e-05,
      "loss": 1.6909,
      "step": 213
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9874863691181267e-05,
      "loss": 1.6357,
      "step": 214
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9872720036243803e-05,
      "loss": 1.725,
      "step": 215
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9870558293486658e-05,
      "loss": 1.7006,
      "step": 216
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.986837846687035e-05,
      "loss": 1.6993,
      "step": 217
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.986618056038856e-05,
      "loss": 1.6428,
      "step": 218
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.986396457806806e-05,
      "loss": 1.7556,
      "step": 219
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9861730523968762e-05,
      "loss": 1.7063,
      "step": 220
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.985947840218367e-05,
      "loss": 1.7021,
      "step": 221
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9857208216838905e-05,
      "loss": 1.7497,
      "step": 222
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9854919972093676e-05,
      "loss": 1.6359,
      "step": 223
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9852613672140274e-05,
      "loss": 1.6713,
      "step": 224
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9850289321204074e-05,
      "loss": 1.6783,
      "step": 225
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9847946923543525e-05,
      "loss": 1.7054,
      "step": 226
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9845586483450137e-05,
      "loss": 1.6797,
      "step": 227
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.984320800524847e-05,
      "loss": 1.6676,
      "step": 228
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9840811493296134e-05,
      "loss": 1.7225,
      "step": 229
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9838396951983787e-05,
      "loss": 1.6516,
      "step": 230
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.983596438573511e-05,
      "loss": 1.6386,
      "step": 231
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.983351379900681e-05,
      "loss": 1.7135,
      "step": 232
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9831045196288613e-05,
      "loss": 1.6781,
      "step": 233
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.982855858210324e-05,
      "loss": 1.6845,
      "step": 234
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9826053961006426e-05,
      "loss": 1.679,
      "step": 235
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9823531337586884e-05,
      "loss": 1.6082,
      "step": 236
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9820990716466323e-05,
      "loss": 1.7089,
      "step": 237
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.981843210229941e-05,
      "loss": 1.7175,
      "step": 238
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9815855499773784e-05,
      "loss": 1.6602,
      "step": 239
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9813260913610048e-05,
      "loss": 1.6522,
      "step": 240
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.981064834856174e-05,
      "loss": 1.642,
      "step": 241
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9808017809415344e-05,
      "loss": 1.5974,
      "step": 242
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9805369300990275e-05,
      "loss": 1.6775,
      "step": 243
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9802702828138866e-05,
      "loss": 1.6241,
      "step": 244
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9800018395746362e-05,
      "loss": 1.6341,
      "step": 245
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9797316008730922e-05,
      "loss": 1.7105,
      "step": 246
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9794595672043586e-05,
      "loss": 1.6753,
      "step": 247
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9791857390668287e-05,
      "loss": 1.6561,
      "step": 248
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9789101169621827e-05,
      "loss": 1.6477,
      "step": 249
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9786327013953893e-05,
      "loss": 1.6462,
      "step": 250
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9783534928747006e-05,
      "loss": 1.6183,
      "step": 251
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9780724919116556e-05,
      "loss": 1.707,
      "step": 252
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.977789699021076e-05,
      "loss": 1.6888,
      "step": 253
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9775051147210673e-05,
      "loss": 1.6725,
      "step": 254
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9772187395330166e-05,
      "loss": 1.6209,
      "step": 255
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9769305739815918e-05,
      "loss": 1.6534,
      "step": 256
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9766406185947416e-05,
      "loss": 1.7137,
      "step": 257
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9763488739036933e-05,
      "loss": 1.7033,
      "step": 258
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9760553404429528e-05,
      "loss": 1.6526,
      "step": 259
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.975760018750303e-05,
      "loss": 1.6358,
      "step": 260
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.975462909366803e-05,
      "loss": 1.5986,
      "step": 261
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9751640128367872e-05,
      "loss": 1.6659,
      "step": 262
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9748633297078642e-05,
      "loss": 1.7274,
      "step": 263
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.974560860530916e-05,
      "loss": 1.6532,
      "step": 264
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9742566058600968e-05,
      "loss": 1.6748,
      "step": 265
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9739505662528314e-05,
      "loss": 1.5444,
      "step": 266
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9736427422698152e-05,
      "loss": 1.6507,
      "step": 267
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9733331344750132e-05,
      "loss": 1.6663,
      "step": 268
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9730217434356584e-05,
      "loss": 1.6506,
      "step": 269
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9727085697222497e-05,
      "loss": 1.6102,
      "step": 270
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.972393613908554e-05,
      "loss": 1.6284,
      "step": 271
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.972076876571601e-05,
      "loss": 1.6605,
      "step": 272
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9717583582916862e-05,
      "loss": 1.6571,
      "step": 273
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9714380596523676e-05,
      "loss": 1.6522,
      "step": 274
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9711159812404638e-05,
      "loss": 1.6334,
      "step": 275
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9707921236460556e-05,
      "loss": 1.6803,
      "step": 276
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.970466487462482e-05,
      "loss": 1.6925,
      "step": 277
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9701390732863426e-05,
      "loss": 1.6437,
      "step": 278
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9698098817174922e-05,
      "loss": 1.6897,
      "step": 279
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.969478913359043e-05,
      "loss": 1.655,
      "step": 280
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9691461688173632e-05,
      "loss": 1.6167,
      "step": 281
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.968811648702074e-05,
      "loss": 1.6628,
      "step": 282
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9684753536260496e-05,
      "loss": 1.6205,
      "step": 283
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.968137284205417e-05,
      "loss": 1.6794,
      "step": 284
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9677974410595534e-05,
      "loss": 1.6334,
      "step": 285
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9674558248110856e-05,
      "loss": 1.6216,
      "step": 286
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9671124360858896e-05,
      "loss": 1.6624,
      "step": 287
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9667672755130873e-05,
      "loss": 1.6262,
      "step": 288
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9664203437250485e-05,
      "loss": 1.633,
      "step": 289
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.966071641357387e-05,
      "loss": 1.6174,
      "step": 290
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9657211690489602e-05,
      "loss": 1.6414,
      "step": 291
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.96536892744187e-05,
      "loss": 1.614,
      "step": 292
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9650149171814573e-05,
      "loss": 1.6524,
      "step": 293
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.964659138916305e-05,
      "loss": 1.6088,
      "step": 294
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9643015932982355e-05,
      "loss": 1.6395,
      "step": 295
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.9639422809823075e-05,
      "loss": 1.666,
      "step": 296
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.963581202626818e-05,
      "loss": 1.6756,
      "step": 297
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9632183588932987e-05,
      "loss": 1.6583,
      "step": 298
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.962853750446516e-05,
      "loss": 1.6158,
      "step": 299
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.96248737795447e-05,
      "loss": 1.6744,
      "step": 300
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.962119242088391e-05,
      "loss": 1.6503,
      "step": 301
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.961749343522742e-05,
      "loss": 1.6641,
      "step": 302
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.961377682935214e-05,
      "loss": 1.6229,
      "step": 303
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9610042610067266e-05,
      "loss": 1.6643,
      "step": 304
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9606290784214268e-05,
      "loss": 1.6225,
      "step": 305
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.960252135866687e-05,
      "loss": 1.6285,
      "step": 306
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.959873434033104e-05,
      "loss": 1.6346,
      "step": 307
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9594929736144978e-05,
      "loss": 1.6327,
      "step": 308
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9591107553079096e-05,
      "loss": 1.6837,
      "step": 309
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9587267798136027e-05,
      "loss": 1.618,
      "step": 310
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.958341047835059e-05,
      "loss": 1.6611,
      "step": 311
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.957953560078978e-05,
      "loss": 1.5892,
      "step": 312
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9575643172552765e-05,
      "loss": 1.6573,
      "step": 313
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.957173320077087e-05,
      "loss": 1.6387,
      "step": 314
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.956780569260755e-05,
      "loss": 1.6468,
      "step": 315
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9563860655258408e-05,
      "loss": 1.6431,
      "step": 316
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9559898095951137e-05,
      "loss": 1.6084,
      "step": 317
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9555918021945556e-05,
      "loss": 1.577,
      "step": 318
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9551920440533562e-05,
      "loss": 1.6124,
      "step": 319
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.954790535903912e-05,
      "loss": 1.6638,
      "step": 320
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.954387278481827e-05,
      "loss": 1.6443,
      "step": 321
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9539822725259095e-05,
      "loss": 1.7034,
      "step": 322
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.953575518778171e-05,
      "loss": 1.6255,
      "step": 323
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.953167017983826e-05,
      "loss": 1.6374,
      "step": 324
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9527567708912887e-05,
      "loss": 1.6564,
      "step": 325
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9523447782521734e-05,
      "loss": 1.6089,
      "step": 326
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9519310408212918e-05,
      "loss": 1.685,
      "step": 327
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9515155593566536e-05,
      "loss": 1.6374,
      "step": 328
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9510983346194617e-05,
      "loss": 1.6884,
      "step": 329
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9506793673741145e-05,
      "loss": 1.6406,
      "step": 330
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.950258658388202e-05,
      "loss": 1.6164,
      "step": 331
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.949836208432506e-05,
      "loss": 1.5929,
      "step": 332
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.949412018280997e-05,
      "loss": 1.6822,
      "step": 333
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9489860887108344e-05,
      "loss": 1.5831,
      "step": 334
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.948558420502364e-05,
      "loss": 1.6763,
      "step": 335
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.948129014439117e-05,
      "loss": 1.6545,
      "step": 336
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.947697871307809e-05,
      "loss": 1.6325,
      "step": 337
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.947264991898338e-05,
      "loss": 1.6796,
      "step": 338
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.946830377003782e-05,
      "loss": 1.6948,
      "step": 339
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9463940274204e-05,
      "loss": 1.638,
      "step": 340
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.9459559439476282e-05,
      "loss": 1.6297,
      "step": 341
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.94551612738808e-05,
      "loss": 1.6604,
      "step": 342
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.9450745785475427e-05,
      "loss": 1.6476,
      "step": 343
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.94463129823498e-05,
      "loss": 1.6686,
      "step": 344
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.944186287262525e-05,
      "loss": 1.6702,
      "step": 345
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.943739546445483e-05,
      "loss": 1.6678,
      "step": 346
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.9432910766023284e-05,
      "loss": 1.6907,
      "step": 347
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.9428408785547033e-05,
      "loss": 1.6784,
      "step": 348
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.9423889531274157e-05,
      "loss": 1.6818,
      "step": 349
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.941935301148439e-05,
      "loss": 1.6243,
      "step": 350
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.9414799234489084e-05,
      "loss": 1.6198,
      "step": 351
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.941022820863123e-05,
      "loss": 1.6236,
      "step": 352
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.9405639942285406e-05,
      "loss": 1.6224,
      "step": 353
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.9401034443857776e-05,
      "loss": 1.6567,
      "step": 354
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.9396411721786078e-05,
      "loss": 1.7099,
      "step": 355
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.939177178453961e-05,
      "loss": 1.6412,
      "step": 356
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.9387114640619198e-05,
      "loss": 1.6372,
      "step": 357
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.938244029855721e-05,
      "loss": 1.651,
      "step": 358
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.9377748766917502e-05,
      "loss": 1.6032,
      "step": 359
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.9373040054295437e-05,
      "loss": 1.6179,
      "step": 360
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9368314169317858e-05,
      "loss": 1.6082,
      "step": 361
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9363571120643055e-05,
      "loss": 1.6624,
      "step": 362
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9358810916960774e-05,
      "loss": 1.5887,
      "step": 363
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.935403356699219e-05,
      "loss": 1.6192,
      "step": 364
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9349239079489893e-05,
      "loss": 1.6571,
      "step": 365
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9344427463237868e-05,
      "loss": 1.6024,
      "step": 366
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.933959872705147e-05,
      "loss": 1.5968,
      "step": 367
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9334752879777447e-05,
      "loss": 1.6258,
      "step": 368
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.9329889930293873e-05,
      "loss": 1.5858,
      "step": 369
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.932500988751016e-05,
      "loss": 1.6323,
      "step": 370
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.9320112760367044e-05,
      "loss": 1.5766,
      "step": 371
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.9315198557836555e-05,
      "loss": 1.623,
      "step": 372
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.9310267288922003e-05,
      "loss": 1.6172,
      "step": 373
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.9305318962657978e-05,
      "loss": 1.6035,
      "step": 374
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.9300353588110307e-05,
      "loss": 1.5915,
      "step": 375
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.9295371174376063e-05,
      "loss": 1.639,
      "step": 376
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9290371730583527e-05,
      "loss": 1.6439,
      "step": 377
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9285355265892184e-05,
      "loss": 1.6535,
      "step": 378
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9280321789492708e-05,
      "loss": 1.6903,
      "step": 379
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9275271310606933e-05,
      "loss": 1.5702,
      "step": 380
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.9270203838487847e-05,
      "loss": 1.6285,
      "step": 381
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.9265119382419568e-05,
      "loss": 1.5934,
      "step": 382
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.9260017951717334e-05,
      "loss": 1.6,
      "step": 383
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.9254899555727483e-05,
      "loss": 1.6378,
      "step": 384
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.924976420382743e-05,
      "loss": 1.5822,
      "step": 385
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.9244611905425663e-05,
      "loss": 1.6013,
      "step": 386
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.9239442669961708e-05,
      "loss": 1.6487,
      "step": 387
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.9234256506906128e-05,
      "loss": 1.5913,
      "step": 388
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.9229053425760494e-05,
      "loss": 1.6578,
      "step": 389
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.9223833436057383e-05,
      "loss": 1.5695,
      "step": 390
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.921859654736034e-05,
      "loss": 1.6656,
      "step": 391
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.9213342769263867e-05,
      "loss": 1.6092,
      "step": 392
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9208072111393425e-05,
      "loss": 1.6162,
      "step": 393
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9202784583405386e-05,
      "loss": 1.6474,
      "step": 394
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9197480194987038e-05,
      "loss": 1.6096,
      "step": 395
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9192158955856556e-05,
      "loss": 1.6178,
      "step": 396
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9186820875762982e-05,
      "loss": 1.6237,
      "step": 397
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9181465964486222e-05,
      "loss": 1.63,
      "step": 398
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.917609423183702e-05,
      "loss": 1.6314,
      "step": 399
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9170705687656918e-05,
      "loss": 1.6044,
      "step": 400
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9165300341818283e-05,
      "loss": 1.6481,
      "step": 401
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.9159878204224254e-05,
      "loss": 1.6592,
      "step": 402
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.915443928480873e-05,
      "loss": 1.6291,
      "step": 403
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.914898359353636e-05,
      "loss": 1.3749,
      "step": 404
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.9143511140402532e-05,
      "loss": 1.3916,
      "step": 405
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.9138021935433323e-05,
      "loss": 1.3096,
      "step": 406
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.9132515988685507e-05,
      "loss": 1.338,
      "step": 407
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.912699331024654e-05,
      "loss": 1.2852,
      "step": 408
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.912145391023453e-05,
      "loss": 1.3468,
      "step": 409
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.9115897798798205e-05,
      "loss": 1.3444,
      "step": 410
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.9110324986116938e-05,
      "loss": 1.3227,
      "step": 411
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.9104735482400672e-05,
      "loss": 1.351,
      "step": 412
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.909912929788995e-05,
      "loss": 1.3704,
      "step": 413
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.9093506442855858e-05,
      "loss": 1.3497,
      "step": 414
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.9087866927600042e-05,
      "loss": 1.3437,
      "step": 415
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.908221076245466e-05,
      "loss": 1.3284,
      "step": 416
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.907653795778238e-05,
      "loss": 1.2519,
      "step": 417
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.907084852397635e-05,
      "loss": 1.3062,
      "step": 418
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.9065142471460195e-05,
      "loss": 1.3462,
      "step": 419
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.9059419810687964e-05,
      "loss": 1.3003,
      "step": 420
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.9053680552144165e-05,
      "loss": 1.3354,
      "step": 421
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.904792470634369e-05,
      "loss": 1.332,
      "step": 422
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.9042152283831827e-05,
      "loss": 1.3326,
      "step": 423
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.9036363295184245e-05,
      "loss": 1.3092,
      "step": 424
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.903055775100695e-05,
      "loss": 1.3659,
      "step": 425
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.9024735661936285e-05,
      "loss": 1.3425,
      "step": 426
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.901889703863891e-05,
      "loss": 1.3381,
      "step": 427
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.901304189181177e-05,
      "loss": 1.2971,
      "step": 428
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.9007170232182083e-05,
      "loss": 1.3093,
      "step": 429
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.900128207050732e-05,
      "loss": 1.3624,
      "step": 430
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8995377417575193e-05,
      "loss": 1.3123,
      "step": 431
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.8989456284203614e-05,
      "loss": 1.3137,
      "step": 432
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.898351868124071e-05,
      "loss": 1.3411,
      "step": 433
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8977564619564757e-05,
      "loss": 1.3214,
      "step": 434
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8971594110084207e-05,
      "loss": 1.3392,
      "step": 435
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8965607163737628e-05,
      "loss": 1.3274,
      "step": 436
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.8959603791493714e-05,
      "loss": 1.3059,
      "step": 437
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8953584004351243e-05,
      "loss": 1.2688,
      "step": 438
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.894754781333908e-05,
      "loss": 1.3328,
      "step": 439
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.8941495229516135e-05,
      "loss": 1.3194,
      "step": 440
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.893542626397135e-05,
      "loss": 1.2464,
      "step": 441
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8929340927823686e-05,
      "loss": 1.2884,
      "step": 442
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8923239232222087e-05,
      "loss": 1.3719,
      "step": 443
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.8917121188345483e-05,
      "loss": 1.3473,
      "step": 444
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.891098680740274e-05,
      "loss": 1.376,
      "step": 445
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8904836100632668e-05,
      "loss": 1.3038,
      "step": 446
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8898669079303983e-05,
      "loss": 1.3065,
      "step": 447
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8892485754715292e-05,
      "loss": 1.3636,
      "step": 448
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.8886286138195063e-05,
      "loss": 1.3509,
      "step": 449
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8880070241101623e-05,
      "loss": 1.3688,
      "step": 450
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.8873838074823126e-05,
      "loss": 1.3147,
      "step": 451
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.886758965077753e-05,
      "loss": 1.3358,
      "step": 452
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.886132498041257e-05,
      "loss": 1.3193,
      "step": 453
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.885504407520576e-05,
      "loss": 1.3146,
      "step": 454
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8848746946664355e-05,
      "loss": 1.3473,
      "step": 455
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.884243360632533e-05,
      "loss": 1.3289,
      "step": 456
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.8836104065755356e-05,
      "loss": 1.3539,
      "step": 457
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8829758336550794e-05,
      "loss": 1.3262,
      "step": 458
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8823396430337663e-05,
      "loss": 1.3705,
      "step": 459
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8817018358771612e-05,
      "loss": 1.373,
      "step": 460
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.8810624133537913e-05,
      "loss": 1.3184,
      "step": 461
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.8804213766351437e-05,
      "loss": 1.3159,
      "step": 462
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.879778726895662e-05,
      "loss": 1.2995,
      "step": 463
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.879134465312745e-05,
      "loss": 1.3383,
      "step": 464
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.8784885930667455e-05,
      "loss": 1.388,
      "step": 465
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.8778411113409668e-05,
      "loss": 1.3413,
      "step": 466
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.8771920213216604e-05,
      "loss": 1.3515,
      "step": 467
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.876541324198025e-05,
      "loss": 1.3786,
      "step": 468
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.875889021162203e-05,
      "loss": 1.3491,
      "step": 469
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.87523511340928e-05,
      "loss": 1.2793,
      "step": 470
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.87457960213728e-05,
      "loss": 1.3165,
      "step": 471
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.8739224885471672e-05,
      "loss": 1.3319,
      "step": 472
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.873263773842839e-05,
      "loss": 1.3327,
      "step": 473
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.8726034592311275e-05,
      "loss": 1.3433,
      "step": 474
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.871941545921796e-05,
      "loss": 1.3474,
      "step": 475
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.8712780351275358e-05,
      "loss": 1.3265,
      "step": 476
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.8706129280639667e-05,
      "loss": 1.2739,
      "step": 477
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.869946225949631e-05,
      "loss": 1.3437,
      "step": 478
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.869277930005995e-05,
      "loss": 1.3181,
      "step": 479
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.8686080414574437e-05,
      "loss": 1.2839,
      "step": 480
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.867936561531281e-05,
      "loss": 1.322,
      "step": 481
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.867263491457726e-05,
      "loss": 1.2906,
      "step": 482
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.8665888324699104e-05,
      "loss": 1.2876,
      "step": 483
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.865912585803878e-05,
      "loss": 1.3869,
      "step": 484
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.865234752698581e-05,
      "loss": 1.2987,
      "step": 485
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.864555334395878e-05,
      "loss": 1.3754,
      "step": 486
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.8638743321405316e-05,
      "loss": 1.3833,
      "step": 487
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.863191747180207e-05,
      "loss": 1.3147,
      "step": 488
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.8625075807654688e-05,
      "loss": 1.2685,
      "step": 489
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.861821834149779e-05,
      "loss": 1.3257,
      "step": 490
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.861134508589494e-05,
      "loss": 1.3299,
      "step": 491
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.8604456053438645e-05,
      "loss": 1.3173,
      "step": 492
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.85975512567503e-05,
      "loss": 1.3567,
      "step": 493
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.8590630708480194e-05,
      "loss": 1.2953,
      "step": 494
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.8583694421307473e-05,
      "loss": 1.3632,
      "step": 495
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.8576742407940106e-05,
      "loss": 1.3415,
      "step": 496
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.8569774681114892e-05,
      "loss": 1.3506,
      "step": 497
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.8562791253597402e-05,
      "loss": 1.3236,
      "step": 498
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.8555792138181988e-05,
      "loss": 1.3063,
      "step": 499
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.8548777347691732e-05,
      "loss": 1.394,
      "step": 500
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.854174689497844e-05,
      "loss": 1.3066,
      "step": 501
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.8534700792922606e-05,
      "loss": 1.3251,
      "step": 502
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.8527639054433405e-05,
      "loss": 1.3087,
      "step": 503
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.8520561692448655e-05,
      "loss": 1.3111,
      "step": 504
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.8513468719934797e-05,
      "loss": 1.3511,
      "step": 505
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.850636014988687e-05,
      "loss": 1.3265,
      "step": 506
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.8499235995328497e-05,
      "loss": 1.35,
      "step": 507
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.849209626931184e-05,
      "loss": 1.3514,
      "step": 508
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.8484940984917608e-05,
      "loss": 1.3146,
      "step": 509
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.8477770155254996e-05,
      "loss": 1.3193,
      "step": 510
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.847058379346169e-05,
      "loss": 1.3805,
      "step": 511
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.8463381912703833e-05,
      "loss": 1.349,
      "step": 512
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.8456164526175996e-05,
      "loss": 1.3453,
      "step": 513
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.8448931647101158e-05,
      "loss": 1.3418,
      "step": 514
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.8441683288730686e-05,
      "loss": 1.3319,
      "step": 515
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.8434419464344305e-05,
      "loss": 1.3735,
      "step": 516
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.842714018725007e-05,
      "loss": 1.3719,
      "step": 517
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8419845470784354e-05,
      "loss": 1.295,
      "step": 518
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8412535328311813e-05,
      "loss": 1.3286,
      "step": 519
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.840520977322537e-05,
      "loss": 1.3416,
      "step": 520
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.8397868818946173e-05,
      "loss": 1.3667,
      "step": 521
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.83905124789236e-05,
      "loss": 1.3562,
      "step": 522
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8383140766635205e-05,
      "loss": 1.363,
      "step": 523
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.837575369558671e-05,
      "loss": 1.366,
      "step": 524
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.8368351279311972e-05,
      "loss": 1.2956,
      "step": 525
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.836093353137297e-05,
      "loss": 1.3471,
      "step": 526
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.835350046535976e-05,
      "loss": 1.3533,
      "step": 527
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.834605209489047e-05,
      "loss": 1.326,
      "step": 528
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8338588433611275e-05,
      "loss": 1.3208,
      "step": 529
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.8331109495196348e-05,
      "loss": 1.3187,
      "step": 530
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8323615293347862e-05,
      "loss": 1.3292,
      "step": 531
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8316105841795952e-05,
      "loss": 1.3326,
      "step": 532
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.8308581154298687e-05,
      "loss": 1.3299,
      "step": 533
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.830104124464206e-05,
      "loss": 1.2837,
      "step": 534
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8293486126639942e-05,
      "loss": 1.3731,
      "step": 535
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8285915814134074e-05,
      "loss": 1.3829,
      "step": 536
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8278330320994035e-05,
      "loss": 1.3568,
      "step": 537
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.8270729661117205e-05,
      "loss": 1.2714,
      "step": 538
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.826311384842877e-05,
      "loss": 1.3703,
      "step": 539
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.8255482896881665e-05,
      "loss": 1.3299,
      "step": 540
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.824783682045656e-05,
      "loss": 1.3332,
      "step": 541
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.824017563316184e-05,
      "loss": 1.3677,
      "step": 542
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8232499349033576e-05,
      "loss": 1.3257,
      "step": 543
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8224807982135496e-05,
      "loss": 1.3776,
      "step": 544
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.821710154655896e-05,
      "loss": 1.3331,
      "step": 545
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.8209380056422936e-05,
      "loss": 1.3573,
      "step": 546
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8201643525873964e-05,
      "loss": 1.3839,
      "step": 547
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8193891969086164e-05,
      "loss": 1.3254,
      "step": 548
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.818612540026116e-05,
      "loss": 1.2981,
      "step": 549
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.8178343833628095e-05,
      "loss": 1.3095,
      "step": 550
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8170547283443584e-05,
      "loss": 1.3586,
      "step": 551
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8162735763991688e-05,
      "loss": 1.348,
      "step": 552
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8154909289583908e-05,
      "loss": 1.3892,
      "step": 553
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.8147067874559134e-05,
      "loss": 1.3269,
      "step": 554
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.8139211533283623e-05,
      "loss": 1.3275,
      "step": 555
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.813134028015099e-05,
      "loss": 1.3582,
      "step": 556
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.8123454129582164e-05,
      "loss": 1.3583,
      "step": 557
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.811555309602537e-05,
      "loss": 1.3834,
      "step": 558
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.8107637193956102e-05,
      "loss": 1.3064,
      "step": 559
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.8099706437877082e-05,
      "loss": 1.3888,
      "step": 560
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.8091760842318266e-05,
      "loss": 1.325,
      "step": 561
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.808380042183678e-05,
      "loss": 1.3143,
      "step": 562
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.8075825191016923e-05,
      "loss": 1.3076,
      "step": 563
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.806783516447012e-05,
      "loss": 1.3487,
      "step": 564
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.8059830356834908e-05,
      "loss": 1.3044,
      "step": 565
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.8051810782776905e-05,
      "loss": 1.3292,
      "step": 566
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.8043776456988772e-05,
      "loss": 1.341,
      "step": 567
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.803572739419021e-05,
      "loss": 1.3344,
      "step": 568
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.8027663609127914e-05,
      "loss": 1.3343,
      "step": 569
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.8019585116575554e-05,
      "loss": 1.3162,
      "step": 570
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.8011491931333743e-05,
      "loss": 1.3283,
      "step": 571
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.800338406823001e-05,
      "loss": 1.3576,
      "step": 572
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7995261542118783e-05,
      "loss": 1.3393,
      "step": 573
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7987124367881346e-05,
      "loss": 1.3497,
      "step": 574
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7978972560425826e-05,
      "loss": 1.3584,
      "step": 575
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7970806134687157e-05,
      "loss": 1.3271,
      "step": 576
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7962625105627053e-05,
      "loss": 1.3616,
      "step": 577
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.795442948823399e-05,
      "loss": 1.3163,
      "step": 578
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.794621929752316e-05,
      "loss": 1.3117,
      "step": 579
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7937994548536467e-05,
      "loss": 1.2947,
      "step": 580
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.792975525634248e-05,
      "loss": 1.3857,
      "step": 581
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.792150143603641e-05,
      "loss": 1.3646,
      "step": 582
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7913233102740096e-05,
      "loss": 1.3635,
      "step": 583
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.790495027160195e-05,
      "loss": 1.2955,
      "step": 584
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7896652957796962e-05,
      "loss": 1.3251,
      "step": 585
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7888341176526645e-05,
      "loss": 1.3493,
      "step": 586
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7880014943019022e-05,
      "loss": 1.3521,
      "step": 587
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7871674272528598e-05,
      "loss": 1.3305,
      "step": 588
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7863319180336314e-05,
      "loss": 1.3178,
      "step": 589
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.785494968174955e-05,
      "loss": 1.3602,
      "step": 590
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.7846565792102065e-05,
      "loss": 1.2989,
      "step": 591
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.7838167526754002e-05,
      "loss": 1.2728,
      "step": 592
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.7829754901091824e-05,
      "loss": 1.3277,
      "step": 593
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.7821327930528304e-05,
      "loss": 1.379,
      "step": 594
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.781288663050252e-05,
      "loss": 1.3665,
      "step": 595
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.7804431016479765e-05,
      "loss": 1.33,
      "step": 596
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.7795961103951594e-05,
      "loss": 1.3508,
      "step": 597
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.778747690843573e-05,
      "loss": 1.3503,
      "step": 598
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.7778978445476088e-05,
      "loss": 1.3034,
      "step": 599
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.7770465730642702e-05,
      "loss": 1.3114,
      "step": 600
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.7761938779531724e-05,
      "loss": 1.3634,
      "step": 601
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.775339760776539e-05,
      "loss": 1.3301,
      "step": 602
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.774484223099199e-05,
      "loss": 1.3288,
      "step": 603
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.7736272664885842e-05,
      "loss": 1.3816,
      "step": 604
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.7727688925147247e-05,
      "loss": 1.3541,
      "step": 605
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.7719091027502492e-05,
      "loss": 1.3134,
      "step": 606
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.7710478987703784e-05,
      "loss": 1.3072,
      "step": 607
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.770185282152926e-05,
      "loss": 1.3446,
      "step": 608
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.7693212544782916e-05,
      "loss": 1.3732,
      "step": 609
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.7684558173294618e-05,
      "loss": 1.3248,
      "step": 610
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.767588972292005e-05,
      "loss": 1.335,
      "step": 611
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.7667207209540686e-05,
      "loss": 1.321,
      "step": 612
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.765851064906376e-05,
      "loss": 1.3829,
      "step": 613
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.7649800057422256e-05,
      "loss": 1.3317,
      "step": 614
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.7641075450574858e-05,
      "loss": 1.2972,
      "step": 615
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.7632336844505925e-05,
      "loss": 1.3327,
      "step": 616
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.7623584255225462e-05,
      "loss": 1.3811,
      "step": 617
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.76148176987691e-05,
      "loss": 1.3476,
      "step": 618
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.7606037191198052e-05,
      "loss": 1.3702,
      "step": 619
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.7597242748599096e-05,
      "loss": 1.3655,
      "step": 620
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.758843438708454e-05,
      "loss": 1.3339,
      "step": 621
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.7579612122792187e-05,
      "loss": 1.2943,
      "step": 622
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.7570775971885322e-05,
      "loss": 1.3555,
      "step": 623
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.7561925950552666e-05,
      "loss": 1.3055,
      "step": 624
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.755306207500834e-05,
      "loss": 1.3418,
      "step": 625
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.754418436149187e-05,
      "loss": 1.3703,
      "step": 626
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.753529282626812e-05,
      "loss": 1.33,
      "step": 627
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.752638748562728e-05,
      "loss": 1.3452,
      "step": 628
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.7517468355884832e-05,
      "loss": 1.3662,
      "step": 629
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.7508535453381523e-05,
      "loss": 1.3886,
      "step": 630
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.749958879448333e-05,
      "loss": 1.3492,
      "step": 631
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.749062839558144e-05,
      "loss": 1.2925,
      "step": 632
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.7481654273092204e-05,
      "loss": 1.2977,
      "step": 633
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.7472666443457123e-05,
      "loss": 1.3367,
      "step": 634
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.7463664923142806e-05,
      "loss": 1.3478,
      "step": 635
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.7454649728640944e-05,
      "loss": 1.3712,
      "step": 636
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.7445620876468293e-05,
      "loss": 1.3015,
      "step": 637
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.743657838316661e-05,
      "loss": 1.3202,
      "step": 638
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.7427522265302665e-05,
      "loss": 1.3499,
      "step": 639
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.7418452539468177e-05,
      "loss": 1.3647,
      "step": 640
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.7409369222279794e-05,
      "loss": 1.3602,
      "step": 641
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.740027233037908e-05,
      "loss": 1.3313,
      "step": 642
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7391161880432446e-05,
      "loss": 1.3256,
      "step": 643
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7382037889131168e-05,
      "loss": 1.2922,
      "step": 644
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7372900373191316e-05,
      "loss": 1.3975,
      "step": 645
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7363749349353737e-05,
      "loss": 1.3123,
      "step": 646
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.7354584834384036e-05,
      "loss": 1.273,
      "step": 647
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.7345406845072525e-05,
      "loss": 1.3425,
      "step": 648
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.733621539823421e-05,
      "loss": 1.3907,
      "step": 649
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.7327010510708747e-05,
      "loss": 1.3303,
      "step": 650
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.7317792199360422e-05,
      "loss": 1.3635,
      "step": 651
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.7308560481078108e-05,
      "loss": 1.3318,
      "step": 652
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.7299315372775253e-05,
      "loss": 1.3575,
      "step": 653
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.729005689138982e-05,
      "loss": 1.3421,
      "step": 654
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.728078505388428e-05,
      "loss": 1.3375,
      "step": 655
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7271499877245582e-05,
      "loss": 1.3219,
      "step": 656
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7262201378485107e-05,
      "loss": 1.3228,
      "step": 657
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.725288957463864e-05,
      "loss": 1.3578,
      "step": 658
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7243564482766344e-05,
      "loss": 1.3432,
      "step": 659
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.723422611995273e-05,
      "loss": 1.3154,
      "step": 660
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.7224874503306625e-05,
      "loss": 1.3684,
      "step": 661
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.721550964996113e-05,
      "loss": 1.3653,
      "step": 662
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.7206131577073605e-05,
      "loss": 1.3417,
      "step": 663
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7196740301825625e-05,
      "loss": 1.3575,
      "step": 664
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7187335841422953e-05,
      "loss": 1.3788,
      "step": 665
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.717791821309551e-05,
      "loss": 1.282,
      "step": 666
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.7168487434097346e-05,
      "loss": 1.3218,
      "step": 667
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.715904352170659e-05,
      "loss": 1.3473,
      "step": 668
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7149586493225453e-05,
      "loss": 1.3465,
      "step": 669
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.7140116365980164e-05,
      "loss": 1.3759,
      "step": 670
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.713063315732095e-05,
      "loss": 1.2637,
      "step": 671
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.7121136884622006e-05,
      "loss": 1.2885,
      "step": 672
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.7111627565281462e-05,
      "loss": 1.3476,
      "step": 673
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.710210521672135e-05,
      "loss": 1.3134,
      "step": 674
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.7092569856387577e-05,
      "loss": 1.3685,
      "step": 675
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.7083021501749886e-05,
      "loss": 1.3066,
      "step": 676
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.7073460170301824e-05,
      "loss": 1.3344,
      "step": 677
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.7063885879560717e-05,
      "loss": 1.3476,
      "step": 678
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.705429864706763e-05,
      "loss": 1.3515,
      "step": 679
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.704469849038734e-05,
      "loss": 1.3376,
      "step": 680
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.7035085427108308e-05,
      "loss": 1.3466,
      "step": 681
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.702545947484263e-05,
      "loss": 1.3561,
      "step": 682
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.7015820651226027e-05,
      "loss": 1.3793,
      "step": 683
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.7006168973917797e-05,
      "loss": 1.3419,
      "step": 684
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.6996504460600786e-05,
      "loss": 1.3783,
      "step": 685
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.6986827128981354e-05,
      "loss": 1.3641,
      "step": 686
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.6977136996789354e-05,
      "loss": 1.3647,
      "step": 687
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.6967434081778086e-05,
      "loss": 1.3614,
      "step": 688
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.695771840172427e-05,
      "loss": 1.3507,
      "step": 689
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.694798997442801e-05,
      "loss": 1.3552,
      "step": 690
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.6938248817712767e-05,
      "loss": 1.3138,
      "step": 691
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.6928494949425328e-05,
      "loss": 1.3399,
      "step": 692
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.6918728387435758e-05,
      "loss": 1.3413,
      "step": 693
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.6908949149637387e-05,
      "loss": 1.336,
      "step": 694
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.6899157253946767e-05,
      "loss": 1.3454,
      "step": 695
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.688935271830364e-05,
      "loss": 1.3203,
      "step": 696
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.68795355606709e-05,
      "loss": 1.3484,
      "step": 697
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.6869705799034578e-05,
      "loss": 1.2956,
      "step": 698
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.6859863451403783e-05,
      "loss": 1.3376,
      "step": 699
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.6850008535810696e-05,
      "loss": 1.3634,
      "step": 700
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.684014107031051e-05,
      "loss": 1.3714,
      "step": 701
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.6830261072981423e-05,
      "loss": 1.3794,
      "step": 702
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.6820368561924586e-05,
      "loss": 1.2815,
      "step": 703
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.6810463555264073e-05,
      "loss": 1.3725,
      "step": 704
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.6800546071146863e-05,
      "loss": 1.3348,
      "step": 705
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.679061612774279e-05,
      "loss": 1.3873,
      "step": 706
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.6780673743244502e-05,
      "loss": 1.2803,
      "step": 707
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.6770718935867457e-05,
      "loss": 1.3647,
      "step": 708
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.6760751723849874e-05,
      "loss": 1.3653,
      "step": 709
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.6750772125452677e-05,
      "loss": 1.3768,
      "step": 710
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.674078015895951e-05,
      "loss": 1.3462,
      "step": 711
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.6730775842676658e-05,
      "loss": 1.3496,
      "step": 712
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.6720759194933037e-05,
      "loss": 1.3259,
      "step": 713
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.6710730234080165e-05,
      "loss": 1.3641,
      "step": 714
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.67006889784921e-05,
      "loss": 1.3361,
      "step": 715
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.6690635446565444e-05,
      "loss": 1.3233,
      "step": 716
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.668056965671928e-05,
      "loss": 1.438,
      "step": 717
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.667049162739515e-05,
      "loss": 1.3713,
      "step": 718
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.6660401377057018e-05,
      "loss": 1.3685,
      "step": 719
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.6650298924191245e-05,
      "loss": 1.3678,
      "step": 720
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.664018428730655e-05,
      "loss": 1.3532,
      "step": 721
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.6630057484933956e-05,
      "loss": 1.3755,
      "step": 722
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.66199185356268e-05,
      "loss": 1.328,
      "step": 723
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.660976745796065e-05,
      "loss": 1.3796,
      "step": 724
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.6599604270533308e-05,
      "loss": 1.348,
      "step": 725
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.6589428991964768e-05,
      "loss": 1.3613,
      "step": 726
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.6579241640897153e-05,
      "loss": 1.3064,
      "step": 727
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.656904223599473e-05,
      "loss": 1.3538,
      "step": 728
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.6558830795943833e-05,
      "loss": 1.3258,
      "step": 729
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.6548607339452853e-05,
      "loss": 1.3739,
      "step": 730
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.653837188525219e-05,
      "loss": 1.3467,
      "step": 731
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.6528124452094238e-05,
      "loss": 1.4009,
      "step": 732
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.6517865058753316e-05,
      "loss": 1.3137,
      "step": 733
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.6507593724025683e-05,
      "loss": 1.3395,
      "step": 734
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.6497310466729448e-05,
      "loss": 1.3088,
      "step": 735
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.6487015305704586e-05,
      "loss": 1.348,
      "step": 736
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.647670825981287e-05,
      "loss": 1.3324,
      "step": 737
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.646638934793784e-05,
      "loss": 1.3686,
      "step": 738
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.6456058588984792e-05,
      "loss": 1.3026,
      "step": 739
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.644571600188072e-05,
      "loss": 1.2777,
      "step": 740
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.643536160557428e-05,
      "loss": 1.3711,
      "step": 741
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.6424995419035777e-05,
      "loss": 1.3306,
      "step": 742
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.641461746125711e-05,
      "loss": 1.354,
      "step": 743
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.6404227751251747e-05,
      "loss": 1.3456,
      "step": 744
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.639382630805468e-05,
      "loss": 1.3295,
      "step": 745
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.6383413150722417e-05,
      "loss": 1.358,
      "step": 746
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.6372988298332896e-05,
      "loss": 1.3787,
      "step": 747
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.636255176998551e-05,
      "loss": 1.3403,
      "step": 748
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.6352103584801024e-05,
      "loss": 1.3818,
      "step": 749
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.6341643761921573e-05,
      "loss": 1.3326,
      "step": 750
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.6331172320510607e-05,
      "loss": 1.3696,
      "step": 751
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6320689279752866e-05,
      "loss": 1.3143,
      "step": 752
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6310194658854338e-05,
      "loss": 1.3178,
      "step": 753
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6299688477042227e-05,
      "loss": 1.3526,
      "step": 754
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.628917075356492e-05,
      "loss": 1.369,
      "step": 755
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.627864150769194e-05,
      "loss": 1.3234,
      "step": 756
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.626810075871394e-05,
      "loss": 1.3799,
      "step": 757
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.6257548525942635e-05,
      "loss": 1.3351,
      "step": 758
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.6246984828710773e-05,
      "loss": 1.3313,
      "step": 759
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.623640968637212e-05,
      "loss": 1.3487,
      "step": 760
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6225823118301395e-05,
      "loss": 1.3711,
      "step": 761
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6215225143894274e-05,
      "loss": 1.3409,
      "step": 762
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6204615782567304e-05,
      "loss": 1.3299,
      "step": 763
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.619399505375791e-05,
      "loss": 1.3477,
      "step": 764
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.618336297692434e-05,
      "loss": 1.3563,
      "step": 765
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.617271957154563e-05,
      "loss": 1.3073,
      "step": 766
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.6162064857121578e-05,
      "loss": 1.3605,
      "step": 767
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.615139885317269e-05,
      "loss": 1.3638,
      "step": 768
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.6140721579240164e-05,
      "loss": 1.3407,
      "step": 769
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.6130033054885844e-05,
      "loss": 1.3423,
      "step": 770
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.6119333299692183e-05,
      "loss": 1.378,
      "step": 771
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.6108622333262217e-05,
      "loss": 1.3401,
      "step": 772
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.609790017521952e-05,
      "loss": 1.3513,
      "step": 773
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.6087166845208158e-05,
      "loss": 1.3732,
      "step": 774
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.6076422362892687e-05,
      "loss": 1.3336,
      "step": 775
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.6065666747958073e-05,
      "loss": 1.3472,
      "step": 776
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.605490002010969e-05,
      "loss": 1.3117,
      "step": 777
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.6044122199073276e-05,
      "loss": 1.318,
      "step": 778
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.6033333304594886e-05,
      "loss": 1.3407,
      "step": 779
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.6022533356440858e-05,
      "loss": 1.3792,
      "step": 780
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.6011722374397787e-05,
      "loss": 1.3514,
      "step": 781
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.600090037827249e-05,
      "loss": 1.338,
      "step": 782
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.5990067387891953e-05,
      "loss": 1.3477,
      "step": 783
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.597922342310331e-05,
      "loss": 1.3407,
      "step": 784
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.5968368503773794e-05,
      "loss": 1.3436,
      "step": 785
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.595750264979072e-05,
      "loss": 1.3383,
      "step": 786
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.594662588106142e-05,
      "loss": 1.3711,
      "step": 787
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.593573821751324e-05,
      "loss": 1.3911,
      "step": 788
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.592483967909348e-05,
      "loss": 1.3059,
      "step": 789
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.5913930285769356e-05,
      "loss": 1.3385,
      "step": 790
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.590301005752798e-05,
      "loss": 1.3039,
      "step": 791
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.5892079014376313e-05,
      "loss": 1.3651,
      "step": 792
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.588113717634113e-05,
      "loss": 1.3557,
      "step": 793
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.5870184563468982e-05,
      "loss": 1.303,
      "step": 794
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.5859221195826163e-05,
      "loss": 1.347,
      "step": 795
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.584824709349867e-05,
      "loss": 1.3422,
      "step": 796
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.583726227659216e-05,
      "loss": 1.3287,
      "step": 797
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.582626676523193e-05,
      "loss": 1.2864,
      "step": 798
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.5815260579562868e-05,
      "loss": 1.3118,
      "step": 799
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.580424373974942e-05,
      "loss": 1.3256,
      "step": 800
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.579321626597554e-05,
      "loss": 1.3104,
      "step": 801
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.578217817844468e-05,
      "loss": 1.3214,
      "step": 802
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.577112949737973e-05,
      "loss": 1.3059,
      "step": 803
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.576007024302299e-05,
      "loss": 1.3529,
      "step": 804
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.574900043563612e-05,
      "loss": 1.3473,
      "step": 805
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.5737920095500138e-05,
      "loss": 1.3597,
      "step": 806
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.572682924291534e-05,
      "loss": 1.0561,
      "step": 807
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.5715727898201286e-05,
      "loss": 1.0105,
      "step": 808
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.570461608169676e-05,
      "loss": 0.9776,
      "step": 809
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.569349381375974e-05,
      "loss": 0.9904,
      "step": 810
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.568236111476733e-05,
      "loss": 0.9595,
      "step": 811
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.5671218005115767e-05,
      "loss": 0.9937,
      "step": 812
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.566006450522035e-05,
      "loss": 0.9525,
      "step": 813
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.5648900635515422e-05,
      "loss": 0.9771,
      "step": 814
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.5637726416454308e-05,
      "loss": 0.9343,
      "step": 815
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.5626541868509318e-05,
      "loss": 0.9788,
      "step": 816
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.5615347012171665e-05,
      "loss": 0.9637,
      "step": 817
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.5604141867951457e-05,
      "loss": 0.9081,
      "step": 818
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.5592926456377648e-05,
      "loss": 0.9614,
      "step": 819
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.5581700797998007e-05,
      "loss": 0.9636,
      "step": 820
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.5570464913379074e-05,
      "loss": 0.935,
      "step": 821
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.555921882310612e-05,
      "loss": 0.9698,
      "step": 822
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.5547962547783126e-05,
      "loss": 0.9182,
      "step": 823
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.553669610803271e-05,
      "loss": 0.9461,
      "step": 824
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.552541952449614e-05,
      "loss": 0.9373,
      "step": 825
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.551413281783325e-05,
      "loss": 0.9326,
      "step": 826
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.5502836008722425e-05,
      "loss": 0.9433,
      "step": 827
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.549152911786056e-05,
      "loss": 0.9111,
      "step": 828
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.5480212165963025e-05,
      "loss": 0.942,
      "step": 829
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.546888517376361e-05,
      "loss": 0.9219,
      "step": 830
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.5457548162014516e-05,
      "loss": 0.9334,
      "step": 831
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.544620115148629e-05,
      "loss": 0.9222,
      "step": 832
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.54348441629678e-05,
      "loss": 0.9417,
      "step": 833
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.54234772172662e-05,
      "loss": 0.9023,
      "step": 834
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.5412100335206873e-05,
      "loss": 0.9719,
      "step": 835
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.540071353763342e-05,
      "loss": 0.9751,
      "step": 836
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.5389316845407605e-05,
      "loss": 0.9144,
      "step": 837
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.537791027940932e-05,
      "loss": 0.9096,
      "step": 838
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.5366493860536536e-05,
      "loss": 0.9186,
      "step": 839
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.5355067609705298e-05,
      "loss": 0.9131,
      "step": 840
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.534363154784964e-05,
      "loss": 0.9575,
      "step": 841
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.5332185695921586e-05,
      "loss": 0.8903,
      "step": 842
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.532073007489109e-05,
      "loss": 0.9696,
      "step": 843
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.5309264705746013e-05,
      "loss": 0.9654,
      "step": 844
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.5297789609492062e-05,
      "loss": 0.937,
      "step": 845
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.5286304807152773e-05,
      "loss": 0.9121,
      "step": 846
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.5274810319769465e-05,
      "loss": 0.8926,
      "step": 847
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.52633061684012e-05,
      "loss": 0.9152,
      "step": 848
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5251792374124734e-05,
      "loss": 0.9556,
      "step": 849
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5240268958034512e-05,
      "loss": 0.9124,
      "step": 850
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5228735941242599e-05,
      "loss": 0.9587,
      "step": 851
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5217193344878634e-05,
      "loss": 0.9384,
      "step": 852
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5205641190089822e-05,
      "loss": 0.9734,
      "step": 853
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5194079498040882e-05,
      "loss": 0.9049,
      "step": 854
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5182508289913995e-05,
      "loss": 0.9122,
      "step": 855
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5170927586908787e-05,
      "loss": 0.9138,
      "step": 856
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.515933741024227e-05,
      "loss": 0.9419,
      "step": 857
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.514773778114882e-05,
      "loss": 0.949,
      "step": 858
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5136128720880131e-05,
      "loss": 0.9616,
      "step": 859
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5124510250705169e-05,
      "loss": 0.9173,
      "step": 860
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5112882391910147e-05,
      "loss": 0.9747,
      "step": 861
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5101245165798477e-05,
      "loss": 0.9773,
      "step": 862
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5089598593690726e-05,
      "loss": 0.9404,
      "step": 863
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5077942696924597e-05,
      "loss": 0.9596,
      "step": 864
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5066277496854864e-05,
      "loss": 0.9776,
      "step": 865
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5054603014853353e-05,
      "loss": 0.9325,
      "step": 866
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5042919272308895e-05,
      "loss": 0.9594,
      "step": 867
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5031226290627286e-05,
      "loss": 0.978,
      "step": 868
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5019524091231245e-05,
      "loss": 0.9305,
      "step": 869
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5007812695560385e-05,
      "loss": 0.9424,
      "step": 870
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.4996092125071162e-05,
      "loss": 0.9237,
      "step": 871
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.4984362401236854e-05,
      "loss": 0.9229,
      "step": 872
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4972623545547486e-05,
      "loss": 0.9196,
      "step": 873
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4960875579509835e-05,
      "loss": 0.941,
      "step": 874
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4949118524647359e-05,
      "loss": 0.9388,
      "step": 875
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4937352402500169e-05,
      "loss": 0.9838,
      "step": 876
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4925577234624986e-05,
      "loss": 0.9745,
      "step": 877
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4913793042595109e-05,
      "loss": 0.9327,
      "step": 878
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4901999848000365e-05,
      "loss": 0.9205,
      "step": 879
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4890197672447082e-05,
      "loss": 0.944,
      "step": 880
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4878386537558032e-05,
      "loss": 0.9392,
      "step": 881
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4866566464972407e-05,
      "loss": 0.9546,
      "step": 882
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4854737476345778e-05,
      "loss": 0.988,
      "step": 883
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.484289959335004e-05,
      "loss": 0.9747,
      "step": 884
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.483105283767339e-05,
      "loss": 0.942,
      "step": 885
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4819197231020282e-05,
      "loss": 0.9184,
      "step": 886
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4807332795111387e-05,
      "loss": 0.9173,
      "step": 887
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4795459551683542e-05,
      "loss": 0.9564,
      "step": 888
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4783577522489733e-05,
      "loss": 0.9674,
      "step": 889
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4771686729299038e-05,
      "loss": 0.9651,
      "step": 890
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4759787193896587e-05,
      "loss": 0.9137,
      "step": 891
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4747878938083532e-05,
      "loss": 0.9492,
      "step": 892
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.4735961983676998e-05,
      "loss": 0.9204,
      "step": 893
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.472403635251005e-05,
      "loss": 0.9601,
      "step": 894
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.471210206643165e-05,
      "loss": 0.9068,
      "step": 895
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.470015914730661e-05,
      "loss": 0.9577,
      "step": 896
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.4688207617015564e-05,
      "loss": 0.9256,
      "step": 897
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.4676247497454925e-05,
      "loss": 0.928,
      "step": 898
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.4664278810536835e-05,
      "loss": 0.9735,
      "step": 899
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.4652301578189141e-05,
      "loss": 0.9352,
      "step": 900
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.4640315822355342e-05,
      "loss": 0.9198,
      "step": 901
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.462832156499455e-05,
      "loss": 0.9596,
      "step": 902
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.4616318828081454e-05,
      "loss": 0.9301,
      "step": 903
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.4604307633606283e-05,
      "loss": 0.9351,
      "step": 904
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.4592288003574757e-05,
      "loss": 0.9256,
      "step": 905
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.458025996000805e-05,
      "loss": 0.9385,
      "step": 906
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.4568223524942754e-05,
      "loss": 0.9891,
      "step": 907
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.4556178720430828e-05,
      "loss": 0.9621,
      "step": 908
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.4544125568539573e-05,
      "loss": 0.9482,
      "step": 909
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.4532064091351583e-05,
      "loss": 0.9648,
      "step": 910
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.4519994310964697e-05,
      "loss": 0.9577,
      "step": 911
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.4507916249491975e-05,
      "loss": 0.9347,
      "step": 912
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.4495829929061644e-05,
      "loss": 0.923,
      "step": 913
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.4483735371817061e-05,
      "loss": 0.9556,
      "step": 914
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.4471632599916677e-05,
      "loss": 0.9115,
      "step": 915
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.445952163553399e-05,
      "loss": 0.9036,
      "step": 916
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.4447402500857513e-05,
      "loss": 1.0009,
      "step": 917
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.4435275218090718e-05,
      "loss": 0.9334,
      "step": 918
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.4423139809452016e-05,
      "loss": 0.9628,
      "step": 919
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.4410996297174702e-05,
      "loss": 0.9312,
      "step": 920
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.4398844703506911e-05,
      "loss": 0.9492,
      "step": 921
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.4386685050711593e-05,
      "loss": 0.9992,
      "step": 922
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.437451736106646e-05,
      "loss": 0.9552,
      "step": 923
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.4362341656863946e-05,
      "loss": 0.9427,
      "step": 924
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.4350157960411168e-05,
      "loss": 0.9297,
      "step": 925
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.4337966294029892e-05,
      "loss": 0.9348,
      "step": 926
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.4325766680056481e-05,
      "loss": 0.8734,
      "step": 927
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.4313559140841854e-05,
      "loss": 0.9382,
      "step": 928
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.430134369875146e-05,
      "loss": 0.952,
      "step": 929
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.4289120376165223e-05,
      "loss": 0.9555,
      "step": 930
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.4276889195477498e-05,
      "loss": 0.9553,
      "step": 931
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.4264650179097049e-05,
      "loss": 0.9087,
      "step": 932
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.4252403349446986e-05,
      "loss": 0.9454,
      "step": 933
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.4240148728964737e-05,
      "loss": 0.9445,
      "step": 934
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.4227886340102005e-05,
      "loss": 0.9277,
      "step": 935
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.4215616205324725e-05,
      "loss": 0.9167,
      "step": 936
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.420333834711302e-05,
      "loss": 0.9798,
      "step": 937
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.4191052787961168e-05,
      "loss": 0.9613,
      "step": 938
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.4178759550377552e-05,
      "loss": 0.9232,
      "step": 939
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.4166458656884626e-05,
      "loss": 0.9418,
      "step": 940
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.4154150130018867e-05,
      "loss": 0.9418,
      "step": 941
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.4141833992330735e-05,
      "loss": 0.9733,
      "step": 942
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.4129510266384645e-05,
      "loss": 0.9622,
      "step": 943
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.4117178974758903e-05,
      "loss": 0.9238,
      "step": 944
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.4104840140045678e-05,
      "loss": 0.9648,
      "step": 945
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.4092493784850966e-05,
      "loss": 0.9498,
      "step": 946
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.4080139931794532e-05,
      "loss": 0.938,
      "step": 947
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.4067778603509881e-05,
      "loss": 0.958,
      "step": 948
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.4055409822644216e-05,
      "loss": 0.9097,
      "step": 949
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.4043033611858394e-05,
      "loss": 0.945,
      "step": 950
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.403064999382688e-05,
      "loss": 0.9529,
      "step": 951
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.4018258991237712e-05,
      "loss": 0.9774,
      "step": 952
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.400586062679246e-05,
      "loss": 0.9422,
      "step": 953
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.3993454923206176e-05,
      "loss": 0.9396,
      "step": 954
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.3981041903207364e-05,
      "loss": 0.9664,
      "step": 955
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.3968621589537928e-05,
      "loss": 0.9662,
      "step": 956
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.3956194004953134e-05,
      "loss": 0.9303,
      "step": 957
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.3943759172221578e-05,
      "loss": 0.9351,
      "step": 958
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.3931317114125123e-05,
      "loss": 0.9723,
      "step": 959
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.3918867853458878e-05,
      "loss": 0.9515,
      "step": 960
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.3906411413031142e-05,
      "loss": 0.926,
      "step": 961
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.3893947815663378e-05,
      "loss": 0.973,
      "step": 962
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.3881477084190146e-05,
      "loss": 0.9544,
      "step": 963
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.3868999241459094e-05,
      "loss": 0.9218,
      "step": 964
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.3856514310330881e-05,
      "loss": 0.9498,
      "step": 965
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.3844022313679167e-05,
      "loss": 0.9422,
      "step": 966
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.3831523274390546e-05,
      "loss": 0.9281,
      "step": 967
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.3819017215364523e-05,
      "loss": 0.9445,
      "step": 968
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.3806504159513461e-05,
      "loss": 0.9596,
      "step": 969
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.3793984129762539e-05,
      "loss": 0.9308,
      "step": 970
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.3781457149049715e-05,
      "loss": 0.9762,
      "step": 971
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.3768923240325684e-05,
      "loss": 0.9731,
      "step": 972
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.3756382426553834e-05,
      "loss": 0.9925,
      "step": 973
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.3743834730710197e-05,
      "loss": 0.9314,
      "step": 974
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.3731280175783423e-05,
      "loss": 0.906,
      "step": 975
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.371871878477472e-05,
      "loss": 0.9898,
      "step": 976
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.3706150580697826e-05,
      "loss": 0.927,
      "step": 977
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.3693575586578958e-05,
      "loss": 0.9463,
      "step": 978
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.368099382545678e-05,
      "loss": 0.9362,
      "step": 979
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.3668405320382341e-05,
      "loss": 0.9139,
      "step": 980
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.3655810094419057e-05,
      "loss": 0.9287,
      "step": 981
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.3643208170642655e-05,
      "loss": 0.957,
      "step": 982
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.3630599572141129e-05,
      "loss": 0.985,
      "step": 983
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.3617984322014702e-05,
      "loss": 0.9467,
      "step": 984
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.360536244337579e-05,
      "loss": 0.9695,
      "step": 985
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.3592733959348945e-05,
      "loss": 0.9293,
      "step": 986
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.3580098893070828e-05,
      "loss": 0.9161,
      "step": 987
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.3567457267690152e-05,
      "loss": 0.9154,
      "step": 988
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.3554809106367654e-05,
      "loss": 0.9385,
      "step": 989
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.3542154432276044e-05,
      "loss": 0.9654,
      "step": 990
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.3529493268599957e-05,
      "loss": 0.9407,
      "step": 991
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.3516825638535927e-05,
      "loss": 0.9126,
      "step": 992
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.3504151565292328e-05,
      "loss": 0.93,
      "step": 993
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.3491471072089345e-05,
      "loss": 0.9169,
      "step": 994
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.3478784182158918e-05,
      "loss": 0.9703,
      "step": 995
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.346609091874471e-05,
      "loss": 0.9643,
      "step": 996
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.3453391305102061e-05,
      "loss": 0.9612,
      "step": 997
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.344068536449795e-05,
      "loss": 0.9157,
      "step": 998
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.342797312021094e-05,
      "loss": 0.9486,
      "step": 999
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.3415254595531141e-05,
      "loss": 0.943,
      "step": 1000
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.3402529813760185e-05,
      "loss": 0.955,
      "step": 1001
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.3389798798211147e-05,
      "loss": 0.9201,
      "step": 1002
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.3377061572208535e-05,
      "loss": 0.9031,
      "step": 1003
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.336431815908823e-05,
      "loss": 0.9227,
      "step": 1004
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.3351568582197456e-05,
      "loss": 0.9771,
      "step": 1005
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.3338812864894725e-05,
      "loss": 0.9224,
      "step": 1006
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.3326051030549794e-05,
      "loss": 0.9559,
      "step": 1007
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.3313283102543637e-05,
      "loss": 0.9619,
      "step": 1008
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.3300509104268382e-05,
      "loss": 0.9359,
      "step": 1009
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3287729059127288e-05,
      "loss": 0.9497,
      "step": 1010
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3274942990534685e-05,
      "loss": 0.9449,
      "step": 1011
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3262150921915943e-05,
      "loss": 0.9595,
      "step": 1012
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3249352876707423e-05,
      "loss": 0.9229,
      "step": 1013
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3236548878356433e-05,
      "loss": 0.9175,
      "step": 1014
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3223738950321189e-05,
      "loss": 0.9103,
      "step": 1015
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.321092311607078e-05,
      "loss": 0.9541,
      "step": 1016
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3198101399085097e-05,
      "loss": 0.9547,
      "step": 1017
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.3185273822854832e-05,
      "loss": 0.9689,
      "step": 1018
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.317244041088139e-05,
      "loss": 0.9035,
      "step": 1019
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.3159601186676883e-05,
      "loss": 0.9191,
      "step": 1020
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.3146756173764061e-05,
      "loss": 0.9181,
      "step": 1021
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.313390539567629e-05,
      "loss": 0.9244,
      "step": 1022
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.312104887595749e-05,
      "loss": 0.9206,
      "step": 1023
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.31081866381621e-05,
      "loss": 0.9739,
      "step": 1024
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.3095318705855041e-05,
      "loss": 0.9901,
      "step": 1025
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.308244510261166e-05,
      "loss": 0.917,
      "step": 1026
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.3069565852017703e-05,
      "loss": 0.9329,
      "step": 1027
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.3056680977669253e-05,
      "loss": 0.992,
      "step": 1028
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.3043790503172707e-05,
      "loss": 0.9517,
      "step": 1029
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.3030894452144706e-05,
      "loss": 0.937,
      "step": 1030
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.3017992848212127e-05,
      "loss": 0.9621,
      "step": 1031
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.3005085715012003e-05,
      "loss": 0.9391,
      "step": 1032
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.2992173076191508e-05,
      "loss": 0.9536,
      "step": 1033
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.2979254955407901e-05,
      "loss": 0.9609,
      "step": 1034
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.296633137632848e-05,
      "loss": 0.9242,
      "step": 1035
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.295340236263055e-05,
      "loss": 0.9093,
      "step": 1036
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.2940467938001362e-05,
      "loss": 0.9441,
      "step": 1037
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.2927528126138098e-05,
      "loss": 0.9759,
      "step": 1038
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.2914582950747794e-05,
      "loss": 0.9725,
      "step": 1039
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.2901632435547317e-05,
      "loss": 0.9798,
      "step": 1040
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.2888676604263322e-05,
      "loss": 0.9388,
      "step": 1041
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.2875715480632195e-05,
      "loss": 0.956,
      "step": 1042
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.2862749088400026e-05,
      "loss": 0.9585,
      "step": 1043
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.2849777451322553e-05,
      "loss": 0.962,
      "step": 1044
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.2836800593165127e-05,
      "loss": 0.9633,
      "step": 1045
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.282381853770266e-05,
      "loss": 0.9084,
      "step": 1046
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.2810831308719585e-05,
      "loss": 0.9546,
      "step": 1047
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.279783893000982e-05,
      "loss": 0.9687,
      "step": 1048
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.2784841425376714e-05,
      "loss": 0.9251,
      "step": 1049
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.2771838818633011e-05,
      "loss": 0.9513,
      "step": 1050
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.2758831133600793e-05,
      "loss": 0.9074,
      "step": 1051
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.2745818394111456e-05,
      "loss": 0.9565,
      "step": 1052
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.2732800624005648e-05,
      "loss": 0.9309,
      "step": 1053
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.2719777847133241e-05,
      "loss": 0.9366,
      "step": 1054
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.270675008735328e-05,
      "loss": 0.9491,
      "step": 1055
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.2693717368533929e-05,
      "loss": 0.9966,
      "step": 1056
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.2680679714552449e-05,
      "loss": 0.9446,
      "step": 1057
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.2667637149295136e-05,
      "loss": 0.9468,
      "step": 1058
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.265458969665729e-05,
      "loss": 0.9313,
      "step": 1059
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.2641537380543154e-05,
      "loss": 0.9777,
      "step": 1060
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.2628480224865895e-05,
      "loss": 0.9639,
      "step": 1061
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.261541825354754e-05,
      "loss": 0.9194,
      "step": 1062
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.2602351490518935e-05,
      "loss": 0.9443,
      "step": 1063
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.2589279959719711e-05,
      "loss": 0.9391,
      "step": 1064
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.2576203685098233e-05,
      "loss": 0.9446,
      "step": 1065
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.2563122690611556e-05,
      "loss": 0.9241,
      "step": 1066
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.2550037000225382e-05,
      "loss": 0.9255,
      "step": 1067
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.2536946637914014e-05,
      "loss": 0.9284,
      "step": 1068
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.2523851627660318e-05,
      "loss": 0.9558,
      "step": 1069
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.2510751993455679e-05,
      "loss": 0.9833,
      "step": 1070
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.2497647759299945e-05,
      "loss": 0.9481,
      "step": 1071
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.2484538949201395e-05,
      "loss": 0.9486,
      "step": 1072
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.2471425587176696e-05,
      "loss": 0.963,
      "step": 1073
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.2458307697250847e-05,
      "loss": 0.9396,
      "step": 1074
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.2445185303457151e-05,
      "loss": 0.9461,
      "step": 1075
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.2432058429837153e-05,
      "loss": 0.9815,
      "step": 1076
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.2418927100440614e-05,
      "loss": 0.9273,
      "step": 1077
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.2405791339325455e-05,
      "loss": 0.9333,
      "step": 1078
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.2392651170557714e-05,
      "loss": 0.9351,
      "step": 1079
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.237950661821151e-05,
      "loss": 0.9746,
      "step": 1080
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.2366357706368987e-05,
      "loss": 0.9626,
      "step": 1081
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.2353204459120279e-05,
      "loss": 0.9384,
      "step": 1082
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.2340046900563462e-05,
      "loss": 0.9122,
      "step": 1083
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.2326885054804517e-05,
      "loss": 0.9603,
      "step": 1084
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.2313718945957264e-05,
      "loss": 0.9683,
      "step": 1085
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.230054859814335e-05,
      "loss": 0.9576,
      "step": 1086
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.2287374035492184e-05,
      "loss": 0.9815,
      "step": 1087
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.2274195282140891e-05,
      "loss": 0.9346,
      "step": 1088
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.2261012362234275e-05,
      "loss": 0.9332,
      "step": 1089
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.224782529992478e-05,
      "loss": 0.9414,
      "step": 1090
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.2234634119372436e-05,
      "loss": 0.9822,
      "step": 1091
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.2221438844744813e-05,
      "loss": 0.9624,
      "step": 1092
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.220823950021699e-05,
      "loss": 0.9367,
      "step": 1093
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.2195036109971495e-05,
      "loss": 0.9422,
      "step": 1094
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.2181828698198275e-05,
      "loss": 0.9957,
      "step": 1095
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.2168617289094637e-05,
      "loss": 0.9845,
      "step": 1096
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.2155401906865225e-05,
      "loss": 0.9378,
      "step": 1097
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.2142182575721946e-05,
      "loss": 0.9269,
      "step": 1098
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.212895931988395e-05,
      "loss": 1.0148,
      "step": 1099
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.2115732163577579e-05,
      "loss": 0.9149,
      "step": 1100
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.2102501131036313e-05,
      "loss": 0.9516,
      "step": 1101
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.2089266246500743e-05,
      "loss": 0.9408,
      "step": 1102
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.2076027534218518e-05,
      "loss": 0.9626,
      "step": 1103
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.2062785018444288e-05,
      "loss": 0.9191,
      "step": 1104
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.2049538723439679e-05,
      "loss": 0.9625,
      "step": 1105
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.2036288673473243e-05,
      "loss": 0.931,
      "step": 1106
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.2023034892820409e-05,
      "loss": 0.9307,
      "step": 1107
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.2009777405763446e-05,
      "loss": 0.9469,
      "step": 1108
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.1996516236591398e-05,
      "loss": 0.948,
      "step": 1109
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.1983251409600073e-05,
      "loss": 0.9153,
      "step": 1110
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.1969982949091969e-05,
      "loss": 0.8975,
      "step": 1111
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.1956710879376249e-05,
      "loss": 0.9811,
      "step": 1112
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.1943435224768682e-05,
      "loss": 0.932,
      "step": 1113
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.1930156009591607e-05,
      "loss": 0.981,
      "step": 1114
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.1916873258173886e-05,
      "loss": 0.9405,
      "step": 1115
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.1903586994850865e-05,
      "loss": 0.9703,
      "step": 1116
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.1890297243964315e-05,
      "loss": 0.9079,
      "step": 1117
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.18770040298624e-05,
      "loss": 1.0058,
      "step": 1118
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.1863707376899638e-05,
      "loss": 0.9228,
      "step": 1119
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.1850407309436831e-05,
      "loss": 0.9599,
      "step": 1120
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.1837103851841047e-05,
      "loss": 0.9113,
      "step": 1121
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.1823797028485566e-05,
      "loss": 0.9178,
      "step": 1122
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.1810486863749828e-05,
      "loss": 0.9359,
      "step": 1123
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.1797173382019402e-05,
      "loss": 0.9497,
      "step": 1124
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.1783856607685926e-05,
      "loss": 0.9465,
      "step": 1125
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.177053656514708e-05,
      "loss": 0.9799,
      "step": 1126
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.1757213278806523e-05,
      "loss": 0.9439,
      "step": 1127
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.174388677307386e-05,
      "loss": 0.9392,
      "step": 1128
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.1730557072364601e-05,
      "loss": 0.9193,
      "step": 1129
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.1717224201100095e-05,
      "loss": 0.9771,
      "step": 1130
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.1703888183707513e-05,
      "loss": 0.9426,
      "step": 1131
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.1690549044619781e-05,
      "loss": 0.9638,
      "step": 1132
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.1677206808275552e-05,
      "loss": 0.9796,
      "step": 1133
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.1663861499119148e-05,
      "loss": 1.0189,
      "step": 1134
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.165051314160052e-05,
      "loss": 0.9612,
      "step": 1135
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.1637161760175215e-05,
      "loss": 0.9325,
      "step": 1136
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.1623807379304304e-05,
      "loss": 0.9791,
      "step": 1137
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.1610450023454363e-05,
      "loss": 0.9616,
      "step": 1138
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.1597089717097413e-05,
      "loss": 0.9865,
      "step": 1139
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.1583726484710892e-05,
      "loss": 0.9618,
      "step": 1140
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.1570360350777583e-05,
      "loss": 0.9592,
      "step": 1141
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.1556991339785595e-05,
      "loss": 0.9343,
      "step": 1142
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.1543619476228306e-05,
      "loss": 0.9666,
      "step": 1143
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.1530244784604319e-05,
      "loss": 0.925,
      "step": 1144
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.1516867289417415e-05,
      "loss": 0.9951,
      "step": 1145
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.1503487015176527e-05,
      "loss": 1.0007,
      "step": 1146
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.1490103986395658e-05,
      "loss": 0.9597,
      "step": 1147
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.147671822759387e-05,
      "loss": 0.918,
      "step": 1148
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.1463329763295228e-05,
      "loss": 0.8704,
      "step": 1149
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.1449938618028746e-05,
      "loss": 0.931,
      "step": 1150
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1436544816328357e-05,
      "loss": 0.98,
      "step": 1151
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1423148382732854e-05,
      "loss": 0.9324,
      "step": 1152
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1409749341785859e-05,
      "loss": 0.9296,
      "step": 1153
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1396347718035764e-05,
      "loss": 0.9811,
      "step": 1154
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.1382943536035695e-05,
      "loss": 0.9389,
      "step": 1155
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.1369536820343475e-05,
      "loss": 0.9618,
      "step": 1156
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.1356127595521555e-05,
      "loss": 1.0006,
      "step": 1157
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.1342715886136987e-05,
      "loss": 0.9288,
      "step": 1158
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.132930171676138e-05,
      "loss": 0.9351,
      "step": 1159
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.1315885111970844e-05,
      "loss": 0.9691,
      "step": 1160
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.1302466096345955e-05,
      "loss": 0.9417,
      "step": 1161
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.1289044694471705e-05,
      "loss": 0.9011,
      "step": 1162
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.1275620930937452e-05,
      "loss": 0.9757,
      "step": 1163
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.1262194830336888e-05,
      "loss": 0.9197,
      "step": 1164
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.1248766417267985e-05,
      "loss": 0.9637,
      "step": 1165
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.1235335716332953e-05,
      "loss": 0.9324,
      "step": 1166
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1221902752138192e-05,
      "loss": 0.9551,
      "step": 1167
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1208467549294247e-05,
      "loss": 0.9174,
      "step": 1168
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1195030132415765e-05,
      "loss": 1.0413,
      "step": 1169
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.1181590526121451e-05,
      "loss": 0.9601,
      "step": 1170
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.116814875503402e-05,
      "loss": 0.9581,
      "step": 1171
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.1154704843780154e-05,
      "loss": 0.967,
      "step": 1172
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.1141258816990457e-05,
      "loss": 0.9633,
      "step": 1173
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.11278106992994e-05,
      "loss": 0.9551,
      "step": 1174
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.1114360515345301e-05,
      "loss": 0.9311,
      "step": 1175
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.1100908289770253e-05,
      "loss": 0.9303,
      "step": 1176
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.1087454047220087e-05,
      "loss": 0.95,
      "step": 1177
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.1073997812344337e-05,
      "loss": 0.9403,
      "step": 1178
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.1060539609796183e-05,
      "loss": 0.9147,
      "step": 1179
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.1047079464232411e-05,
      "loss": 0.9559,
      "step": 1180
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.1033617400313363e-05,
      "loss": 0.9911,
      "step": 1181
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.1020153442702903e-05,
      "loss": 0.96,
      "step": 1182
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.100668761606836e-05,
      "loss": 0.9994,
      "step": 1183
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.0993219945080486e-05,
      "loss": 0.9692,
      "step": 1184
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.0979750454413414e-05,
      "loss": 0.9768,
      "step": 1185
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.096627916874461e-05,
      "loss": 0.9499,
      "step": 1186
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.095280611275483e-05,
      "loss": 0.9779,
      "step": 1187
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.0939331311128073e-05,
      "loss": 0.9592,
      "step": 1188
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.0925854788551537e-05,
      "loss": 0.9215,
      "step": 1189
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.091237656971557e-05,
      "loss": 0.9613,
      "step": 1190
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.0898896679313628e-05,
      "loss": 0.9374,
      "step": 1191
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0885415142042235e-05,
      "loss": 0.9616,
      "step": 1192
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0871931982600923e-05,
      "loss": 0.9484,
      "step": 1193
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0858447225692203e-05,
      "loss": 0.9428,
      "step": 1194
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0844960896021515e-05,
      "loss": 0.9606,
      "step": 1195
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.083147301829717e-05,
      "loss": 0.9394,
      "step": 1196
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.0817983617230326e-05,
      "loss": 0.9377,
      "step": 1197
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.0804492717534924e-05,
      "loss": 0.9666,
      "step": 1198
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.0791000343927658e-05,
      "loss": 0.9139,
      "step": 1199
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.0777506521127915e-05,
      "loss": 0.9281,
      "step": 1200
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.0764011273857741e-05,
      "loss": 0.9415,
      "step": 1201
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.0750514626841794e-05,
      "loss": 0.9426,
      "step": 1202
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.073701660480729e-05,
      "loss": 0.9313,
      "step": 1203
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.072351723248397e-05,
      "loss": 0.96,
      "step": 1204
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.0710016534604046e-05,
      "loss": 0.9681,
      "step": 1205
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.069651453590216e-05,
      "loss": 0.9302,
      "step": 1206
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.0683011261115337e-05,
      "loss": 0.9264,
      "step": 1207
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.066950673498294e-05,
      "loss": 0.9616,
      "step": 1208
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.065600098224662e-05,
      "loss": 0.9592,
      "step": 1209
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.0642494027650283e-05,
      "loss": 0.67,
      "step": 1210
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.0628985895940032e-05,
      "loss": 0.6031,
      "step": 1211
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.0615476611864129e-05,
      "loss": 0.586,
      "step": 1212
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.0601966200172943e-05,
      "loss": 0.6238,
      "step": 1213
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.0588454685618916e-05,
      "loss": 0.5547,
      "step": 1214
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.0574942092956503e-05,
      "loss": 0.5582,
      "step": 1215
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.056142844694214e-05,
      "loss": 0.5402,
      "step": 1216
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.0547913772334192e-05,
      "loss": 0.5617,
      "step": 1217
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.0534398093892906e-05,
      "loss": 0.6016,
      "step": 1218
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.0520881436380366e-05,
      "loss": 0.6178,
      "step": 1219
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.0507363824560458e-05,
      "loss": 0.5514,
      "step": 1220
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.0493845283198804e-05,
      "loss": 0.5955,
      "step": 1221
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.0480325837062746e-05,
      "loss": 0.5633,
      "step": 1222
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.0466805510921265e-05,
      "loss": 0.548,
      "step": 1223
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.0453284329544968e-05,
      "loss": 0.5857,
      "step": 1224
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.0439762317706028e-05,
      "loss": 0.5543,
      "step": 1225
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.0426239500178128e-05,
      "loss": 0.5735,
      "step": 1226
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.041271590173644e-05,
      "loss": 0.5529,
      "step": 1227
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.0399191547157555e-05,
      "loss": 0.5552,
      "step": 1228
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.0385666461219463e-05,
      "loss": 0.6182,
      "step": 1229
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.0372140668701483e-05,
      "loss": 0.5467,
      "step": 1230
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.0358614194384233e-05,
      "loss": 0.5419,
      "step": 1231
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.0345087063049582e-05,
      "loss": 0.551,
      "step": 1232
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.0331559299480593e-05,
      "loss": 0.5604,
      "step": 1233
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.03180309284615e-05,
      "loss": 0.5121,
      "step": 1234
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.0304501974777646e-05,
      "loss": 0.5738,
      "step": 1235
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.0290972463215436e-05,
      "loss": 0.5268,
      "step": 1236
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.0277442418562302e-05,
      "loss": 0.5737,
      "step": 1237
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.026391186560665e-05,
      "loss": 0.5441,
      "step": 1238
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.0250380829137822e-05,
      "loss": 0.4986,
      "step": 1239
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.023684933394604e-05,
      "loss": 0.5316,
      "step": 1240
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.022331740482237e-05,
      "loss": 0.5645,
      "step": 1241
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.0209785066558672e-05,
      "loss": 0.5658,
      "step": 1242
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.0196252343947555e-05,
      "loss": 0.5651,
      "step": 1243
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.0182719261782333e-05,
      "loss": 0.5345,
      "step": 1244
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.0169185844856982e-05,
      "loss": 0.565,
      "step": 1245
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.0155652117966084e-05,
      "loss": 0.5432,
      "step": 1246
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.0142118105904795e-05,
      "loss": 0.5632,
      "step": 1247
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.012858383346879e-05,
      "loss": 0.5806,
      "step": 1248
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.011504932545422e-05,
      "loss": 0.5695,
      "step": 1249
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.0101514606657675e-05,
      "loss": 0.5319,
      "step": 1250
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.0087979701876123e-05,
      "loss": 0.5374,
      "step": 1251
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.0074444635906875e-05,
      "loss": 0.5871,
      "step": 1252
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.0060909433547536e-05,
      "loss": 0.5658,
      "step": 1253
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.0047374119595972e-05,
      "loss": 0.5422,
      "step": 1254
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.0033838718850238e-05,
      "loss": 0.5202,
      "step": 1255
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.0020303256108554e-05,
      "loss": 0.5818,
      "step": 1256
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.0006767756169255e-05,
      "loss": 0.5888,
      "step": 1257
    },
    {
      "epoch": 3.12,
      "learning_rate": 9.993232243830747e-06,
      "loss": 0.5577,
      "step": 1258
    },
    {
      "epoch": 3.12,
      "learning_rate": 9.979696743891449e-06,
      "loss": 0.5353,
      "step": 1259
    },
    {
      "epoch": 3.13,
      "learning_rate": 9.966161281149766e-06,
      "loss": 0.5595,
      "step": 1260
    },
    {
      "epoch": 3.13,
      "learning_rate": 9.952625880404031e-06,
      "loss": 0.5486,
      "step": 1261
    },
    {
      "epoch": 3.13,
      "learning_rate": 9.939090566452466e-06,
      "loss": 0.5114,
      "step": 1262
    },
    {
      "epoch": 3.13,
      "learning_rate": 9.92555536409313e-06,
      "loss": 0.5599,
      "step": 1263
    },
    {
      "epoch": 3.14,
      "learning_rate": 9.912020298123884e-06,
      "loss": 0.5412,
      "step": 1264
    },
    {
      "epoch": 3.14,
      "learning_rate": 9.89848539334233e-06,
      "loss": 0.5131,
      "step": 1265
    },
    {
      "epoch": 3.14,
      "learning_rate": 9.884950674545785e-06,
      "loss": 0.5732,
      "step": 1266
    },
    {
      "epoch": 3.14,
      "learning_rate": 9.871416166531216e-06,
      "loss": 0.5418,
      "step": 1267
    },
    {
      "epoch": 3.15,
      "learning_rate": 9.857881894095208e-06,
      "loss": 0.5254,
      "step": 1268
    },
    {
      "epoch": 3.15,
      "learning_rate": 9.844347882033917e-06,
      "loss": 0.5644,
      "step": 1269
    },
    {
      "epoch": 3.15,
      "learning_rate": 9.83081415514302e-06,
      "loss": 0.5732,
      "step": 1270
    },
    {
      "epoch": 3.15,
      "learning_rate": 9.817280738217667e-06,
      "loss": 0.5275,
      "step": 1271
    },
    {
      "epoch": 3.16,
      "learning_rate": 9.803747656052446e-06,
      "loss": 0.5464,
      "step": 1272
    },
    {
      "epoch": 3.16,
      "learning_rate": 9.790214933441331e-06,
      "loss": 0.5372,
      "step": 1273
    },
    {
      "epoch": 3.16,
      "learning_rate": 9.776682595177633e-06,
      "loss": 0.5706,
      "step": 1274
    },
    {
      "epoch": 3.16,
      "learning_rate": 9.763150666053961e-06,
      "loss": 0.5683,
      "step": 1275
    },
    {
      "epoch": 3.17,
      "learning_rate": 9.749619170862181e-06,
      "loss": 0.5291,
      "step": 1276
    },
    {
      "epoch": 3.17,
      "learning_rate": 9.736088134393351e-06,
      "loss": 0.5746,
      "step": 1277
    },
    {
      "epoch": 3.17,
      "learning_rate": 9.722557581437701e-06,
      "loss": 0.5661,
      "step": 1278
    },
    {
      "epoch": 3.17,
      "learning_rate": 9.709027536784566e-06,
      "loss": 0.5464,
      "step": 1279
    },
    {
      "epoch": 3.18,
      "learning_rate": 9.695498025222358e-06,
      "loss": 0.569,
      "step": 1280
    },
    {
      "epoch": 3.18,
      "learning_rate": 9.681969071538502e-06,
      "loss": 0.55,
      "step": 1281
    },
    {
      "epoch": 3.18,
      "learning_rate": 9.66844070051941e-06,
      "loss": 0.5437,
      "step": 1282
    },
    {
      "epoch": 3.18,
      "learning_rate": 9.654912936950423e-06,
      "loss": 0.5561,
      "step": 1283
    },
    {
      "epoch": 3.19,
      "learning_rate": 9.64138580561577e-06,
      "loss": 0.553,
      "step": 1284
    },
    {
      "epoch": 3.19,
      "learning_rate": 9.627859331298522e-06,
      "loss": 0.5647,
      "step": 1285
    },
    {
      "epoch": 3.19,
      "learning_rate": 9.614333538780543e-06,
      "loss": 0.5341,
      "step": 1286
    },
    {
      "epoch": 3.19,
      "learning_rate": 9.60080845284245e-06,
      "loss": 0.5663,
      "step": 1287
    },
    {
      "epoch": 3.2,
      "learning_rate": 9.587284098263566e-06,
      "loss": 0.5354,
      "step": 1288
    },
    {
      "epoch": 3.2,
      "learning_rate": 9.573760499821874e-06,
      "loss": 0.5434,
      "step": 1289
    },
    {
      "epoch": 3.2,
      "learning_rate": 9.560237682293975e-06,
      "loss": 0.5589,
      "step": 1290
    },
    {
      "epoch": 3.2,
      "learning_rate": 9.546715670455032e-06,
      "loss": 0.5288,
      "step": 1291
    },
    {
      "epoch": 3.21,
      "learning_rate": 9.533194489078737e-06,
      "loss": 0.5487,
      "step": 1292
    },
    {
      "epoch": 3.21,
      "learning_rate": 9.519674162937259e-06,
      "loss": 0.5683,
      "step": 1293
    },
    {
      "epoch": 3.21,
      "learning_rate": 9.506154716801198e-06,
      "loss": 0.5441,
      "step": 1294
    },
    {
      "epoch": 3.21,
      "learning_rate": 9.492636175439547e-06,
      "loss": 0.5849,
      "step": 1295
    },
    {
      "epoch": 3.22,
      "learning_rate": 9.479118563619638e-06,
      "loss": 0.5925,
      "step": 1296
    },
    {
      "epoch": 3.22,
      "learning_rate": 9.465601906107098e-06,
      "loss": 0.5574,
      "step": 1297
    },
    {
      "epoch": 3.22,
      "learning_rate": 9.45208622766581e-06,
      "loss": 0.4983,
      "step": 1298
    },
    {
      "epoch": 3.22,
      "learning_rate": 9.438571553057861e-06,
      "loss": 0.5264,
      "step": 1299
    },
    {
      "epoch": 3.23,
      "learning_rate": 9.425057907043498e-06,
      "loss": 0.553,
      "step": 1300
    },
    {
      "epoch": 3.23,
      "learning_rate": 9.411545314381087e-06,
      "loss": 0.5437,
      "step": 1301
    },
    {
      "epoch": 3.23,
      "learning_rate": 9.398033799827062e-06,
      "loss": 0.5295,
      "step": 1302
    },
    {
      "epoch": 3.23,
      "learning_rate": 9.384523388135878e-06,
      "loss": 0.5452,
      "step": 1303
    },
    {
      "epoch": 3.24,
      "learning_rate": 9.371014104059973e-06,
      "loss": 0.5532,
      "step": 1304
    },
    {
      "epoch": 3.24,
      "learning_rate": 9.357505972349724e-06,
      "loss": 0.5465,
      "step": 1305
    },
    {
      "epoch": 3.24,
      "learning_rate": 9.343999017753385e-06,
      "loss": 0.5014,
      "step": 1306
    },
    {
      "epoch": 3.24,
      "learning_rate": 9.330493265017062e-06,
      "loss": 0.5636,
      "step": 1307
    },
    {
      "epoch": 3.25,
      "learning_rate": 9.316988738884663e-06,
      "loss": 0.568,
      "step": 1308
    },
    {
      "epoch": 3.25,
      "learning_rate": 9.30348546409784e-06,
      "loss": 0.5286,
      "step": 1309
    },
    {
      "epoch": 3.25,
      "learning_rate": 9.289983465395955e-06,
      "loss": 0.5225,
      "step": 1310
    },
    {
      "epoch": 3.25,
      "learning_rate": 9.276482767516031e-06,
      "loss": 0.5209,
      "step": 1311
    },
    {
      "epoch": 3.26,
      "learning_rate": 9.262983395192713e-06,
      "loss": 0.5274,
      "step": 1312
    },
    {
      "epoch": 3.26,
      "learning_rate": 9.249485373158209e-06,
      "loss": 0.5324,
      "step": 1313
    },
    {
      "epoch": 3.26,
      "learning_rate": 9.23598872614226e-06,
      "loss": 0.5413,
      "step": 1314
    },
    {
      "epoch": 3.26,
      "learning_rate": 9.222493478872087e-06,
      "loss": 0.5435,
      "step": 1315
    },
    {
      "epoch": 3.27,
      "learning_rate": 9.208999656072344e-06,
      "loss": 0.552,
      "step": 1316
    },
    {
      "epoch": 3.27,
      "learning_rate": 9.195507282465079e-06,
      "loss": 0.5271,
      "step": 1317
    },
    {
      "epoch": 3.27,
      "learning_rate": 9.182016382769678e-06,
      "loss": 0.5708,
      "step": 1318
    },
    {
      "epoch": 3.27,
      "learning_rate": 9.168526981702832e-06,
      "loss": 0.5671,
      "step": 1319
    },
    {
      "epoch": 3.27,
      "learning_rate": 9.15503910397849e-06,
      "loss": 0.574,
      "step": 1320
    },
    {
      "epoch": 3.28,
      "learning_rate": 9.141552774307799e-06,
      "loss": 0.5076,
      "step": 1321
    },
    {
      "epoch": 3.28,
      "learning_rate": 9.12806801739908e-06,
      "loss": 0.5274,
      "step": 1322
    },
    {
      "epoch": 3.28,
      "learning_rate": 9.114584857957768e-06,
      "loss": 0.549,
      "step": 1323
    },
    {
      "epoch": 3.28,
      "learning_rate": 9.101103320686376e-06,
      "loss": 0.5569,
      "step": 1324
    },
    {
      "epoch": 3.29,
      "learning_rate": 9.087623430284435e-06,
      "loss": 0.5386,
      "step": 1325
    },
    {
      "epoch": 3.29,
      "learning_rate": 9.074145211448467e-06,
      "loss": 0.5067,
      "step": 1326
    },
    {
      "epoch": 3.29,
      "learning_rate": 9.060668688871925e-06,
      "loss": 0.5421,
      "step": 1327
    },
    {
      "epoch": 3.29,
      "learning_rate": 9.047193887245169e-06,
      "loss": 0.5469,
      "step": 1328
    },
    {
      "epoch": 3.3,
      "learning_rate": 9.033720831255391e-06,
      "loss": 0.5408,
      "step": 1329
    },
    {
      "epoch": 3.3,
      "learning_rate": 9.020249545586587e-06,
      "loss": 0.5521,
      "step": 1330
    },
    {
      "epoch": 3.3,
      "learning_rate": 9.006780054919516e-06,
      "loss": 0.5509,
      "step": 1331
    },
    {
      "epoch": 3.3,
      "learning_rate": 8.993312383931643e-06,
      "loss": 0.5382,
      "step": 1332
    },
    {
      "epoch": 3.31,
      "learning_rate": 8.979846557297099e-06,
      "loss": 0.5526,
      "step": 1333
    },
    {
      "epoch": 3.31,
      "learning_rate": 8.966382599686638e-06,
      "loss": 0.5867,
      "step": 1334
    },
    {
      "epoch": 3.31,
      "learning_rate": 8.952920535767592e-06,
      "loss": 0.565,
      "step": 1335
    },
    {
      "epoch": 3.31,
      "learning_rate": 8.939460390203818e-06,
      "loss": 0.5311,
      "step": 1336
    },
    {
      "epoch": 3.32,
      "learning_rate": 8.926002187655665e-06,
      "loss": 0.5813,
      "step": 1337
    },
    {
      "epoch": 3.32,
      "learning_rate": 8.912545952779915e-06,
      "loss": 0.5478,
      "step": 1338
    },
    {
      "epoch": 3.32,
      "learning_rate": 8.89909171022975e-06,
      "loss": 0.5788,
      "step": 1339
    },
    {
      "epoch": 3.32,
      "learning_rate": 8.8856394846547e-06,
      "loss": 0.5701,
      "step": 1340
    },
    {
      "epoch": 3.33,
      "learning_rate": 8.872189300700602e-06,
      "loss": 0.5303,
      "step": 1341
    },
    {
      "epoch": 3.33,
      "learning_rate": 8.85874118300955e-06,
      "loss": 0.5429,
      "step": 1342
    },
    {
      "epoch": 3.33,
      "learning_rate": 8.84529515621985e-06,
      "loss": 0.5896,
      "step": 1343
    },
    {
      "epoch": 3.33,
      "learning_rate": 8.831851244965982e-06,
      "loss": 0.5269,
      "step": 1344
    },
    {
      "epoch": 3.34,
      "learning_rate": 8.818409473878554e-06,
      "loss": 0.5372,
      "step": 1345
    },
    {
      "epoch": 3.34,
      "learning_rate": 8.804969867584241e-06,
      "loss": 0.5842,
      "step": 1346
    },
    {
      "epoch": 3.34,
      "learning_rate": 8.791532450705755e-06,
      "loss": 0.5582,
      "step": 1347
    },
    {
      "epoch": 3.34,
      "learning_rate": 8.778097247861808e-06,
      "loss": 0.506,
      "step": 1348
    },
    {
      "epoch": 3.35,
      "learning_rate": 8.764664283667047e-06,
      "loss": 0.5612,
      "step": 1349
    },
    {
      "epoch": 3.35,
      "learning_rate": 8.751233582732015e-06,
      "loss": 0.5905,
      "step": 1350
    },
    {
      "epoch": 3.35,
      "learning_rate": 8.737805169663113e-06,
      "loss": 0.6127,
      "step": 1351
    },
    {
      "epoch": 3.35,
      "learning_rate": 8.724379069062552e-06,
      "loss": 0.5404,
      "step": 1352
    },
    {
      "epoch": 3.36,
      "learning_rate": 8.7109553055283e-06,
      "loss": 0.5745,
      "step": 1353
    },
    {
      "epoch": 3.36,
      "learning_rate": 8.697533903654048e-06,
      "loss": 0.5517,
      "step": 1354
    },
    {
      "epoch": 3.36,
      "learning_rate": 8.684114888029158e-06,
      "loss": 0.5533,
      "step": 1355
    },
    {
      "epoch": 3.36,
      "learning_rate": 8.670698283238623e-06,
      "loss": 0.5618,
      "step": 1356
    },
    {
      "epoch": 3.37,
      "learning_rate": 8.657284113863014e-06,
      "loss": 0.547,
      "step": 1357
    },
    {
      "epoch": 3.37,
      "learning_rate": 8.643872404478448e-06,
      "loss": 0.5156,
      "step": 1358
    },
    {
      "epoch": 3.37,
      "learning_rate": 8.630463179656527e-06,
      "loss": 0.5699,
      "step": 1359
    },
    {
      "epoch": 3.37,
      "learning_rate": 8.617056463964307e-06,
      "loss": 0.5711,
      "step": 1360
    },
    {
      "epoch": 3.38,
      "learning_rate": 8.603652281964241e-06,
      "loss": 0.5604,
      "step": 1361
    },
    {
      "epoch": 3.38,
      "learning_rate": 8.590250658214148e-06,
      "loss": 0.553,
      "step": 1362
    },
    {
      "epoch": 3.38,
      "learning_rate": 8.576851617267151e-06,
      "loss": 0.5645,
      "step": 1363
    },
    {
      "epoch": 3.38,
      "learning_rate": 8.563455183671648e-06,
      "loss": 0.5141,
      "step": 1364
    },
    {
      "epoch": 3.39,
      "learning_rate": 8.550061381971257e-06,
      "loss": 0.5415,
      "step": 1365
    },
    {
      "epoch": 3.39,
      "learning_rate": 8.536670236704772e-06,
      "loss": 0.5289,
      "step": 1366
    },
    {
      "epoch": 3.39,
      "learning_rate": 8.523281772406129e-06,
      "loss": 0.5517,
      "step": 1367
    },
    {
      "epoch": 3.39,
      "learning_rate": 8.509896013604342e-06,
      "loss": 0.5453,
      "step": 1368
    },
    {
      "epoch": 3.4,
      "learning_rate": 8.496512984823475e-06,
      "loss": 0.5289,
      "step": 1369
    },
    {
      "epoch": 3.4,
      "learning_rate": 8.483132710582585e-06,
      "loss": 0.5521,
      "step": 1370
    },
    {
      "epoch": 3.4,
      "learning_rate": 8.469755215395686e-06,
      "loss": 0.5493,
      "step": 1371
    },
    {
      "epoch": 3.4,
      "learning_rate": 8.456380523771698e-06,
      "loss": 0.5615,
      "step": 1372
    },
    {
      "epoch": 3.41,
      "learning_rate": 8.443008660214409e-06,
      "loss": 0.5508,
      "step": 1373
    },
    {
      "epoch": 3.41,
      "learning_rate": 8.42963964922242e-06,
      "loss": 0.5263,
      "step": 1374
    },
    {
      "epoch": 3.41,
      "learning_rate": 8.416273515289111e-06,
      "loss": 0.5635,
      "step": 1375
    },
    {
      "epoch": 3.41,
      "learning_rate": 8.402910282902588e-06,
      "loss": 0.579,
      "step": 1376
    },
    {
      "epoch": 3.42,
      "learning_rate": 8.38954997654564e-06,
      "loss": 0.5853,
      "step": 1377
    },
    {
      "epoch": 3.42,
      "learning_rate": 8.3761926206957e-06,
      "loss": 0.5396,
      "step": 1378
    },
    {
      "epoch": 3.42,
      "learning_rate": 8.362838239824788e-06,
      "loss": 0.5704,
      "step": 1379
    },
    {
      "epoch": 3.42,
      "learning_rate": 8.349486858399481e-06,
      "loss": 0.5734,
      "step": 1380
    },
    {
      "epoch": 3.43,
      "learning_rate": 8.336138500880857e-06,
      "loss": 0.5306,
      "step": 1381
    },
    {
      "epoch": 3.43,
      "learning_rate": 8.322793191724453e-06,
      "loss": 0.5512,
      "step": 1382
    },
    {
      "epoch": 3.43,
      "learning_rate": 8.309450955380224e-06,
      "loss": 0.5316,
      "step": 1383
    },
    {
      "epoch": 3.43,
      "learning_rate": 8.296111816292494e-06,
      "loss": 0.5745,
      "step": 1384
    },
    {
      "epoch": 3.44,
      "learning_rate": 8.282775798899908e-06,
      "loss": 0.5602,
      "step": 1385
    },
    {
      "epoch": 3.44,
      "learning_rate": 8.269442927635402e-06,
      "loss": 0.5296,
      "step": 1386
    },
    {
      "epoch": 3.44,
      "learning_rate": 8.25611322692614e-06,
      "loss": 0.5813,
      "step": 1387
    },
    {
      "epoch": 3.44,
      "learning_rate": 8.242786721193479e-06,
      "loss": 0.5375,
      "step": 1388
    },
    {
      "epoch": 3.45,
      "learning_rate": 8.229463434852922e-06,
      "loss": 0.5508,
      "step": 1389
    },
    {
      "epoch": 3.45,
      "learning_rate": 8.216143392314075e-06,
      "loss": 0.5477,
      "step": 1390
    },
    {
      "epoch": 3.45,
      "learning_rate": 8.202826617980602e-06,
      "loss": 0.5327,
      "step": 1391
    },
    {
      "epoch": 3.45,
      "learning_rate": 8.189513136250175e-06,
      "loss": 0.5853,
      "step": 1392
    },
    {
      "epoch": 3.46,
      "learning_rate": 8.176202971514437e-06,
      "loss": 0.5578,
      "step": 1393
    },
    {
      "epoch": 3.46,
      "learning_rate": 8.162896148158954e-06,
      "loss": 0.5412,
      "step": 1394
    },
    {
      "epoch": 3.46,
      "learning_rate": 8.149592690563172e-06,
      "loss": 0.5687,
      "step": 1395
    },
    {
      "epoch": 3.46,
      "learning_rate": 8.136292623100365e-06,
      "loss": 0.5113,
      "step": 1396
    },
    {
      "epoch": 3.47,
      "learning_rate": 8.122995970137602e-06,
      "loss": 0.5578,
      "step": 1397
    },
    {
      "epoch": 3.47,
      "learning_rate": 8.10970275603569e-06,
      "loss": 0.5906,
      "step": 1398
    },
    {
      "epoch": 3.47,
      "learning_rate": 8.096413005149139e-06,
      "loss": 0.5705,
      "step": 1399
    },
    {
      "epoch": 3.47,
      "learning_rate": 8.083126741826116e-06,
      "loss": 0.5656,
      "step": 1400
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.069843990408398e-06,
      "loss": 0.5492,
      "step": 1401
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.056564775231323e-06,
      "loss": 0.5908,
      "step": 1402
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.043289120623754e-06,
      "loss": 0.5521,
      "step": 1403
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.030017050908035e-06,
      "loss": 0.5789,
      "step": 1404
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.016748590399932e-06,
      "loss": 0.5479,
      "step": 1405
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.003483763408604e-06,
      "loss": 0.5264,
      "step": 1406
    },
    {
      "epoch": 3.49,
      "learning_rate": 7.990222594236557e-06,
      "loss": 0.5602,
      "step": 1407
    },
    {
      "epoch": 3.49,
      "learning_rate": 7.976965107179591e-06,
      "loss": 0.5374,
      "step": 1408
    },
    {
      "epoch": 3.5,
      "learning_rate": 7.963711326526757e-06,
      "loss": 0.5351,
      "step": 1409
    },
    {
      "epoch": 3.5,
      "learning_rate": 7.950461276560324e-06,
      "loss": 0.5682,
      "step": 1410
    },
    {
      "epoch": 3.5,
      "learning_rate": 7.937214981555717e-06,
      "loss": 0.5587,
      "step": 1411
    },
    {
      "epoch": 3.5,
      "learning_rate": 7.923972465781487e-06,
      "loss": 0.5592,
      "step": 1412
    },
    {
      "epoch": 3.51,
      "learning_rate": 7.910733753499258e-06,
      "loss": 0.5745,
      "step": 1413
    },
    {
      "epoch": 3.51,
      "learning_rate": 7.89749886896369e-06,
      "loss": 0.5535,
      "step": 1414
    },
    {
      "epoch": 3.51,
      "learning_rate": 7.884267836422426e-06,
      "loss": 0.539,
      "step": 1415
    },
    {
      "epoch": 3.51,
      "learning_rate": 7.871040680116053e-06,
      "loss": 0.5558,
      "step": 1416
    },
    {
      "epoch": 3.52,
      "learning_rate": 7.857817424278056e-06,
      "loss": 0.5303,
      "step": 1417
    },
    {
      "epoch": 3.52,
      "learning_rate": 7.844598093134776e-06,
      "loss": 0.5608,
      "step": 1418
    },
    {
      "epoch": 3.52,
      "learning_rate": 7.831382710905364e-06,
      "loss": 0.5402,
      "step": 1419
    },
    {
      "epoch": 3.52,
      "learning_rate": 7.818171301801728e-06,
      "loss": 0.5287,
      "step": 1420
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.80496389002851e-06,
      "loss": 0.5832,
      "step": 1421
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.791760499783016e-06,
      "loss": 0.5388,
      "step": 1422
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.778561155255192e-06,
      "loss": 0.5511,
      "step": 1423
    },
    {
      "epoch": 3.53,
      "learning_rate": 7.76536588062757e-06,
      "loss": 0.5482,
      "step": 1424
    },
    {
      "epoch": 3.54,
      "learning_rate": 7.75217470007522e-06,
      "loss": 0.5674,
      "step": 1425
    },
    {
      "epoch": 3.54,
      "learning_rate": 7.738987637765725e-06,
      "loss": 0.5577,
      "step": 1426
    },
    {
      "epoch": 3.54,
      "learning_rate": 7.725804717859114e-06,
      "loss": 0.5672,
      "step": 1427
    },
    {
      "epoch": 3.54,
      "learning_rate": 7.712625964507818e-06,
      "loss": 0.5676,
      "step": 1428
    },
    {
      "epoch": 3.55,
      "learning_rate": 7.699451401856651e-06,
      "loss": 0.5649,
      "step": 1429
    },
    {
      "epoch": 3.55,
      "learning_rate": 7.686281054042738e-06,
      "loss": 0.5636,
      "step": 1430
    },
    {
      "epoch": 3.55,
      "learning_rate": 7.673114945195488e-06,
      "loss": 0.5462,
      "step": 1431
    },
    {
      "epoch": 3.55,
      "learning_rate": 7.65995309943654e-06,
      "loss": 0.5294,
      "step": 1432
    },
    {
      "epoch": 3.56,
      "learning_rate": 7.646795540879723e-06,
      "loss": 0.56,
      "step": 1433
    },
    {
      "epoch": 3.56,
      "learning_rate": 7.633642293631016e-06,
      "loss": 0.6007,
      "step": 1434
    },
    {
      "epoch": 3.56,
      "learning_rate": 7.620493381788492e-06,
      "loss": 0.5719,
      "step": 1435
    },
    {
      "epoch": 3.56,
      "learning_rate": 7.607348829442287e-06,
      "loss": 0.5634,
      "step": 1436
    },
    {
      "epoch": 3.57,
      "learning_rate": 7.594208660674547e-06,
      "loss": 0.5554,
      "step": 1437
    },
    {
      "epoch": 3.57,
      "learning_rate": 7.5810728995593875e-06,
      "loss": 0.5295,
      "step": 1438
    },
    {
      "epoch": 3.57,
      "learning_rate": 7.567941570162849e-06,
      "loss": 0.5749,
      "step": 1439
    },
    {
      "epoch": 3.57,
      "learning_rate": 7.554814696542852e-06,
      "loss": 0.5668,
      "step": 1440
    },
    {
      "epoch": 3.58,
      "learning_rate": 7.541692302749157e-06,
      "loss": 0.5483,
      "step": 1441
    },
    {
      "epoch": 3.58,
      "learning_rate": 7.528574412823311e-06,
      "loss": 0.5507,
      "step": 1442
    },
    {
      "epoch": 3.58,
      "learning_rate": 7.51546105079861e-06,
      "loss": 0.5447,
      "step": 1443
    },
    {
      "epoch": 3.58,
      "learning_rate": 7.502352240700062e-06,
      "loss": 0.5579,
      "step": 1444
    },
    {
      "epoch": 3.59,
      "learning_rate": 7.489248006544323e-06,
      "loss": 0.5735,
      "step": 1445
    },
    {
      "epoch": 3.59,
      "learning_rate": 7.476148372339682e-06,
      "loss": 0.5624,
      "step": 1446
    },
    {
      "epoch": 3.59,
      "learning_rate": 7.463053362085988e-06,
      "loss": 0.5461,
      "step": 1447
    },
    {
      "epoch": 3.59,
      "learning_rate": 7.44996299977462e-06,
      "loss": 0.5879,
      "step": 1448
    },
    {
      "epoch": 3.59,
      "learning_rate": 7.436877309388446e-06,
      "loss": 0.531,
      "step": 1449
    },
    {
      "epoch": 3.6,
      "learning_rate": 7.423796314901769e-06,
      "loss": 0.5535,
      "step": 1450
    },
    {
      "epoch": 3.6,
      "learning_rate": 7.410720040280291e-06,
      "loss": 0.5713,
      "step": 1451
    },
    {
      "epoch": 3.6,
      "learning_rate": 7.397648509481068e-06,
      "loss": 0.5821,
      "step": 1452
    },
    {
      "epoch": 3.6,
      "learning_rate": 7.384581746452464e-06,
      "loss": 0.5383,
      "step": 1453
    },
    {
      "epoch": 3.61,
      "learning_rate": 7.371519775134107e-06,
      "loss": 0.5362,
      "step": 1454
    },
    {
      "epoch": 3.61,
      "learning_rate": 7.358462619456849e-06,
      "loss": 0.5481,
      "step": 1455
    },
    {
      "epoch": 3.61,
      "learning_rate": 7.345410303342715e-06,
      "loss": 0.57,
      "step": 1456
    },
    {
      "epoch": 3.61,
      "learning_rate": 7.332362850704866e-06,
      "loss": 0.5677,
      "step": 1457
    },
    {
      "epoch": 3.62,
      "learning_rate": 7.3193202854475554e-06,
      "loss": 0.5773,
      "step": 1458
    },
    {
      "epoch": 3.62,
      "learning_rate": 7.306282631466075e-06,
      "loss": 0.5619,
      "step": 1459
    },
    {
      "epoch": 3.62,
      "learning_rate": 7.293249912646725e-06,
      "loss": 0.5441,
      "step": 1460
    },
    {
      "epoch": 3.62,
      "learning_rate": 7.2802221528667604e-06,
      "loss": 0.5665,
      "step": 1461
    },
    {
      "epoch": 3.63,
      "learning_rate": 7.267199375994355e-06,
      "loss": 0.554,
      "step": 1462
    },
    {
      "epoch": 3.63,
      "learning_rate": 7.25418160588855e-06,
      "loss": 0.5712,
      "step": 1463
    },
    {
      "epoch": 3.63,
      "learning_rate": 7.241168866399213e-06,
      "loss": 0.5282,
      "step": 1464
    },
    {
      "epoch": 3.63,
      "learning_rate": 7.228161181366991e-06,
      "loss": 0.5543,
      "step": 1465
    },
    {
      "epoch": 3.64,
      "learning_rate": 7.215158574623285e-06,
      "loss": 0.5428,
      "step": 1466
    },
    {
      "epoch": 3.64,
      "learning_rate": 7.202161069990181e-06,
      "loss": 0.5437,
      "step": 1467
    },
    {
      "epoch": 3.64,
      "learning_rate": 7.189168691280417e-06,
      "loss": 0.5539,
      "step": 1468
    },
    {
      "epoch": 3.64,
      "learning_rate": 7.176181462297343e-06,
      "loss": 0.5165,
      "step": 1469
    },
    {
      "epoch": 3.65,
      "learning_rate": 7.1631994068348755e-06,
      "loss": 0.5484,
      "step": 1470
    },
    {
      "epoch": 3.65,
      "learning_rate": 7.15022254867745e-06,
      "loss": 0.5712,
      "step": 1471
    },
    {
      "epoch": 3.65,
      "learning_rate": 7.137250911599978e-06,
      "loss": 0.5305,
      "step": 1472
    },
    {
      "epoch": 3.65,
      "learning_rate": 7.124284519367808e-06,
      "loss": 0.5539,
      "step": 1473
    },
    {
      "epoch": 3.66,
      "learning_rate": 7.111323395736683e-06,
      "loss": 0.5604,
      "step": 1474
    },
    {
      "epoch": 3.66,
      "learning_rate": 7.098367564452685e-06,
      "loss": 0.5419,
      "step": 1475
    },
    {
      "epoch": 3.66,
      "learning_rate": 7.085417049252209e-06,
      "loss": 0.5393,
      "step": 1476
    },
    {
      "epoch": 3.66,
      "learning_rate": 7.072471873861905e-06,
      "loss": 0.5311,
      "step": 1477
    },
    {
      "epoch": 3.67,
      "learning_rate": 7.059532061998639e-06,
      "loss": 0.5445,
      "step": 1478
    },
    {
      "epoch": 3.67,
      "learning_rate": 7.046597637369456e-06,
      "loss": 0.5336,
      "step": 1479
    },
    {
      "epoch": 3.67,
      "learning_rate": 7.033668623671524e-06,
      "loss": 0.5425,
      "step": 1480
    },
    {
      "epoch": 3.67,
      "learning_rate": 7.020745044592103e-06,
      "loss": 0.5246,
      "step": 1481
    },
    {
      "epoch": 3.68,
      "learning_rate": 7.007826923808496e-06,
      "loss": 0.5463,
      "step": 1482
    },
    {
      "epoch": 3.68,
      "learning_rate": 6.9949142849880015e-06,
      "loss": 0.555,
      "step": 1483
    },
    {
      "epoch": 3.68,
      "learning_rate": 6.982007151787875e-06,
      "loss": 0.5259,
      "step": 1484
    },
    {
      "epoch": 3.68,
      "learning_rate": 6.969105547855293e-06,
      "loss": 0.563,
      "step": 1485
    },
    {
      "epoch": 3.69,
      "learning_rate": 6.9562094968272955e-06,
      "loss": 0.5923,
      "step": 1486
    },
    {
      "epoch": 3.69,
      "learning_rate": 6.943319022330747e-06,
      "loss": 0.5414,
      "step": 1487
    },
    {
      "epoch": 3.69,
      "learning_rate": 6.930434147982299e-06,
      "loss": 0.553,
      "step": 1488
    },
    {
      "epoch": 3.69,
      "learning_rate": 6.917554897388342e-06,
      "loss": 0.5619,
      "step": 1489
    },
    {
      "epoch": 3.7,
      "learning_rate": 6.904681294144964e-06,
      "loss": 0.5524,
      "step": 1490
    },
    {
      "epoch": 3.7,
      "learning_rate": 6.891813361837903e-06,
      "loss": 0.5394,
      "step": 1491
    },
    {
      "epoch": 3.7,
      "learning_rate": 6.878951124042514e-06,
      "loss": 0.5462,
      "step": 1492
    },
    {
      "epoch": 3.7,
      "learning_rate": 6.8660946043237145e-06,
      "loss": 0.5474,
      "step": 1493
    },
    {
      "epoch": 3.71,
      "learning_rate": 6.8532438262359404e-06,
      "loss": 0.5165,
      "step": 1494
    },
    {
      "epoch": 3.71,
      "learning_rate": 6.840398813323121e-06,
      "loss": 0.5458,
      "step": 1495
    },
    {
      "epoch": 3.71,
      "learning_rate": 6.827559589118613e-06,
      "loss": 0.5402,
      "step": 1496
    },
    {
      "epoch": 3.71,
      "learning_rate": 6.814726177145172e-06,
      "loss": 0.5494,
      "step": 1497
    },
    {
      "epoch": 3.72,
      "learning_rate": 6.801898600914905e-06,
      "loss": 0.5291,
      "step": 1498
    },
    {
      "epoch": 3.72,
      "learning_rate": 6.7890768839292265e-06,
      "loss": 0.5329,
      "step": 1499
    },
    {
      "epoch": 3.72,
      "learning_rate": 6.776261049678814e-06,
      "loss": 0.523,
      "step": 1500
    },
    {
      "epoch": 3.72,
      "learning_rate": 6.763451121643573e-06,
      "loss": 0.5897,
      "step": 1501
    },
    {
      "epoch": 3.73,
      "learning_rate": 6.750647123292582e-06,
      "loss": 0.5451,
      "step": 1502
    },
    {
      "epoch": 3.73,
      "learning_rate": 6.73784907808406e-06,
      "loss": 0.5997,
      "step": 1503
    },
    {
      "epoch": 3.73,
      "learning_rate": 6.725057009465315e-06,
      "loss": 0.5603,
      "step": 1504
    },
    {
      "epoch": 3.73,
      "learning_rate": 6.712270940872713e-06,
      "loss": 0.6094,
      "step": 1505
    },
    {
      "epoch": 3.74,
      "learning_rate": 6.699490895731618e-06,
      "loss": 0.571,
      "step": 1506
    },
    {
      "epoch": 3.74,
      "learning_rate": 6.686716897456365e-06,
      "loss": 0.524,
      "step": 1507
    },
    {
      "epoch": 3.74,
      "learning_rate": 6.673948969450207e-06,
      "loss": 0.5815,
      "step": 1508
    },
    {
      "epoch": 3.74,
      "learning_rate": 6.661187135105278e-06,
      "loss": 0.526,
      "step": 1509
    },
    {
      "epoch": 3.75,
      "learning_rate": 6.648431417802546e-06,
      "loss": 0.5515,
      "step": 1510
    },
    {
      "epoch": 3.75,
      "learning_rate": 6.635681840911772e-06,
      "loss": 0.539,
      "step": 1511
    },
    {
      "epoch": 3.75,
      "learning_rate": 6.62293842779147e-06,
      "loss": 0.5501,
      "step": 1512
    },
    {
      "epoch": 3.75,
      "learning_rate": 6.610201201788857e-06,
      "loss": 0.5365,
      "step": 1513
    },
    {
      "epoch": 3.76,
      "learning_rate": 6.597470186239818e-06,
      "loss": 0.5751,
      "step": 1514
    },
    {
      "epoch": 3.76,
      "learning_rate": 6.58474540446886e-06,
      "loss": 0.5514,
      "step": 1515
    },
    {
      "epoch": 3.76,
      "learning_rate": 6.572026879789064e-06,
      "loss": 0.5435,
      "step": 1516
    },
    {
      "epoch": 3.76,
      "learning_rate": 6.559314635502052e-06,
      "loss": 0.5587,
      "step": 1517
    },
    {
      "epoch": 3.77,
      "learning_rate": 6.54660869489794e-06,
      "loss": 0.5941,
      "step": 1518
    },
    {
      "epoch": 3.77,
      "learning_rate": 6.5339090812552945e-06,
      "loss": 0.5411,
      "step": 1519
    },
    {
      "epoch": 3.77,
      "learning_rate": 6.521215817841089e-06,
      "loss": 0.5652,
      "step": 1520
    },
    {
      "epoch": 3.77,
      "learning_rate": 6.508528927910662e-06,
      "loss": 0.554,
      "step": 1521
    },
    {
      "epoch": 3.78,
      "learning_rate": 6.495848434707676e-06,
      "loss": 0.5328,
      "step": 1522
    },
    {
      "epoch": 3.78,
      "learning_rate": 6.483174361464079e-06,
      "loss": 0.5668,
      "step": 1523
    },
    {
      "epoch": 3.78,
      "learning_rate": 6.470506731400046e-06,
      "loss": 0.5734,
      "step": 1524
    },
    {
      "epoch": 3.78,
      "learning_rate": 6.457845567723959e-06,
      "loss": 0.5364,
      "step": 1525
    },
    {
      "epoch": 3.79,
      "learning_rate": 6.445190893632347e-06,
      "loss": 0.5559,
      "step": 1526
    },
    {
      "epoch": 3.79,
      "learning_rate": 6.43254273230985e-06,
      "loss": 0.5667,
      "step": 1527
    },
    {
      "epoch": 3.79,
      "learning_rate": 6.419901106929175e-06,
      "loss": 0.5555,
      "step": 1528
    },
    {
      "epoch": 3.79,
      "learning_rate": 6.4072660406510576e-06,
      "loss": 0.5609,
      "step": 1529
    },
    {
      "epoch": 3.8,
      "learning_rate": 6.394637556624213e-06,
      "loss": 0.5912,
      "step": 1530
    },
    {
      "epoch": 3.8,
      "learning_rate": 6.3820156779853e-06,
      "loss": 0.6158,
      "step": 1531
    },
    {
      "epoch": 3.8,
      "learning_rate": 6.3694004278588765e-06,
      "loss": 0.5622,
      "step": 1532
    },
    {
      "epoch": 3.8,
      "learning_rate": 6.3567918293573486e-06,
      "loss": 0.5676,
      "step": 1533
    },
    {
      "epoch": 3.81,
      "learning_rate": 6.344189905580944e-06,
      "loss": 0.5743,
      "step": 1534
    },
    {
      "epoch": 3.81,
      "learning_rate": 6.331594679617661e-06,
      "loss": 0.5539,
      "step": 1535
    },
    {
      "epoch": 3.81,
      "learning_rate": 6.319006174543223e-06,
      "loss": 0.5363,
      "step": 1536
    },
    {
      "epoch": 3.81,
      "learning_rate": 6.306424413421045e-06,
      "loss": 0.5377,
      "step": 1537
    },
    {
      "epoch": 3.82,
      "learning_rate": 6.293849419302179e-06,
      "loss": 0.4815,
      "step": 1538
    },
    {
      "epoch": 3.82,
      "learning_rate": 6.281281215225286e-06,
      "loss": 0.5815,
      "step": 1539
    },
    {
      "epoch": 3.82,
      "learning_rate": 6.268719824216583e-06,
      "loss": 0.5743,
      "step": 1540
    },
    {
      "epoch": 3.82,
      "learning_rate": 6.256165269289807e-06,
      "loss": 0.5669,
      "step": 1541
    },
    {
      "epoch": 3.83,
      "learning_rate": 6.243617573446171e-06,
      "loss": 0.5575,
      "step": 1542
    },
    {
      "epoch": 3.83,
      "learning_rate": 6.231076759674315e-06,
      "loss": 0.5369,
      "step": 1543
    },
    {
      "epoch": 3.83,
      "learning_rate": 6.2185428509502855e-06,
      "loss": 0.5679,
      "step": 1544
    },
    {
      "epoch": 3.83,
      "learning_rate": 6.206015870237465e-06,
      "loss": 0.5769,
      "step": 1545
    },
    {
      "epoch": 3.84,
      "learning_rate": 6.193495840486543e-06,
      "loss": 0.5303,
      "step": 1546
    },
    {
      "epoch": 3.84,
      "learning_rate": 6.1809827846354805e-06,
      "loss": 0.5664,
      "step": 1547
    },
    {
      "epoch": 3.84,
      "learning_rate": 6.168476725609458e-06,
      "loss": 0.5506,
      "step": 1548
    },
    {
      "epoch": 3.84,
      "learning_rate": 6.155977686320837e-06,
      "loss": 0.5515,
      "step": 1549
    },
    {
      "epoch": 3.85,
      "learning_rate": 6.143485689669122e-06,
      "loss": 0.5501,
      "step": 1550
    },
    {
      "epoch": 3.85,
      "learning_rate": 6.13100075854091e-06,
      "loss": 0.5266,
      "step": 1551
    },
    {
      "epoch": 3.85,
      "learning_rate": 6.118522915809854e-06,
      "loss": 0.5689,
      "step": 1552
    },
    {
      "epoch": 3.85,
      "learning_rate": 6.106052184336626e-06,
      "loss": 0.5825,
      "step": 1553
    },
    {
      "epoch": 3.86,
      "learning_rate": 6.0935885869688595e-06,
      "loss": 0.5818,
      "step": 1554
    },
    {
      "epoch": 3.86,
      "learning_rate": 6.081132146541126e-06,
      "loss": 0.5361,
      "step": 1555
    },
    {
      "epoch": 3.86,
      "learning_rate": 6.068682885874879e-06,
      "loss": 0.5951,
      "step": 1556
    },
    {
      "epoch": 3.86,
      "learning_rate": 6.0562408277784255e-06,
      "loss": 0.5704,
      "step": 1557
    },
    {
      "epoch": 3.87,
      "learning_rate": 6.0438059950468675e-06,
      "loss": 0.5139,
      "step": 1558
    },
    {
      "epoch": 3.87,
      "learning_rate": 6.0313784104620774e-06,
      "loss": 0.5699,
      "step": 1559
    },
    {
      "epoch": 3.87,
      "learning_rate": 6.018958096792642e-06,
      "loss": 0.5679,
      "step": 1560
    },
    {
      "epoch": 3.87,
      "learning_rate": 6.006545076793828e-06,
      "loss": 0.5682,
      "step": 1561
    },
    {
      "epoch": 3.88,
      "learning_rate": 5.994139373207545e-06,
      "loss": 0.5473,
      "step": 1562
    },
    {
      "epoch": 3.88,
      "learning_rate": 5.981741008762288e-06,
      "loss": 0.5714,
      "step": 1563
    },
    {
      "epoch": 3.88,
      "learning_rate": 5.96935000617312e-06,
      "loss": 0.5341,
      "step": 1564
    },
    {
      "epoch": 3.88,
      "learning_rate": 5.956966388141605e-06,
      "loss": 0.5591,
      "step": 1565
    },
    {
      "epoch": 3.89,
      "learning_rate": 5.944590177355783e-06,
      "loss": 0.585,
      "step": 1566
    },
    {
      "epoch": 3.89,
      "learning_rate": 5.932221396490121e-06,
      "loss": 0.5174,
      "step": 1567
    },
    {
      "epoch": 3.89,
      "learning_rate": 5.919860068205472e-06,
      "loss": 0.5409,
      "step": 1568
    },
    {
      "epoch": 3.89,
      "learning_rate": 5.907506215149037e-06,
      "loss": 0.519,
      "step": 1569
    },
    {
      "epoch": 3.9,
      "learning_rate": 5.895159859954323e-06,
      "loss": 0.5899,
      "step": 1570
    },
    {
      "epoch": 3.9,
      "learning_rate": 5.8828210252411e-06,
      "loss": 0.5479,
      "step": 1571
    },
    {
      "epoch": 3.9,
      "learning_rate": 5.870489733615357e-06,
      "loss": 0.5534,
      "step": 1572
    },
    {
      "epoch": 3.9,
      "learning_rate": 5.858166007669266e-06,
      "loss": 0.5322,
      "step": 1573
    },
    {
      "epoch": 3.91,
      "learning_rate": 5.845849869981137e-06,
      "loss": 0.5424,
      "step": 1574
    },
    {
      "epoch": 3.91,
      "learning_rate": 5.833541343115377e-06,
      "loss": 0.5458,
      "step": 1575
    },
    {
      "epoch": 3.91,
      "learning_rate": 5.821240449622451e-06,
      "loss": 0.5553,
      "step": 1576
    },
    {
      "epoch": 3.91,
      "learning_rate": 5.808947212038833e-06,
      "loss": 0.5447,
      "step": 1577
    },
    {
      "epoch": 3.92,
      "learning_rate": 5.796661652886983e-06,
      "loss": 0.5627,
      "step": 1578
    },
    {
      "epoch": 3.92,
      "learning_rate": 5.784383794675277e-06,
      "loss": 0.5727,
      "step": 1579
    },
    {
      "epoch": 3.92,
      "learning_rate": 5.772113659897999e-06,
      "loss": 0.5711,
      "step": 1580
    },
    {
      "epoch": 3.92,
      "learning_rate": 5.759851271035266e-06,
      "loss": 0.5464,
      "step": 1581
    },
    {
      "epoch": 3.92,
      "learning_rate": 5.74759665055302e-06,
      "loss": 0.561,
      "step": 1582
    },
    {
      "epoch": 3.93,
      "learning_rate": 5.735349820902955e-06,
      "loss": 0.5672,
      "step": 1583
    },
    {
      "epoch": 3.93,
      "learning_rate": 5.7231108045225026e-06,
      "loss": 0.5406,
      "step": 1584
    },
    {
      "epoch": 3.93,
      "learning_rate": 5.710879623834782e-06,
      "loss": 0.552,
      "step": 1585
    },
    {
      "epoch": 3.93,
      "learning_rate": 5.698656301248541e-06,
      "loss": 0.5257,
      "step": 1586
    },
    {
      "epoch": 3.94,
      "learning_rate": 5.686440859158149e-06,
      "loss": 0.5509,
      "step": 1587
    },
    {
      "epoch": 3.94,
      "learning_rate": 5.674233319943522e-06,
      "loss": 0.554,
      "step": 1588
    },
    {
      "epoch": 3.94,
      "learning_rate": 5.662033705970111e-06,
      "loss": 0.5642,
      "step": 1589
    },
    {
      "epoch": 3.94,
      "learning_rate": 5.649842039588833e-06,
      "loss": 0.5818,
      "step": 1590
    },
    {
      "epoch": 3.95,
      "learning_rate": 5.637658343136058e-06,
      "loss": 0.5526,
      "step": 1591
    },
    {
      "epoch": 3.95,
      "learning_rate": 5.625482638933542e-06,
      "loss": 0.5642,
      "step": 1592
    },
    {
      "epoch": 3.95,
      "learning_rate": 5.613314949288409e-06,
      "loss": 0.5517,
      "step": 1593
    },
    {
      "epoch": 3.95,
      "learning_rate": 5.60115529649309e-06,
      "loss": 0.5589,
      "step": 1594
    },
    {
      "epoch": 3.96,
      "learning_rate": 5.589003702825302e-06,
      "loss": 0.5659,
      "step": 1595
    },
    {
      "epoch": 3.96,
      "learning_rate": 5.576860190547988e-06,
      "loss": 0.5699,
      "step": 1596
    },
    {
      "epoch": 3.96,
      "learning_rate": 5.564724781909285e-06,
      "loss": 0.5559,
      "step": 1597
    },
    {
      "epoch": 3.96,
      "learning_rate": 5.552597499142495e-06,
      "loss": 0.5309,
      "step": 1598
    },
    {
      "epoch": 3.97,
      "learning_rate": 5.540478364466014e-06,
      "loss": 0.5723,
      "step": 1599
    },
    {
      "epoch": 3.97,
      "learning_rate": 5.52836740008333e-06,
      "loss": 0.5386,
      "step": 1600
    },
    {
      "epoch": 3.97,
      "learning_rate": 5.516264628182944e-06,
      "loss": 0.6071,
      "step": 1601
    },
    {
      "epoch": 3.97,
      "learning_rate": 5.504170070938357e-06,
      "loss": 0.5642,
      "step": 1602
    },
    {
      "epoch": 3.98,
      "learning_rate": 5.4920837505080224e-06,
      "loss": 0.5294,
      "step": 1603
    },
    {
      "epoch": 3.98,
      "learning_rate": 5.4800056890353025e-06,
      "loss": 0.5818,
      "step": 1604
    },
    {
      "epoch": 3.98,
      "learning_rate": 5.467935908648421e-06,
      "loss": 0.5519,
      "step": 1605
    },
    {
      "epoch": 3.98,
      "learning_rate": 5.4558744314604275e-06,
      "loss": 0.5417,
      "step": 1606
    },
    {
      "epoch": 3.99,
      "learning_rate": 5.443821279569177e-06,
      "loss": 0.5346,
      "step": 1607
    },
    {
      "epoch": 3.99,
      "learning_rate": 5.431776475057251e-06,
      "loss": 0.5229,
      "step": 1608
    },
    {
      "epoch": 3.99,
      "learning_rate": 5.419740039991954e-06,
      "loss": 0.5189,
      "step": 1609
    },
    {
      "epoch": 3.99,
      "learning_rate": 5.407711996425244e-06,
      "loss": 0.5339,
      "step": 1610
    },
    {
      "epoch": 4.0,
      "learning_rate": 5.39569236639372e-06,
      "loss": 0.5379,
      "step": 1611
    },
    {
      "epoch": 4.0,
      "learning_rate": 5.3836811719185465e-06,
      "loss": 0.5599,
      "step": 1612
    },
    {
      "epoch": 4.0,
      "learning_rate": 5.3716784350054544e-06,
      "loss": 0.3744,
      "step": 1613
    },
    {
      "epoch": 4.0,
      "learning_rate": 5.359684177644659e-06,
      "loss": 0.3225,
      "step": 1614
    },
    {
      "epoch": 4.01,
      "learning_rate": 5.347698421810861e-06,
      "loss": 0.3253,
      "step": 1615
    },
    {
      "epoch": 4.01,
      "learning_rate": 5.335721189463165e-06,
      "loss": 0.2761,
      "step": 1616
    },
    {
      "epoch": 4.01,
      "learning_rate": 5.32375250254508e-06,
      "loss": 0.2888,
      "step": 1617
    },
    {
      "epoch": 4.01,
      "learning_rate": 5.311792382984442e-06,
      "loss": 0.2883,
      "step": 1618
    },
    {
      "epoch": 4.02,
      "learning_rate": 5.299840852693396e-06,
      "loss": 0.2637,
      "step": 1619
    },
    {
      "epoch": 4.02,
      "learning_rate": 5.287897933568356e-06,
      "loss": 0.301,
      "step": 1620
    },
    {
      "epoch": 4.02,
      "learning_rate": 5.2759636474899526e-06,
      "loss": 0.2736,
      "step": 1621
    },
    {
      "epoch": 4.02,
      "learning_rate": 5.264038016323002e-06,
      "loss": 0.2867,
      "step": 1622
    },
    {
      "epoch": 4.03,
      "learning_rate": 5.252121061916466e-06,
      "loss": 0.3176,
      "step": 1623
    },
    {
      "epoch": 4.03,
      "learning_rate": 5.240212806103413e-06,
      "loss": 0.32,
      "step": 1624
    },
    {
      "epoch": 4.03,
      "learning_rate": 5.228313270700961e-06,
      "loss": 0.2836,
      "step": 1625
    },
    {
      "epoch": 4.03,
      "learning_rate": 5.216422477510267e-06,
      "loss": 0.2989,
      "step": 1626
    },
    {
      "epoch": 4.04,
      "learning_rate": 5.204540448316462e-06,
      "loss": 0.2905,
      "step": 1627
    },
    {
      "epoch": 4.04,
      "learning_rate": 5.1926672048886175e-06,
      "loss": 0.2819,
      "step": 1628
    },
    {
      "epoch": 4.04,
      "learning_rate": 5.180802768979723e-06,
      "loss": 0.287,
      "step": 1629
    },
    {
      "epoch": 4.04,
      "learning_rate": 5.168947162326614e-06,
      "loss": 0.3082,
      "step": 1630
    },
    {
      "epoch": 4.05,
      "learning_rate": 5.157100406649967e-06,
      "loss": 0.2616,
      "step": 1631
    },
    {
      "epoch": 4.05,
      "learning_rate": 5.145262523654226e-06,
      "loss": 0.3143,
      "step": 1632
    },
    {
      "epoch": 4.05,
      "learning_rate": 5.1334335350275955e-06,
      "loss": 0.2709,
      "step": 1633
    },
    {
      "epoch": 4.05,
      "learning_rate": 5.12161346244197e-06,
      "loss": 0.2819,
      "step": 1634
    },
    {
      "epoch": 4.06,
      "learning_rate": 5.109802327552922e-06,
      "loss": 0.2741,
      "step": 1635
    },
    {
      "epoch": 4.06,
      "learning_rate": 5.098000151999636e-06,
      "loss": 0.2736,
      "step": 1636
    },
    {
      "epoch": 4.06,
      "learning_rate": 5.0862069574048956e-06,
      "loss": 0.2483,
      "step": 1637
    },
    {
      "epoch": 4.06,
      "learning_rate": 5.074422765375017e-06,
      "loss": 0.289,
      "step": 1638
    },
    {
      "epoch": 4.07,
      "learning_rate": 5.062647597499837e-06,
      "loss": 0.2883,
      "step": 1639
    },
    {
      "epoch": 4.07,
      "learning_rate": 5.050881475352647e-06,
      "loss": 0.276,
      "step": 1640
    },
    {
      "epoch": 4.07,
      "learning_rate": 5.039124420490169e-06,
      "loss": 0.2681,
      "step": 1641
    },
    {
      "epoch": 4.07,
      "learning_rate": 5.027376454452515e-06,
      "loss": 0.3193,
      "step": 1642
    },
    {
      "epoch": 4.08,
      "learning_rate": 5.015637598763148e-06,
      "loss": 0.2961,
      "step": 1643
    },
    {
      "epoch": 4.08,
      "learning_rate": 5.003907874928837e-06,
      "loss": 0.277,
      "step": 1644
    },
    {
      "epoch": 4.08,
      "learning_rate": 4.992187304439616e-06,
      "loss": 0.3068,
      "step": 1645
    },
    {
      "epoch": 4.08,
      "learning_rate": 4.980475908768758e-06,
      "loss": 0.2935,
      "step": 1646
    },
    {
      "epoch": 4.09,
      "learning_rate": 4.96877370937272e-06,
      "loss": 0.2775,
      "step": 1647
    },
    {
      "epoch": 4.09,
      "learning_rate": 4.957080727691107e-06,
      "loss": 0.2773,
      "step": 1648
    },
    {
      "epoch": 4.09,
      "learning_rate": 4.94539698514665e-06,
      "loss": 0.2589,
      "step": 1649
    },
    {
      "epoch": 4.09,
      "learning_rate": 4.933722503145139e-06,
      "loss": 0.2841,
      "step": 1650
    },
    {
      "epoch": 4.1,
      "learning_rate": 4.9220573030754085e-06,
      "loss": 0.2802,
      "step": 1651
    },
    {
      "epoch": 4.1,
      "learning_rate": 4.910401406309276e-06,
      "loss": 0.268,
      "step": 1652
    },
    {
      "epoch": 4.1,
      "learning_rate": 4.898754834201529e-06,
      "loss": 0.2753,
      "step": 1653
    },
    {
      "epoch": 4.1,
      "learning_rate": 4.8871176080898555e-06,
      "loss": 0.2735,
      "step": 1654
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.875489749294835e-06,
      "loss": 0.272,
      "step": 1655
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.863871279119872e-06,
      "loss": 0.2684,
      "step": 1656
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.8522622188511835e-06,
      "loss": 0.2672,
      "step": 1657
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.8406625897577335e-06,
      "loss": 0.2491,
      "step": 1658
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.829072413091219e-06,
      "loss": 0.2756,
      "step": 1659
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.8174917100860076e-06,
      "loss": 0.3082,
      "step": 1660
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.80592050195912e-06,
      "loss": 0.2627,
      "step": 1661
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.79435880991018e-06,
      "loss": 0.2636,
      "step": 1662
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.782806655121368e-06,
      "loss": 0.2869,
      "step": 1663
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.771264058757405e-06,
      "loss": 0.263,
      "step": 1664
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.759731041965485e-06,
      "loss": 0.2647,
      "step": 1665
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.748207625875268e-06,
      "loss": 0.2746,
      "step": 1666
    },
    {
      "epoch": 4.14,
      "learning_rate": 4.736693831598804e-06,
      "loss": 0.2517,
      "step": 1667
    },
    {
      "epoch": 4.14,
      "learning_rate": 4.725189680230539e-06,
      "loss": 0.2932,
      "step": 1668
    },
    {
      "epoch": 4.14,
      "learning_rate": 4.713695192847231e-06,
      "loss": 0.27,
      "step": 1669
    },
    {
      "epoch": 4.14,
      "learning_rate": 4.7022103905079405e-06,
      "loss": 0.2966,
      "step": 1670
    },
    {
      "epoch": 4.15,
      "learning_rate": 4.6907352942539915e-06,
      "loss": 0.2917,
      "step": 1671
    },
    {
      "epoch": 4.15,
      "learning_rate": 4.67926992510891e-06,
      "loss": 0.295,
      "step": 1672
    },
    {
      "epoch": 4.15,
      "learning_rate": 4.66781430407842e-06,
      "loss": 0.31,
      "step": 1673
    },
    {
      "epoch": 4.15,
      "learning_rate": 4.656368452150365e-06,
      "loss": 0.2867,
      "step": 1674
    },
    {
      "epoch": 4.16,
      "learning_rate": 4.64493239029471e-06,
      "loss": 0.3039,
      "step": 1675
    },
    {
      "epoch": 4.16,
      "learning_rate": 4.633506139463466e-06,
      "loss": 0.291,
      "step": 1676
    },
    {
      "epoch": 4.16,
      "learning_rate": 4.622089720590688e-06,
      "loss": 0.2718,
      "step": 1677
    },
    {
      "epoch": 4.16,
      "learning_rate": 4.610683154592398e-06,
      "loss": 0.2713,
      "step": 1678
    },
    {
      "epoch": 4.17,
      "learning_rate": 4.599286462366584e-06,
      "loss": 0.2829,
      "step": 1679
    },
    {
      "epoch": 4.17,
      "learning_rate": 4.58789966479313e-06,
      "loss": 0.2635,
      "step": 1680
    },
    {
      "epoch": 4.17,
      "learning_rate": 4.576522782733802e-06,
      "loss": 0.2587,
      "step": 1681
    },
    {
      "epoch": 4.17,
      "learning_rate": 4.5651558370322005e-06,
      "loss": 0.2894,
      "step": 1682
    },
    {
      "epoch": 4.18,
      "learning_rate": 4.553798848513709e-06,
      "loss": 0.2648,
      "step": 1683
    },
    {
      "epoch": 4.18,
      "learning_rate": 4.5424518379854845e-06,
      "loss": 0.3062,
      "step": 1684
    },
    {
      "epoch": 4.18,
      "learning_rate": 4.531114826236388e-06,
      "loss": 0.268,
      "step": 1685
    },
    {
      "epoch": 4.18,
      "learning_rate": 4.519787834036978e-06,
      "loss": 0.2625,
      "step": 1686
    },
    {
      "epoch": 4.19,
      "learning_rate": 4.5084708821394396e-06,
      "loss": 0.2768,
      "step": 1687
    },
    {
      "epoch": 4.19,
      "learning_rate": 4.497163991277579e-06,
      "loss": 0.2893,
      "step": 1688
    },
    {
      "epoch": 4.19,
      "learning_rate": 4.485867182166752e-06,
      "loss": 0.286,
      "step": 1689
    },
    {
      "epoch": 4.19,
      "learning_rate": 4.474580475503863e-06,
      "loss": 0.2973,
      "step": 1690
    },
    {
      "epoch": 4.2,
      "learning_rate": 4.463303891967295e-06,
      "loss": 0.2932,
      "step": 1691
    },
    {
      "epoch": 4.2,
      "learning_rate": 4.45203745221688e-06,
      "loss": 0.2842,
      "step": 1692
    },
    {
      "epoch": 4.2,
      "learning_rate": 4.4407811768938826e-06,
      "loss": 0.3063,
      "step": 1693
    },
    {
      "epoch": 4.2,
      "learning_rate": 4.429535086620929e-06,
      "loss": 0.2927,
      "step": 1694
    },
    {
      "epoch": 4.21,
      "learning_rate": 4.4182992020019964e-06,
      "loss": 0.287,
      "step": 1695
    },
    {
      "epoch": 4.21,
      "learning_rate": 4.407073543622354e-06,
      "loss": 0.2572,
      "step": 1696
    },
    {
      "epoch": 4.21,
      "learning_rate": 4.395858132048549e-06,
      "loss": 0.2757,
      "step": 1697
    },
    {
      "epoch": 4.21,
      "learning_rate": 4.384652987828339e-06,
      "loss": 0.3002,
      "step": 1698
    },
    {
      "epoch": 4.22,
      "learning_rate": 4.3734581314906874e-06,
      "loss": 0.2686,
      "step": 1699
    },
    {
      "epoch": 4.22,
      "learning_rate": 4.3622735835456935e-06,
      "loss": 0.2696,
      "step": 1700
    },
    {
      "epoch": 4.22,
      "learning_rate": 4.351099364484581e-06,
      "loss": 0.2607,
      "step": 1701
    },
    {
      "epoch": 4.22,
      "learning_rate": 4.3399354947796514e-06,
      "loss": 0.2671,
      "step": 1702
    },
    {
      "epoch": 4.23,
      "learning_rate": 4.3287819948842334e-06,
      "loss": 0.2927,
      "step": 1703
    },
    {
      "epoch": 4.23,
      "learning_rate": 4.3176388852326735e-06,
      "loss": 0.2894,
      "step": 1704
    },
    {
      "epoch": 4.23,
      "learning_rate": 4.306506186240264e-06,
      "loss": 0.2571,
      "step": 1705
    },
    {
      "epoch": 4.23,
      "learning_rate": 4.295383918303241e-06,
      "loss": 0.2879,
      "step": 1706
    },
    {
      "epoch": 4.24,
      "learning_rate": 4.284272101798717e-06,
      "loss": 0.3057,
      "step": 1707
    },
    {
      "epoch": 4.24,
      "learning_rate": 4.273170757084665e-06,
      "loss": 0.2636,
      "step": 1708
    },
    {
      "epoch": 4.24,
      "learning_rate": 4.262079904499864e-06,
      "loss": 0.2702,
      "step": 1709
    },
    {
      "epoch": 4.24,
      "learning_rate": 4.250999564363882e-06,
      "loss": 0.2845,
      "step": 1710
    },
    {
      "epoch": 4.24,
      "learning_rate": 4.239929756977016e-06,
      "loss": 0.2747,
      "step": 1711
    },
    {
      "epoch": 4.25,
      "learning_rate": 4.228870502620274e-06,
      "loss": 0.2879,
      "step": 1712
    },
    {
      "epoch": 4.25,
      "learning_rate": 4.217821821555325e-06,
      "loss": 0.3017,
      "step": 1713
    },
    {
      "epoch": 4.25,
      "learning_rate": 4.206783734024463e-06,
      "loss": 0.255,
      "step": 1714
    },
    {
      "epoch": 4.25,
      "learning_rate": 4.195756260250588e-06,
      "loss": 0.2909,
      "step": 1715
    },
    {
      "epoch": 4.26,
      "learning_rate": 4.184739420437134e-06,
      "loss": 0.2858,
      "step": 1716
    },
    {
      "epoch": 4.26,
      "learning_rate": 4.173733234768075e-06,
      "loss": 0.2985,
      "step": 1717
    },
    {
      "epoch": 4.26,
      "learning_rate": 4.162737723407844e-06,
      "loss": 0.274,
      "step": 1718
    },
    {
      "epoch": 4.26,
      "learning_rate": 4.1517529065013375e-06,
      "loss": 0.2788,
      "step": 1719
    },
    {
      "epoch": 4.27,
      "learning_rate": 4.1407788041738366e-06,
      "loss": 0.274,
      "step": 1720
    },
    {
      "epoch": 4.27,
      "learning_rate": 4.129815436531018e-06,
      "loss": 0.2889,
      "step": 1721
    },
    {
      "epoch": 4.27,
      "learning_rate": 4.118862823658872e-06,
      "loss": 0.3119,
      "step": 1722
    },
    {
      "epoch": 4.27,
      "learning_rate": 4.107920985623688e-06,
      "loss": 0.2827,
      "step": 1723
    },
    {
      "epoch": 4.28,
      "learning_rate": 4.096989942472023e-06,
      "loss": 0.2906,
      "step": 1724
    },
    {
      "epoch": 4.28,
      "learning_rate": 4.086069714230646e-06,
      "loss": 0.2548,
      "step": 1725
    },
    {
      "epoch": 4.28,
      "learning_rate": 4.075160320906524e-06,
      "loss": 0.2555,
      "step": 1726
    },
    {
      "epoch": 4.28,
      "learning_rate": 4.064261782486759e-06,
      "loss": 0.2921,
      "step": 1727
    },
    {
      "epoch": 4.29,
      "learning_rate": 4.0533741189385815e-06,
      "loss": 0.2647,
      "step": 1728
    },
    {
      "epoch": 4.29,
      "learning_rate": 4.042497350209282e-06,
      "loss": 0.2575,
      "step": 1729
    },
    {
      "epoch": 4.29,
      "learning_rate": 4.031631496226208e-06,
      "loss": 0.2882,
      "step": 1730
    },
    {
      "epoch": 4.29,
      "learning_rate": 4.020776576896692e-06,
      "loss": 0.2789,
      "step": 1731
    },
    {
      "epoch": 4.3,
      "learning_rate": 4.00993261210805e-06,
      "loss": 0.3184,
      "step": 1732
    },
    {
      "epoch": 4.3,
      "learning_rate": 3.99909962172751e-06,
      "loss": 0.2801,
      "step": 1733
    },
    {
      "epoch": 4.3,
      "learning_rate": 3.988277625602216e-06,
      "loss": 0.3053,
      "step": 1734
    },
    {
      "epoch": 4.3,
      "learning_rate": 3.97746664355915e-06,
      "loss": 0.2799,
      "step": 1735
    },
    {
      "epoch": 4.31,
      "learning_rate": 3.96666669540512e-06,
      "loss": 0.2492,
      "step": 1736
    },
    {
      "epoch": 4.31,
      "learning_rate": 3.955877800926728e-06,
      "loss": 0.2946,
      "step": 1737
    },
    {
      "epoch": 4.31,
      "learning_rate": 3.945099979890311e-06,
      "loss": 0.2574,
      "step": 1738
    },
    {
      "epoch": 4.31,
      "learning_rate": 3.934333252041933e-06,
      "loss": 0.2905,
      "step": 1739
    },
    {
      "epoch": 4.32,
      "learning_rate": 3.923577637107314e-06,
      "loss": 0.2693,
      "step": 1740
    },
    {
      "epoch": 4.32,
      "learning_rate": 3.912833154791842e-06,
      "loss": 0.296,
      "step": 1741
    },
    {
      "epoch": 4.32,
      "learning_rate": 3.902099824780481e-06,
      "loss": 0.2821,
      "step": 1742
    },
    {
      "epoch": 4.32,
      "learning_rate": 3.8913776667377824e-06,
      "loss": 0.2981,
      "step": 1743
    },
    {
      "epoch": 4.33,
      "learning_rate": 3.880666700307819e-06,
      "loss": 0.2524,
      "step": 1744
    },
    {
      "epoch": 4.33,
      "learning_rate": 3.86996694511416e-06,
      "loss": 0.2923,
      "step": 1745
    },
    {
      "epoch": 4.33,
      "learning_rate": 3.859278420759841e-06,
      "loss": 0.2872,
      "step": 1746
    },
    {
      "epoch": 4.33,
      "learning_rate": 3.8486011468273145e-06,
      "loss": 0.2646,
      "step": 1747
    },
    {
      "epoch": 4.34,
      "learning_rate": 3.837935142878427e-06,
      "loss": 0.284,
      "step": 1748
    },
    {
      "epoch": 4.34,
      "learning_rate": 3.827280428454371e-06,
      "loss": 0.2964,
      "step": 1749
    },
    {
      "epoch": 4.34,
      "learning_rate": 3.816637023075663e-06,
      "loss": 0.279,
      "step": 1750
    },
    {
      "epoch": 4.34,
      "learning_rate": 3.806004946242092e-06,
      "loss": 0.2943,
      "step": 1751
    },
    {
      "epoch": 4.35,
      "learning_rate": 3.7953842174327004e-06,
      "loss": 0.2865,
      "step": 1752
    },
    {
      "epoch": 4.35,
      "learning_rate": 3.7847748561057296e-06,
      "loss": 0.2751,
      "step": 1753
    },
    {
      "epoch": 4.35,
      "learning_rate": 3.774176881698607e-06,
      "loss": 0.3087,
      "step": 1754
    },
    {
      "epoch": 4.35,
      "learning_rate": 3.763590313627885e-06,
      "loss": 0.2831,
      "step": 1755
    },
    {
      "epoch": 4.36,
      "learning_rate": 3.7530151712892317e-06,
      "loss": 0.2891,
      "step": 1756
    },
    {
      "epoch": 4.36,
      "learning_rate": 3.7424514740573713e-06,
      "loss": 0.2758,
      "step": 1757
    },
    {
      "epoch": 4.36,
      "learning_rate": 3.731899241286061e-06,
      "loss": 0.277,
      "step": 1758
    },
    {
      "epoch": 4.36,
      "learning_rate": 3.7213584923080636e-06,
      "loss": 0.2319,
      "step": 1759
    },
    {
      "epoch": 4.37,
      "learning_rate": 3.7108292464350837e-06,
      "loss": 0.2685,
      "step": 1760
    },
    {
      "epoch": 4.37,
      "learning_rate": 3.7003115229577758e-06,
      "loss": 0.2895,
      "step": 1761
    },
    {
      "epoch": 4.37,
      "learning_rate": 3.6898053411456626e-06,
      "loss": 0.2625,
      "step": 1762
    },
    {
      "epoch": 4.37,
      "learning_rate": 3.679310720247136e-06,
      "loss": 0.2469,
      "step": 1763
    },
    {
      "epoch": 4.38,
      "learning_rate": 3.668827679489393e-06,
      "loss": 0.3028,
      "step": 1764
    },
    {
      "epoch": 4.38,
      "learning_rate": 3.658356238078431e-06,
      "loss": 0.2887,
      "step": 1765
    },
    {
      "epoch": 4.38,
      "learning_rate": 3.6478964151989817e-06,
      "loss": 0.288,
      "step": 1766
    },
    {
      "epoch": 4.38,
      "learning_rate": 3.637448230014495e-06,
      "loss": 0.2521,
      "step": 1767
    },
    {
      "epoch": 4.39,
      "learning_rate": 3.627011701667108e-06,
      "loss": 0.2874,
      "step": 1768
    },
    {
      "epoch": 4.39,
      "learning_rate": 3.616586849277587e-06,
      "loss": 0.2761,
      "step": 1769
    },
    {
      "epoch": 4.39,
      "learning_rate": 3.6061736919453183e-06,
      "loss": 0.2923,
      "step": 1770
    },
    {
      "epoch": 4.39,
      "learning_rate": 3.5957722487482537e-06,
      "loss": 0.2646,
      "step": 1771
    },
    {
      "epoch": 4.4,
      "learning_rate": 3.585382538742892e-06,
      "loss": 0.275,
      "step": 1772
    },
    {
      "epoch": 4.4,
      "learning_rate": 3.575004580964225e-06,
      "loss": 0.32,
      "step": 1773
    },
    {
      "epoch": 4.4,
      "learning_rate": 3.564638394425726e-06,
      "loss": 0.2827,
      "step": 1774
    },
    {
      "epoch": 4.4,
      "learning_rate": 3.5542839981192857e-06,
      "loss": 0.25,
      "step": 1775
    },
    {
      "epoch": 4.41,
      "learning_rate": 3.5439414110152128e-06,
      "loss": 0.2946,
      "step": 1776
    },
    {
      "epoch": 4.41,
      "learning_rate": 3.5336106520621638e-06,
      "loss": 0.2717,
      "step": 1777
    },
    {
      "epoch": 4.41,
      "learning_rate": 3.523291740187137e-06,
      "loss": 0.2815,
      "step": 1778
    },
    {
      "epoch": 4.41,
      "learning_rate": 3.512984694295416e-06,
      "loss": 0.2964,
      "step": 1779
    },
    {
      "epoch": 4.42,
      "learning_rate": 3.5026895332705504e-06,
      "loss": 0.3005,
      "step": 1780
    },
    {
      "epoch": 4.42,
      "learning_rate": 3.4924062759743206e-06,
      "loss": 0.2511,
      "step": 1781
    },
    {
      "epoch": 4.42,
      "learning_rate": 3.482134941246682e-06,
      "loss": 0.3127,
      "step": 1782
    },
    {
      "epoch": 4.42,
      "learning_rate": 3.4718755479057663e-06,
      "loss": 0.2864,
      "step": 1783
    },
    {
      "epoch": 4.43,
      "learning_rate": 3.461628114747809e-06,
      "loss": 0.3049,
      "step": 1784
    },
    {
      "epoch": 4.43,
      "learning_rate": 3.4513926605471504e-06,
      "loss": 0.2732,
      "step": 1785
    },
    {
      "epoch": 4.43,
      "learning_rate": 3.4411692040561675e-06,
      "loss": 0.3056,
      "step": 1786
    },
    {
      "epoch": 4.43,
      "learning_rate": 3.4309577640052715e-06,
      "loss": 0.2807,
      "step": 1787
    },
    {
      "epoch": 4.44,
      "learning_rate": 3.4207583591028495e-06,
      "loss": 0.2776,
      "step": 1788
    },
    {
      "epoch": 4.44,
      "learning_rate": 3.410571008035235e-06,
      "loss": 0.2639,
      "step": 1789
    },
    {
      "epoch": 4.44,
      "learning_rate": 3.4003957294666933e-06,
      "loss": 0.2796,
      "step": 1790
    },
    {
      "epoch": 4.44,
      "learning_rate": 3.390232542039352e-06,
      "loss": 0.2899,
      "step": 1791
    },
    {
      "epoch": 4.45,
      "learning_rate": 3.380081464373205e-06,
      "loss": 0.3102,
      "step": 1792
    },
    {
      "epoch": 4.45,
      "learning_rate": 3.3699425150660447e-06,
      "loss": 0.2833,
      "step": 1793
    },
    {
      "epoch": 4.45,
      "learning_rate": 3.359815712693455e-06,
      "loss": 0.3069,
      "step": 1794
    },
    {
      "epoch": 4.45,
      "learning_rate": 3.3497010758087543e-06,
      "loss": 0.2985,
      "step": 1795
    },
    {
      "epoch": 4.46,
      "learning_rate": 3.3395986229429857e-06,
      "loss": 0.272,
      "step": 1796
    },
    {
      "epoch": 4.46,
      "learning_rate": 3.3295083726048547e-06,
      "loss": 0.3086,
      "step": 1797
    },
    {
      "epoch": 4.46,
      "learning_rate": 3.319430343280725e-06,
      "loss": 0.2948,
      "step": 1798
    },
    {
      "epoch": 4.46,
      "learning_rate": 3.3093645534345585e-06,
      "loss": 0.2599,
      "step": 1799
    },
    {
      "epoch": 4.47,
      "learning_rate": 3.2993110215079004e-06,
      "loss": 0.2955,
      "step": 1800
    },
    {
      "epoch": 4.47,
      "learning_rate": 3.289269765919839e-06,
      "loss": 0.2684,
      "step": 1801
    },
    {
      "epoch": 4.47,
      "learning_rate": 3.2792408050669634e-06,
      "loss": 0.2721,
      "step": 1802
    },
    {
      "epoch": 4.47,
      "learning_rate": 3.269224157323346e-06,
      "loss": 0.2798,
      "step": 1803
    },
    {
      "epoch": 4.48,
      "learning_rate": 3.259219841040492e-06,
      "loss": 0.3035,
      "step": 1804
    },
    {
      "epoch": 4.48,
      "learning_rate": 3.249227874547325e-06,
      "loss": 0.307,
      "step": 1805
    },
    {
      "epoch": 4.48,
      "learning_rate": 3.23924827615013e-06,
      "loss": 0.3091,
      "step": 1806
    },
    {
      "epoch": 4.48,
      "learning_rate": 3.2292810641325434e-06,
      "loss": 0.2683,
      "step": 1807
    },
    {
      "epoch": 4.49,
      "learning_rate": 3.2193262567554994e-06,
      "loss": 0.2856,
      "step": 1808
    },
    {
      "epoch": 4.49,
      "learning_rate": 3.2093838722572148e-06,
      "loss": 0.2411,
      "step": 1809
    },
    {
      "epoch": 4.49,
      "learning_rate": 3.199453928853139e-06,
      "loss": 0.3,
      "step": 1810
    },
    {
      "epoch": 4.49,
      "learning_rate": 3.189536444735928e-06,
      "loss": 0.2608,
      "step": 1811
    },
    {
      "epoch": 4.5,
      "learning_rate": 3.17963143807542e-06,
      "loss": 0.2815,
      "step": 1812
    },
    {
      "epoch": 4.5,
      "learning_rate": 3.16973892701858e-06,
      "loss": 0.2692,
      "step": 1813
    },
    {
      "epoch": 4.5,
      "learning_rate": 3.1598589296894933e-06,
      "loss": 0.26,
      "step": 1814
    },
    {
      "epoch": 4.5,
      "learning_rate": 3.149991464189307e-06,
      "loss": 0.2725,
      "step": 1815
    },
    {
      "epoch": 4.51,
      "learning_rate": 3.1401365485962187e-06,
      "loss": 0.2796,
      "step": 1816
    },
    {
      "epoch": 4.51,
      "learning_rate": 3.1302942009654236e-06,
      "loss": 0.2782,
      "step": 1817
    },
    {
      "epoch": 4.51,
      "learning_rate": 3.120464439329102e-06,
      "loss": 0.2755,
      "step": 1818
    },
    {
      "epoch": 4.51,
      "learning_rate": 3.110647281696363e-06,
      "loss": 0.2972,
      "step": 1819
    },
    {
      "epoch": 4.52,
      "learning_rate": 3.100842746053233e-06,
      "loss": 0.2658,
      "step": 1820
    },
    {
      "epoch": 4.52,
      "learning_rate": 3.091050850362616e-06,
      "loss": 0.2733,
      "step": 1821
    },
    {
      "epoch": 4.52,
      "learning_rate": 3.0812716125642438e-06,
      "loss": 0.2832,
      "step": 1822
    },
    {
      "epoch": 4.52,
      "learning_rate": 3.0715050505746768e-06,
      "loss": 0.2898,
      "step": 1823
    },
    {
      "epoch": 4.53,
      "learning_rate": 3.0617511822872337e-06,
      "loss": 0.2727,
      "step": 1824
    },
    {
      "epoch": 4.53,
      "learning_rate": 3.0520100255719933e-06,
      "loss": 0.2553,
      "step": 1825
    },
    {
      "epoch": 4.53,
      "learning_rate": 3.0422815982757325e-06,
      "loss": 0.2619,
      "step": 1826
    },
    {
      "epoch": 4.53,
      "learning_rate": 3.0325659182219157e-06,
      "loss": 0.2689,
      "step": 1827
    },
    {
      "epoch": 4.54,
      "learning_rate": 3.0228630032106466e-06,
      "loss": 0.2797,
      "step": 1828
    },
    {
      "epoch": 4.54,
      "learning_rate": 3.013172871018648e-06,
      "loss": 0.2619,
      "step": 1829
    },
    {
      "epoch": 4.54,
      "learning_rate": 3.003495539399217e-06,
      "loss": 0.2877,
      "step": 1830
    },
    {
      "epoch": 4.54,
      "learning_rate": 2.9938310260822055e-06,
      "loss": 0.2938,
      "step": 1831
    },
    {
      "epoch": 4.55,
      "learning_rate": 2.9841793487739766e-06,
      "loss": 0.2616,
      "step": 1832
    },
    {
      "epoch": 4.55,
      "learning_rate": 2.9745405251573724e-06,
      "loss": 0.2829,
      "step": 1833
    },
    {
      "epoch": 4.55,
      "learning_rate": 2.9649145728916984e-06,
      "loss": 0.2774,
      "step": 1834
    },
    {
      "epoch": 4.55,
      "learning_rate": 2.9553015096126638e-06,
      "loss": 0.2962,
      "step": 1835
    },
    {
      "epoch": 4.56,
      "learning_rate": 2.9457013529323764e-06,
      "loss": 0.2763,
      "step": 1836
    },
    {
      "epoch": 4.56,
      "learning_rate": 2.9361141204392883e-06,
      "loss": 0.2441,
      "step": 1837
    },
    {
      "epoch": 4.56,
      "learning_rate": 2.9265398296981786e-06,
      "loss": 0.317,
      "step": 1838
    },
    {
      "epoch": 4.56,
      "learning_rate": 2.916978498250117e-06,
      "loss": 0.2837,
      "step": 1839
    },
    {
      "epoch": 4.57,
      "learning_rate": 2.907430143612423e-06,
      "loss": 0.2459,
      "step": 1840
    },
    {
      "epoch": 4.57,
      "learning_rate": 2.897894783278652e-06,
      "loss": 0.2768,
      "step": 1841
    },
    {
      "epoch": 4.57,
      "learning_rate": 2.8883724347185405e-06,
      "loss": 0.2997,
      "step": 1842
    },
    {
      "epoch": 4.57,
      "learning_rate": 2.878863115377999e-06,
      "loss": 0.2834,
      "step": 1843
    },
    {
      "epoch": 4.57,
      "learning_rate": 2.869366842679052e-06,
      "loss": 0.302,
      "step": 1844
    },
    {
      "epoch": 4.58,
      "learning_rate": 2.859883634019839e-06,
      "loss": 0.2869,
      "step": 1845
    },
    {
      "epoch": 4.58,
      "learning_rate": 2.8504135067745463e-06,
      "loss": 0.2959,
      "step": 1846
    },
    {
      "epoch": 4.58,
      "learning_rate": 2.840956478293411e-06,
      "loss": 0.28,
      "step": 1847
    },
    {
      "epoch": 4.58,
      "learning_rate": 2.8315125659026577e-06,
      "loss": 0.2762,
      "step": 1848
    },
    {
      "epoch": 4.59,
      "learning_rate": 2.8220817869044925e-06,
      "loss": 0.311,
      "step": 1849
    },
    {
      "epoch": 4.59,
      "learning_rate": 2.812664158577049e-06,
      "loss": 0.3055,
      "step": 1850
    },
    {
      "epoch": 4.59,
      "learning_rate": 2.8032596981743787e-06,
      "loss": 0.2617,
      "step": 1851
    },
    {
      "epoch": 4.59,
      "learning_rate": 2.7938684229263958e-06,
      "loss": 0.2828,
      "step": 1852
    },
    {
      "epoch": 4.6,
      "learning_rate": 2.784490350038872e-06,
      "loss": 0.278,
      "step": 1853
    },
    {
      "epoch": 4.6,
      "learning_rate": 2.7751254966933795e-06,
      "loss": 0.2918,
      "step": 1854
    },
    {
      "epoch": 4.6,
      "learning_rate": 2.765773880047272e-06,
      "loss": 0.2715,
      "step": 1855
    },
    {
      "epoch": 4.6,
      "learning_rate": 2.756435517233662e-06,
      "loss": 0.279,
      "step": 1856
    },
    {
      "epoch": 4.61,
      "learning_rate": 2.7471104253613645e-06,
      "loss": 0.2834,
      "step": 1857
    },
    {
      "epoch": 4.61,
      "learning_rate": 2.7377986215148955e-06,
      "loss": 0.2826,
      "step": 1858
    },
    {
      "epoch": 4.61,
      "learning_rate": 2.7285001227544163e-06,
      "loss": 0.2721,
      "step": 1859
    },
    {
      "epoch": 4.61,
      "learning_rate": 2.719214946115721e-06,
      "loss": 0.2676,
      "step": 1860
    },
    {
      "epoch": 4.62,
      "learning_rate": 2.7099431086101858e-06,
      "loss": 0.2954,
      "step": 1861
    },
    {
      "epoch": 4.62,
      "learning_rate": 2.70068462722475e-06,
      "loss": 0.2888,
      "step": 1862
    },
    {
      "epoch": 4.62,
      "learning_rate": 2.691439518921892e-06,
      "loss": 0.2847,
      "step": 1863
    },
    {
      "epoch": 4.62,
      "learning_rate": 2.682207800639579e-06,
      "loss": 0.2741,
      "step": 1864
    },
    {
      "epoch": 4.63,
      "learning_rate": 2.6729894892912556e-06,
      "loss": 0.2529,
      "step": 1865
    },
    {
      "epoch": 4.63,
      "learning_rate": 2.663784601765792e-06,
      "loss": 0.2697,
      "step": 1866
    },
    {
      "epoch": 4.63,
      "learning_rate": 2.6545931549274785e-06,
      "loss": 0.2946,
      "step": 1867
    },
    {
      "epoch": 4.63,
      "learning_rate": 2.6454151656159666e-06,
      "loss": 0.2622,
      "step": 1868
    },
    {
      "epoch": 4.64,
      "learning_rate": 2.636250650646266e-06,
      "loss": 0.2924,
      "step": 1869
    },
    {
      "epoch": 4.64,
      "learning_rate": 2.627099626808687e-06,
      "loss": 0.3005,
      "step": 1870
    },
    {
      "epoch": 4.64,
      "learning_rate": 2.6179621108688345e-06,
      "loss": 0.2787,
      "step": 1871
    },
    {
      "epoch": 4.64,
      "learning_rate": 2.608838119567555e-06,
      "loss": 0.305,
      "step": 1872
    },
    {
      "epoch": 4.65,
      "learning_rate": 2.599727669620926e-06,
      "loss": 0.275,
      "step": 1873
    },
    {
      "epoch": 4.65,
      "learning_rate": 2.590630777720208e-06,
      "loss": 0.3028,
      "step": 1874
    },
    {
      "epoch": 4.65,
      "learning_rate": 2.581547460531828e-06,
      "loss": 0.2689,
      "step": 1875
    },
    {
      "epoch": 4.65,
      "learning_rate": 2.5724777346973396e-06,
      "loss": 0.2907,
      "step": 1876
    },
    {
      "epoch": 4.66,
      "learning_rate": 2.5634216168333914e-06,
      "loss": 0.3219,
      "step": 1877
    },
    {
      "epoch": 4.66,
      "learning_rate": 2.5543791235317095e-06,
      "loss": 0.2505,
      "step": 1878
    },
    {
      "epoch": 4.66,
      "learning_rate": 2.545350271359055e-06,
      "loss": 0.2851,
      "step": 1879
    },
    {
      "epoch": 4.66,
      "learning_rate": 2.536335076857197e-06,
      "loss": 0.3069,
      "step": 1880
    },
    {
      "epoch": 4.67,
      "learning_rate": 2.527333556542878e-06,
      "loss": 0.289,
      "step": 1881
    },
    {
      "epoch": 4.67,
      "learning_rate": 2.518345726907797e-06,
      "loss": 0.2742,
      "step": 1882
    },
    {
      "epoch": 4.67,
      "learning_rate": 2.5093716044185634e-06,
      "loss": 0.2882,
      "step": 1883
    },
    {
      "epoch": 4.67,
      "learning_rate": 2.5004112055166707e-06,
      "loss": 0.2591,
      "step": 1884
    },
    {
      "epoch": 4.68,
      "learning_rate": 2.4914645466184807e-06,
      "loss": 0.3122,
      "step": 1885
    },
    {
      "epoch": 4.68,
      "learning_rate": 2.48253164411517e-06,
      "loss": 0.2887,
      "step": 1886
    },
    {
      "epoch": 4.68,
      "learning_rate": 2.473612514372724e-06,
      "loss": 0.2642,
      "step": 1887
    },
    {
      "epoch": 4.68,
      "learning_rate": 2.4647071737318817e-06,
      "loss": 0.2607,
      "step": 1888
    },
    {
      "epoch": 4.69,
      "learning_rate": 2.4558156385081334e-06,
      "loss": 0.2739,
      "step": 1889
    },
    {
      "epoch": 4.69,
      "learning_rate": 2.4469379249916614e-06,
      "loss": 0.2857,
      "step": 1890
    },
    {
      "epoch": 4.69,
      "learning_rate": 2.4380740494473407e-06,
      "loss": 0.253,
      "step": 1891
    },
    {
      "epoch": 4.69,
      "learning_rate": 2.4292240281146784e-06,
      "loss": 0.2351,
      "step": 1892
    },
    {
      "epoch": 4.7,
      "learning_rate": 2.420387877207815e-06,
      "loss": 0.2715,
      "step": 1893
    },
    {
      "epoch": 4.7,
      "learning_rate": 2.411565612915463e-06,
      "loss": 0.2672,
      "step": 1894
    },
    {
      "epoch": 4.7,
      "learning_rate": 2.402757251400908e-06,
      "loss": 0.2749,
      "step": 1895
    },
    {
      "epoch": 4.7,
      "learning_rate": 2.393962808801952e-06,
      "loss": 0.2735,
      "step": 1896
    },
    {
      "epoch": 4.71,
      "learning_rate": 2.3851823012309026e-06,
      "loss": 0.268,
      "step": 1897
    },
    {
      "epoch": 4.71,
      "learning_rate": 2.376415744774542e-06,
      "loss": 0.2785,
      "step": 1898
    },
    {
      "epoch": 4.71,
      "learning_rate": 2.3676631554940778e-06,
      "loss": 0.2912,
      "step": 1899
    },
    {
      "epoch": 4.71,
      "learning_rate": 2.358924549425143e-06,
      "loss": 0.2835,
      "step": 1900
    },
    {
      "epoch": 4.72,
      "learning_rate": 2.3501999425777433e-06,
      "loss": 0.311,
      "step": 1901
    },
    {
      "epoch": 4.72,
      "learning_rate": 2.3414893509362425e-06,
      "loss": 0.256,
      "step": 1902
    },
    {
      "epoch": 4.72,
      "learning_rate": 2.3327927904593174e-06,
      "loss": 0.3069,
      "step": 1903
    },
    {
      "epoch": 4.72,
      "learning_rate": 2.3241102770799517e-06,
      "loss": 0.2936,
      "step": 1904
    },
    {
      "epoch": 4.73,
      "learning_rate": 2.3154418267053826e-06,
      "loss": 0.2714,
      "step": 1905
    },
    {
      "epoch": 4.73,
      "learning_rate": 2.3067874552170844e-06,
      "loss": 0.271,
      "step": 1906
    },
    {
      "epoch": 4.73,
      "learning_rate": 2.2981471784707444e-06,
      "loss": 0.2742,
      "step": 1907
    },
    {
      "epoch": 4.73,
      "learning_rate": 2.2895210122962166e-06,
      "loss": 0.2494,
      "step": 1908
    },
    {
      "epoch": 4.74,
      "learning_rate": 2.2809089724975133e-06,
      "loss": 0.2568,
      "step": 1909
    },
    {
      "epoch": 4.74,
      "learning_rate": 2.272311074852754e-06,
      "loss": 0.2544,
      "step": 1910
    },
    {
      "epoch": 4.74,
      "learning_rate": 2.2637273351141633e-06,
      "loss": 0.2749,
      "step": 1911
    },
    {
      "epoch": 4.74,
      "learning_rate": 2.255157769008011e-06,
      "loss": 0.2708,
      "step": 1912
    },
    {
      "epoch": 4.75,
      "learning_rate": 2.2466023922346137e-06,
      "loss": 0.2626,
      "step": 1913
    },
    {
      "epoch": 4.75,
      "learning_rate": 2.2380612204682794e-06,
      "loss": 0.2541,
      "step": 1914
    },
    {
      "epoch": 4.75,
      "learning_rate": 2.229534269357303e-06,
      "loss": 0.2696,
      "step": 1915
    },
    {
      "epoch": 4.75,
      "learning_rate": 2.2210215545239132e-06,
      "loss": 0.2645,
      "step": 1916
    },
    {
      "epoch": 4.76,
      "learning_rate": 2.212523091564267e-06,
      "loss": 0.2648,
      "step": 1917
    },
    {
      "epoch": 4.76,
      "learning_rate": 2.2040388960484093e-06,
      "loss": 0.2793,
      "step": 1918
    },
    {
      "epoch": 4.76,
      "learning_rate": 2.1955689835202354e-06,
      "loss": 0.2599,
      "step": 1919
    },
    {
      "epoch": 4.76,
      "learning_rate": 2.1871133694974856e-06,
      "loss": 0.2645,
      "step": 1920
    },
    {
      "epoch": 4.77,
      "learning_rate": 2.1786720694716946e-06,
      "loss": 0.3006,
      "step": 1921
    },
    {
      "epoch": 4.77,
      "learning_rate": 2.1702450989081814e-06,
      "loss": 0.2839,
      "step": 1922
    },
    {
      "epoch": 4.77,
      "learning_rate": 2.1618324732459993e-06,
      "loss": 0.2749,
      "step": 1923
    },
    {
      "epoch": 4.77,
      "learning_rate": 2.1534342078979355e-06,
      "loss": 0.2859,
      "step": 1924
    },
    {
      "epoch": 4.78,
      "learning_rate": 2.1450503182504524e-06,
      "loss": 0.2808,
      "step": 1925
    },
    {
      "epoch": 4.78,
      "learning_rate": 2.136680819663689e-06,
      "loss": 0.285,
      "step": 1926
    },
    {
      "epoch": 4.78,
      "learning_rate": 2.128325727471409e-06,
      "loss": 0.2711,
      "step": 1927
    },
    {
      "epoch": 4.78,
      "learning_rate": 2.11998505698098e-06,
      "loss": 0.2878,
      "step": 1928
    },
    {
      "epoch": 4.79,
      "learning_rate": 2.111658823473359e-06,
      "loss": 0.2647,
      "step": 1929
    },
    {
      "epoch": 4.79,
      "learning_rate": 2.103347042203041e-06,
      "loss": 0.2673,
      "step": 1930
    },
    {
      "epoch": 4.79,
      "learning_rate": 2.095049728398054e-06,
      "loss": 0.3049,
      "step": 1931
    },
    {
      "epoch": 4.79,
      "learning_rate": 2.086766897259909e-06,
      "loss": 0.2942,
      "step": 1932
    },
    {
      "epoch": 4.8,
      "learning_rate": 2.0784985639635924e-06,
      "loss": 0.2909,
      "step": 1933
    },
    {
      "epoch": 4.8,
      "learning_rate": 2.0702447436575223e-06,
      "loss": 0.2877,
      "step": 1934
    },
    {
      "epoch": 4.8,
      "learning_rate": 2.0620054514635344e-06,
      "loss": 0.2526,
      "step": 1935
    },
    {
      "epoch": 4.8,
      "learning_rate": 2.0537807024768395e-06,
      "loss": 0.2728,
      "step": 1936
    },
    {
      "epoch": 4.81,
      "learning_rate": 2.0455705117660107e-06,
      "loss": 0.2775,
      "step": 1937
    },
    {
      "epoch": 4.81,
      "learning_rate": 2.0373748943729477e-06,
      "loss": 0.2762,
      "step": 1938
    },
    {
      "epoch": 4.81,
      "learning_rate": 2.029193865312844e-06,
      "loss": 0.3021,
      "step": 1939
    },
    {
      "epoch": 4.81,
      "learning_rate": 2.021027439574177e-06,
      "loss": 0.2837,
      "step": 1940
    },
    {
      "epoch": 4.82,
      "learning_rate": 2.012875632118656e-06,
      "loss": 0.2613,
      "step": 1941
    },
    {
      "epoch": 4.82,
      "learning_rate": 2.0047384578812214e-06,
      "loss": 0.2688,
      "step": 1942
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.996615931769992e-06,
      "loss": 0.277,
      "step": 1943
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.9885080686662606e-06,
      "loss": 0.2997,
      "step": 1944
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.9804148834244465e-06,
      "loss": 0.2771,
      "step": 1945
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.972336390872087e-06,
      "loss": 0.3046,
      "step": 1946
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.9642726058097906e-06,
      "loss": 0.2623,
      "step": 1947
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.9562235430112308e-06,
      "loss": 0.2573,
      "step": 1948
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.9481892172231e-06,
      "loss": 0.2812,
      "step": 1949
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.940169643165092e-06,
      "loss": 0.2601,
      "step": 1950
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.9321648355298817e-06,
      "loss": 0.2835,
      "step": 1951
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.924174808983078e-06,
      "loss": 0.272,
      "step": 1952
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.9161995781632227e-06,
      "loss": 0.2653,
      "step": 1953
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.9082391576817373e-06,
      "loss": 0.2806,
      "step": 1954
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.9002935621229212e-06,
      "loss": 0.2672,
      "step": 1955
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.8923628060439037e-06,
      "loss": 0.2896,
      "step": 1956
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.8844469039746305e-06,
      "loss": 0.2938,
      "step": 1957
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.8765458704178385e-06,
      "loss": 0.2795,
      "step": 1958
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.868659719849012e-06,
      "loss": 0.2839,
      "step": 1959
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.8607884667163812e-06,
      "loss": 0.2781,
      "step": 1960
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.8529321254408695e-06,
      "loss": 0.2455,
      "step": 1961
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.8450907104160932e-06,
      "loss": 0.2906,
      "step": 1962
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.8372642360083116e-06,
      "loss": 0.32,
      "step": 1963
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.8294527165564202e-06,
      "loss": 0.2773,
      "step": 1964
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.8216561663719068e-06,
      "loss": 0.2779,
      "step": 1965
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.8138745997388429e-06,
      "loss": 0.274,
      "step": 1966
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.8061080309138379e-06,
      "loss": 0.2633,
      "step": 1967
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.7983564741260373e-06,
      "loss": 0.2829,
      "step": 1968
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.7906199435770688e-06,
      "loss": 0.2709,
      "step": 1969
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.7828984534410421e-06,
      "loss": 0.2654,
      "step": 1970
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.7751920178645055e-06,
      "loss": 0.2633,
      "step": 1971
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.7675006509664238e-06,
      "loss": 0.2598,
      "step": 1972
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.7598243668381632e-06,
      "loss": 0.2458,
      "step": 1973
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.7521631795434435e-06,
      "loss": 0.2584,
      "step": 1974
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.7445171031183416e-06,
      "loss": 0.3044,
      "step": 1975
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.736886151571231e-06,
      "loss": 0.2595,
      "step": 1976
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.7292703388827969e-06,
      "loss": 0.2635,
      "step": 1977
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.7216696790059718e-06,
      "loss": 0.2609,
      "step": 1978
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.7140841858659274e-06,
      "loss": 0.2594,
      "step": 1979
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.7065138733600605e-06,
      "loss": 0.2599,
      "step": 1980
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.6989587553579423e-06,
      "loss": 0.2875,
      "step": 1981
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.6914188457013148e-06,
      "loss": 0.2605,
      "step": 1982
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.6838941582040503e-06,
      "loss": 0.2839,
      "step": 1983
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.67638470665214e-06,
      "loss": 0.2709,
      "step": 1984
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.668890504803653e-06,
      "loss": 0.2691,
      "step": 1985
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.6614115663887277e-06,
      "loss": 0.2662,
      "step": 1986
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.6539479051095298e-06,
      "loss": 0.2768,
      "step": 1987
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.6464995346402445e-06,
      "loss": 0.2761,
      "step": 1988
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.6390664686270342e-06,
      "loss": 0.3066,
      "step": 1989
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.6316487206880306e-06,
      "loss": 0.3139,
      "step": 1990
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.6242463044132916e-06,
      "loss": 0.2958,
      "step": 1991
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.616859233364797e-06,
      "loss": 0.2647,
      "step": 1992
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.6094875210764027e-06,
      "loss": 0.2694,
      "step": 1993
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.6021311810538287e-06,
      "loss": 0.2694,
      "step": 1994
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.5947902267746363e-06,
      "loss": 0.2889,
      "step": 1995
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.587464671688187e-06,
      "loss": 0.2577,
      "step": 1996
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.5801545292156483e-06,
      "loss": 0.2873,
      "step": 1997
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.5728598127499318e-06,
      "loss": 0.2659,
      "step": 1998
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.5655805356556985e-06,
      "loss": 0.2697,
      "step": 1999
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.5583167112693153e-06,
      "loss": 0.2685,
      "step": 2000
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.5510683528988424e-06,
      "loss": 0.2673,
      "step": 2001
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.5438354738240057e-06,
      "loss": 0.2674,
      "step": 2002
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.5366180872961677e-06,
      "loss": 0.2664,
      "step": 2003
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.529416206538311e-06,
      "loss": 0.232,
      "step": 2004
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.5222298447450058e-06,
      "loss": 0.2732,
      "step": 2005
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.5150590150823951e-06,
      "loss": 0.2509,
      "step": 2006
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.5079037306881605e-06,
      "loss": 0.2695,
      "step": 2007
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.5007640046715067e-06,
      "loss": 0.2891,
      "step": 2008
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.49363985011313e-06,
      "loss": 0.305,
      "step": 2009
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.4865312800652055e-06,
      "loss": 0.2841,
      "step": 2010
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.4794383075513453e-06,
      "loss": 0.2987,
      "step": 2011
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.4723609455665966e-06,
      "loss": 0.2767,
      "step": 2012
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.4652992070773953e-06,
      "loss": 0.2832,
      "step": 2013
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.4582531050215643e-06,
      "loss": 0.3071,
      "step": 2014
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.4512226523082696e-06,
      "loss": 0.2665,
      "step": 2015
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.4442078618180122e-06,
      "loss": 0.1892,
      "step": 2016
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.4372087464025985e-06,
      "loss": 0.1629,
      "step": 2017
    },
    {
      "epoch": 5.01,
      "learning_rate": 1.43022531888511e-06,
      "loss": 0.1574,
      "step": 2018
    },
    {
      "epoch": 5.01,
      "learning_rate": 1.4232575920598947e-06,
      "loss": 0.1677,
      "step": 2019
    },
    {
      "epoch": 5.01,
      "learning_rate": 1.4163055786925284e-06,
      "loss": 0.174,
      "step": 2020
    },
    {
      "epoch": 5.01,
      "learning_rate": 1.4093692915198054e-06,
      "loss": 0.1661,
      "step": 2021
    },
    {
      "epoch": 5.02,
      "learning_rate": 1.4024487432497013e-06,
      "loss": 0.1938,
      "step": 2022
    },
    {
      "epoch": 5.02,
      "learning_rate": 1.3955439465613573e-06,
      "loss": 0.1684,
      "step": 2023
    },
    {
      "epoch": 5.02,
      "learning_rate": 1.388654914105061e-06,
      "loss": 0.1488,
      "step": 2024
    },
    {
      "epoch": 5.02,
      "learning_rate": 1.3817816585022126e-06,
      "loss": 0.1351,
      "step": 2025
    },
    {
      "epoch": 5.03,
      "learning_rate": 1.3749241923453138e-06,
      "loss": 0.1665,
      "step": 2026
    },
    {
      "epoch": 5.03,
      "learning_rate": 1.3680825281979315e-06,
      "loss": 0.1396,
      "step": 2027
    },
    {
      "epoch": 5.03,
      "learning_rate": 1.3612566785946868e-06,
      "loss": 0.1608,
      "step": 2028
    },
    {
      "epoch": 5.03,
      "learning_rate": 1.354446656041224e-06,
      "loss": 0.1731,
      "step": 2029
    },
    {
      "epoch": 5.04,
      "learning_rate": 1.3476524730141938e-06,
      "loss": 0.1506,
      "step": 2030
    },
    {
      "epoch": 5.04,
      "learning_rate": 1.3408741419612215e-06,
      "loss": 0.158,
      "step": 2031
    },
    {
      "epoch": 5.04,
      "learning_rate": 1.3341116753008997e-06,
      "loss": 0.1483,
      "step": 2032
    },
    {
      "epoch": 5.04,
      "learning_rate": 1.3273650854227438e-06,
      "loss": 0.1654,
      "step": 2033
    },
    {
      "epoch": 5.05,
      "learning_rate": 1.3206343846871916e-06,
      "loss": 0.1758,
      "step": 2034
    },
    {
      "epoch": 5.05,
      "learning_rate": 1.3139195854255638e-06,
      "loss": 0.1474,
      "step": 2035
    },
    {
      "epoch": 5.05,
      "learning_rate": 1.3072206999400505e-06,
      "loss": 0.1514,
      "step": 2036
    },
    {
      "epoch": 5.05,
      "learning_rate": 1.30053774050369e-06,
      "loss": 0.135,
      "step": 2037
    },
    {
      "epoch": 5.06,
      "learning_rate": 1.293870719360334e-06,
      "loss": 0.1683,
      "step": 2038
    },
    {
      "epoch": 5.06,
      "learning_rate": 1.2872196487246414e-06,
      "loss": 0.1558,
      "step": 2039
    },
    {
      "epoch": 5.06,
      "learning_rate": 1.2805845407820417e-06,
      "loss": 0.1718,
      "step": 2040
    },
    {
      "epoch": 5.06,
      "learning_rate": 1.2739654076887265e-06,
      "loss": 0.1607,
      "step": 2041
    },
    {
      "epoch": 5.07,
      "learning_rate": 1.2673622615716119e-06,
      "loss": 0.1549,
      "step": 2042
    },
    {
      "epoch": 5.07,
      "learning_rate": 1.2607751145283309e-06,
      "loss": 0.1317,
      "step": 2043
    },
    {
      "epoch": 5.07,
      "learning_rate": 1.2542039786272008e-06,
      "loss": 0.1453,
      "step": 2044
    },
    {
      "epoch": 5.07,
      "learning_rate": 1.247648865907205e-06,
      "loss": 0.1507,
      "step": 2045
    },
    {
      "epoch": 5.08,
      "learning_rate": 1.2411097883779732e-06,
      "loss": 0.1511,
      "step": 2046
    },
    {
      "epoch": 5.08,
      "learning_rate": 1.2345867580197535e-06,
      "loss": 0.1589,
      "step": 2047
    },
    {
      "epoch": 5.08,
      "learning_rate": 1.2280797867833982e-06,
      "loss": 0.1553,
      "step": 2048
    },
    {
      "epoch": 5.08,
      "learning_rate": 1.2215888865903337e-06,
      "loss": 0.1796,
      "step": 2049
    },
    {
      "epoch": 5.09,
      "learning_rate": 1.215114069332546e-06,
      "loss": 0.2041,
      "step": 2050
    },
    {
      "epoch": 5.09,
      "learning_rate": 1.2086553468725515e-06,
      "loss": 0.1559,
      "step": 2051
    },
    {
      "epoch": 5.09,
      "learning_rate": 1.202212731043385e-06,
      "loss": 0.1726,
      "step": 2052
    },
    {
      "epoch": 5.09,
      "learning_rate": 1.195786233648566e-06,
      "loss": 0.1597,
      "step": 2053
    },
    {
      "epoch": 5.1,
      "learning_rate": 1.1893758664620892e-06,
      "loss": 0.1561,
      "step": 2054
    },
    {
      "epoch": 5.1,
      "learning_rate": 1.1829816412283912e-06,
      "loss": 0.1593,
      "step": 2055
    },
    {
      "epoch": 5.1,
      "learning_rate": 1.176603569662339e-06,
      "loss": 0.1476,
      "step": 2056
    },
    {
      "epoch": 5.1,
      "learning_rate": 1.1702416634492075e-06,
      "loss": 0.1608,
      "step": 2057
    },
    {
      "epoch": 5.11,
      "learning_rate": 1.1638959342446454e-06,
      "loss": 0.1622,
      "step": 2058
    },
    {
      "epoch": 5.11,
      "learning_rate": 1.1575663936746729e-06,
      "loss": 0.1699,
      "step": 2059
    },
    {
      "epoch": 5.11,
      "learning_rate": 1.1512530533356448e-06,
      "loss": 0.1395,
      "step": 2060
    },
    {
      "epoch": 5.11,
      "learning_rate": 1.14495592479424e-06,
      "loss": 0.162,
      "step": 2061
    },
    {
      "epoch": 5.12,
      "learning_rate": 1.1386750195874319e-06,
      "loss": 0.1683,
      "step": 2062
    },
    {
      "epoch": 5.12,
      "learning_rate": 1.1324103492224748e-06,
      "loss": 0.1714,
      "step": 2063
    },
    {
      "epoch": 5.12,
      "learning_rate": 1.126161925176874e-06,
      "loss": 0.156,
      "step": 2064
    },
    {
      "epoch": 5.12,
      "learning_rate": 1.1199297588983771e-06,
      "loss": 0.1745,
      "step": 2065
    },
    {
      "epoch": 5.13,
      "learning_rate": 1.1137138618049403e-06,
      "loss": 0.1523,
      "step": 2066
    },
    {
      "epoch": 5.13,
      "learning_rate": 1.107514245284712e-06,
      "loss": 0.1526,
      "step": 2067
    },
    {
      "epoch": 5.13,
      "learning_rate": 1.10133092069602e-06,
      "loss": 0.1558,
      "step": 2068
    },
    {
      "epoch": 5.13,
      "learning_rate": 1.0951638993673331e-06,
      "loss": 0.168,
      "step": 2069
    },
    {
      "epoch": 5.14,
      "learning_rate": 1.0890131925972635e-06,
      "loss": 0.1654,
      "step": 2070
    },
    {
      "epoch": 5.14,
      "learning_rate": 1.082878811654522e-06,
      "loss": 0.1457,
      "step": 2071
    },
    {
      "epoch": 5.14,
      "learning_rate": 1.0767607677779168e-06,
      "loss": 0.1915,
      "step": 2072
    },
    {
      "epoch": 5.14,
      "learning_rate": 1.0706590721763188e-06,
      "loss": 0.152,
      "step": 2073
    },
    {
      "epoch": 5.15,
      "learning_rate": 1.0645737360286513e-06,
      "loss": 0.1552,
      "step": 2074
    },
    {
      "epoch": 5.15,
      "learning_rate": 1.058504770483868e-06,
      "loss": 0.1584,
      "step": 2075
    },
    {
      "epoch": 5.15,
      "learning_rate": 1.0524521866609206e-06,
      "loss": 0.1426,
      "step": 2076
    },
    {
      "epoch": 5.15,
      "learning_rate": 1.0464159956487596e-06,
      "loss": 0.185,
      "step": 2077
    },
    {
      "epoch": 5.16,
      "learning_rate": 1.0403962085062903e-06,
      "loss": 0.1643,
      "step": 2078
    },
    {
      "epoch": 5.16,
      "learning_rate": 1.0343928362623757e-06,
      "loss": 0.1851,
      "step": 2079
    },
    {
      "epoch": 5.16,
      "learning_rate": 1.0284058899157956e-06,
      "loss": 0.1669,
      "step": 2080
    },
    {
      "epoch": 5.16,
      "learning_rate": 1.0224353804352428e-06,
      "loss": 0.1393,
      "step": 2081
    },
    {
      "epoch": 5.17,
      "learning_rate": 1.0164813187592915e-06,
      "loss": 0.1633,
      "step": 2082
    },
    {
      "epoch": 5.17,
      "learning_rate": 1.0105437157963848e-06,
      "loss": 0.147,
      "step": 2083
    },
    {
      "epoch": 5.17,
      "learning_rate": 1.0046225824248102e-06,
      "loss": 0.1626,
      "step": 2084
    },
    {
      "epoch": 5.17,
      "learning_rate": 9.98717929492683e-07,
      "loss": 0.1536,
      "step": 2085
    },
    {
      "epoch": 5.18,
      "learning_rate": 9.9282976781792e-07,
      "loss": 0.1695,
      "step": 2086
    },
    {
      "epoch": 5.18,
      "learning_rate": 9.869581081882329e-07,
      "loss": 0.147,
      "step": 2087
    },
    {
      "epoch": 5.18,
      "learning_rate": 9.811029613610913e-07,
      "loss": 0.1652,
      "step": 2088
    },
    {
      "epoch": 5.18,
      "learning_rate": 9.752643380637151e-07,
      "loss": 0.1689,
      "step": 2089
    },
    {
      "epoch": 5.19,
      "learning_rate": 9.694422489930533e-07,
      "loss": 0.1501,
      "step": 2090
    },
    {
      "epoch": 5.19,
      "learning_rate": 9.63636704815758e-07,
      "loss": 0.1634,
      "step": 2091
    },
    {
      "epoch": 5.19,
      "learning_rate": 9.578477161681765e-07,
      "loss": 0.1523,
      "step": 2092
    },
    {
      "epoch": 5.19,
      "learning_rate": 9.520752936563149e-07,
      "loss": 0.1467,
      "step": 2093
    },
    {
      "epoch": 5.2,
      "learning_rate": 9.463194478558379e-07,
      "loss": 0.1697,
      "step": 2094
    },
    {
      "epoch": 5.2,
      "learning_rate": 9.405801893120347e-07,
      "loss": 0.1556,
      "step": 2095
    },
    {
      "epoch": 5.2,
      "learning_rate": 9.348575285398087e-07,
      "loss": 0.1656,
      "step": 2096
    },
    {
      "epoch": 5.2,
      "learning_rate": 9.291514760236498e-07,
      "loss": 0.1595,
      "step": 2097
    },
    {
      "epoch": 5.21,
      "learning_rate": 9.234620422176211e-07,
      "loss": 0.1582,
      "step": 2098
    },
    {
      "epoch": 5.21,
      "learning_rate": 9.177892375453413e-07,
      "loss": 0.1498,
      "step": 2099
    },
    {
      "epoch": 5.21,
      "learning_rate": 9.1213307239996e-07,
      "loss": 0.173,
      "step": 2100
    },
    {
      "epoch": 5.21,
      "learning_rate": 9.064935571441458e-07,
      "loss": 0.1488,
      "step": 2101
    },
    {
      "epoch": 5.22,
      "learning_rate": 9.008707021100549e-07,
      "loss": 0.1482,
      "step": 2102
    },
    {
      "epoch": 5.22,
      "learning_rate": 8.952645175993302e-07,
      "loss": 0.1672,
      "step": 2103
    },
    {
      "epoch": 5.22,
      "learning_rate": 8.896750138830646e-07,
      "loss": 0.1542,
      "step": 2104
    },
    {
      "epoch": 5.22,
      "learning_rate": 8.841022012017941e-07,
      "loss": 0.1574,
      "step": 2105
    },
    {
      "epoch": 5.22,
      "learning_rate": 8.785460897654729e-07,
      "loss": 0.1461,
      "step": 2106
    },
    {
      "epoch": 5.23,
      "learning_rate": 8.730066897534617e-07,
      "loss": 0.1521,
      "step": 2107
    },
    {
      "epoch": 5.23,
      "learning_rate": 8.674840113144956e-07,
      "loss": 0.1549,
      "step": 2108
    },
    {
      "epoch": 5.23,
      "learning_rate": 8.61978064566682e-07,
      "loss": 0.1624,
      "step": 2109
    },
    {
      "epoch": 5.23,
      "learning_rate": 8.564888595974718e-07,
      "loss": 0.1459,
      "step": 2110
    },
    {
      "epoch": 5.24,
      "learning_rate": 8.510164064636384e-07,
      "loss": 0.1594,
      "step": 2111
    },
    {
      "epoch": 5.24,
      "learning_rate": 8.455607151912725e-07,
      "loss": 0.1577,
      "step": 2112
    },
    {
      "epoch": 5.24,
      "learning_rate": 8.401217957757491e-07,
      "loss": 0.1486,
      "step": 2113
    },
    {
      "epoch": 5.24,
      "learning_rate": 8.34699658181718e-07,
      "loss": 0.1532,
      "step": 2114
    },
    {
      "epoch": 5.25,
      "learning_rate": 8.29294312343083e-07,
      "loss": 0.1494,
      "step": 2115
    },
    {
      "epoch": 5.25,
      "learning_rate": 8.239057681629847e-07,
      "loss": 0.1726,
      "step": 2116
    },
    {
      "epoch": 5.25,
      "learning_rate": 8.185340355137772e-07,
      "loss": 0.1643,
      "step": 2117
    },
    {
      "epoch": 5.25,
      "learning_rate": 8.131791242370191e-07,
      "loss": 0.1577,
      "step": 2118
    },
    {
      "epoch": 5.26,
      "learning_rate": 8.078410441434481e-07,
      "loss": 0.1515,
      "step": 2119
    },
    {
      "epoch": 5.26,
      "learning_rate": 8.025198050129634e-07,
      "loss": 0.1618,
      "step": 2120
    },
    {
      "epoch": 5.26,
      "learning_rate": 7.972154165946155e-07,
      "loss": 0.1407,
      "step": 2121
    },
    {
      "epoch": 5.26,
      "learning_rate": 7.919278886065762e-07,
      "loss": 0.1512,
      "step": 2122
    },
    {
      "epoch": 5.27,
      "learning_rate": 7.866572307361354e-07,
      "loss": 0.1647,
      "step": 2123
    },
    {
      "epoch": 5.27,
      "learning_rate": 7.814034526396641e-07,
      "loss": 0.1499,
      "step": 2124
    },
    {
      "epoch": 5.27,
      "learning_rate": 7.761665639426186e-07,
      "loss": 0.1608,
      "step": 2125
    },
    {
      "epoch": 5.27,
      "learning_rate": 7.70946574239505e-07,
      "loss": 0.1599,
      "step": 2126
    },
    {
      "epoch": 5.28,
      "learning_rate": 7.657434930938745e-07,
      "loss": 0.1395,
      "step": 2127
    },
    {
      "epoch": 5.28,
      "learning_rate": 7.60557330038294e-07,
      "loss": 0.1459,
      "step": 2128
    },
    {
      "epoch": 5.28,
      "learning_rate": 7.553880945743397e-07,
      "loss": 0.1468,
      "step": 2129
    },
    {
      "epoch": 5.28,
      "learning_rate": 7.502357961725703e-07,
      "loss": 0.1463,
      "step": 2130
    },
    {
      "epoch": 5.29,
      "learning_rate": 7.451004442725196e-07,
      "loss": 0.1517,
      "step": 2131
    },
    {
      "epoch": 5.29,
      "learning_rate": 7.399820482826692e-07,
      "loss": 0.1527,
      "step": 2132
    },
    {
      "epoch": 5.29,
      "learning_rate": 7.348806175804335e-07,
      "loss": 0.1589,
      "step": 2133
    },
    {
      "epoch": 5.29,
      "learning_rate": 7.297961615121563e-07,
      "loss": 0.1632,
      "step": 2134
    },
    {
      "epoch": 5.3,
      "learning_rate": 7.247286893930683e-07,
      "loss": 0.1431,
      "step": 2135
    },
    {
      "epoch": 5.3,
      "learning_rate": 7.19678210507293e-07,
      "loss": 0.1689,
      "step": 2136
    },
    {
      "epoch": 5.3,
      "learning_rate": 7.146447341078155e-07,
      "loss": 0.1684,
      "step": 2137
    },
    {
      "epoch": 5.3,
      "learning_rate": 7.096282694164747e-07,
      "loss": 0.166,
      "step": 2138
    },
    {
      "epoch": 5.31,
      "learning_rate": 7.046288256239386e-07,
      "loss": 0.1598,
      "step": 2139
    },
    {
      "epoch": 5.31,
      "learning_rate": 6.996464118896939e-07,
      "loss": 0.1707,
      "step": 2140
    },
    {
      "epoch": 5.31,
      "learning_rate": 6.946810373420254e-07,
      "loss": 0.147,
      "step": 2141
    },
    {
      "epoch": 5.31,
      "learning_rate": 6.897327110779983e-07,
      "loss": 0.1489,
      "step": 2142
    },
    {
      "epoch": 5.32,
      "learning_rate": 6.848014421634497e-07,
      "loss": 0.1628,
      "step": 2143
    },
    {
      "epoch": 5.32,
      "learning_rate": 6.798872396329581e-07,
      "loss": 0.1637,
      "step": 2144
    },
    {
      "epoch": 5.32,
      "learning_rate": 6.749901124898417e-07,
      "loss": 0.1659,
      "step": 2145
    },
    {
      "epoch": 5.32,
      "learning_rate": 6.701100697061302e-07,
      "loss": 0.1671,
      "step": 2146
    },
    {
      "epoch": 5.33,
      "learning_rate": 6.652471202225552e-07,
      "loss": 0.158,
      "step": 2147
    },
    {
      "epoch": 5.33,
      "learning_rate": 6.604012729485299e-07,
      "loss": 0.1457,
      "step": 2148
    },
    {
      "epoch": 5.33,
      "learning_rate": 6.555725367621379e-07,
      "loss": 0.1611,
      "step": 2149
    },
    {
      "epoch": 5.33,
      "learning_rate": 6.507609205101073e-07,
      "loss": 0.1555,
      "step": 2150
    },
    {
      "epoch": 5.34,
      "learning_rate": 6.459664330078108e-07,
      "loss": 0.1601,
      "step": 2151
    },
    {
      "epoch": 5.34,
      "learning_rate": 6.411890830392276e-07,
      "loss": 0.1601,
      "step": 2152
    },
    {
      "epoch": 5.34,
      "learning_rate": 6.364288793569473e-07,
      "loss": 0.1296,
      "step": 2153
    },
    {
      "epoch": 5.34,
      "learning_rate": 6.316858306821449e-07,
      "loss": 0.164,
      "step": 2154
    },
    {
      "epoch": 5.35,
      "learning_rate": 6.269599457045627e-07,
      "loss": 0.1624,
      "step": 2155
    },
    {
      "epoch": 5.35,
      "learning_rate": 6.222512330825003e-07,
      "loss": 0.1376,
      "step": 2156
    },
    {
      "epoch": 5.35,
      "learning_rate": 6.175597014427926e-07,
      "loss": 0.156,
      "step": 2157
    },
    {
      "epoch": 5.35,
      "learning_rate": 6.128853593808026e-07,
      "loss": 0.1787,
      "step": 2158
    },
    {
      "epoch": 5.36,
      "learning_rate": 6.082282154603925e-07,
      "loss": 0.1589,
      "step": 2159
    },
    {
      "epoch": 5.36,
      "learning_rate": 6.035882782139235e-07,
      "loss": 0.1369,
      "step": 2160
    },
    {
      "epoch": 5.36,
      "learning_rate": 5.98965556142227e-07,
      "loss": 0.1635,
      "step": 2161
    },
    {
      "epoch": 5.36,
      "learning_rate": 5.943600577145958e-07,
      "loss": 0.1499,
      "step": 2162
    },
    {
      "epoch": 5.37,
      "learning_rate": 5.897717913687717e-07,
      "loss": 0.1572,
      "step": 2163
    },
    {
      "epoch": 5.37,
      "learning_rate": 5.852007655109171e-07,
      "loss": 0.1554,
      "step": 2164
    },
    {
      "epoch": 5.37,
      "learning_rate": 5.806469885156163e-07,
      "loss": 0.1508,
      "step": 2165
    },
    {
      "epoch": 5.37,
      "learning_rate": 5.761104687258457e-07,
      "loss": 0.1446,
      "step": 2166
    },
    {
      "epoch": 5.38,
      "learning_rate": 5.715912144529701e-07,
      "loss": 0.1699,
      "step": 2167
    },
    {
      "epoch": 5.38,
      "learning_rate": 5.670892339767176e-07,
      "loss": 0.1688,
      "step": 2168
    },
    {
      "epoch": 5.38,
      "learning_rate": 5.626045355451726e-07,
      "loss": 0.1498,
      "step": 2169
    },
    {
      "epoch": 5.38,
      "learning_rate": 5.581371273747527e-07,
      "loss": 0.1564,
      "step": 2170
    },
    {
      "epoch": 5.39,
      "learning_rate": 5.536870176502041e-07,
      "loss": 0.1677,
      "step": 2171
    },
    {
      "epoch": 5.39,
      "learning_rate": 5.492542145245727e-07,
      "loss": 0.1609,
      "step": 2172
    },
    {
      "epoch": 5.39,
      "learning_rate": 5.448387261192046e-07,
      "loss": 0.1318,
      "step": 2173
    },
    {
      "epoch": 5.39,
      "learning_rate": 5.404405605237206e-07,
      "loss": 0.1536,
      "step": 2174
    },
    {
      "epoch": 5.4,
      "learning_rate": 5.36059725796001e-07,
      "loss": 0.1566,
      "step": 2175
    },
    {
      "epoch": 5.4,
      "learning_rate": 5.316962299621808e-07,
      "loss": 0.1897,
      "step": 2176
    },
    {
      "epoch": 5.4,
      "learning_rate": 5.27350081016621e-07,
      "loss": 0.1761,
      "step": 2177
    },
    {
      "epoch": 5.4,
      "learning_rate": 5.230212869219097e-07,
      "loss": 0.1681,
      "step": 2178
    },
    {
      "epoch": 5.41,
      "learning_rate": 5.187098556088299e-07,
      "loss": 0.1737,
      "step": 2179
    },
    {
      "epoch": 5.41,
      "learning_rate": 5.14415794976364e-07,
      "loss": 0.1694,
      "step": 2180
    },
    {
      "epoch": 5.41,
      "learning_rate": 5.101391128916589e-07,
      "loss": 0.1774,
      "step": 2181
    },
    {
      "epoch": 5.41,
      "learning_rate": 5.058798171900325e-07,
      "loss": 0.1511,
      "step": 2182
    },
    {
      "epoch": 5.42,
      "learning_rate": 5.01637915674944e-07,
      "loss": 0.1535,
      "step": 2183
    },
    {
      "epoch": 5.42,
      "learning_rate": 4.974134161179811e-07,
      "loss": 0.1535,
      "step": 2184
    },
    {
      "epoch": 5.42,
      "learning_rate": 4.932063262588582e-07,
      "loss": 0.1585,
      "step": 2185
    },
    {
      "epoch": 5.42,
      "learning_rate": 4.890166538053853e-07,
      "loss": 0.1637,
      "step": 2186
    },
    {
      "epoch": 5.43,
      "learning_rate": 4.848444064334679e-07,
      "loss": 0.1534,
      "step": 2187
    },
    {
      "epoch": 5.43,
      "learning_rate": 4.806895917870813e-07,
      "loss": 0.1812,
      "step": 2188
    },
    {
      "epoch": 5.43,
      "learning_rate": 4.7655221747826885e-07,
      "loss": 0.1688,
      "step": 2189
    },
    {
      "epoch": 5.43,
      "learning_rate": 4.7243229108711485e-07,
      "loss": 0.1501,
      "step": 2190
    },
    {
      "epoch": 5.44,
      "learning_rate": 4.683298201617426e-07,
      "loss": 0.1578,
      "step": 2191
    },
    {
      "epoch": 5.44,
      "learning_rate": 4.642448122182908e-07,
      "loss": 0.1588,
      "step": 2192
    },
    {
      "epoch": 5.44,
      "learning_rate": 4.6017727474090723e-07,
      "loss": 0.1523,
      "step": 2193
    },
    {
      "epoch": 5.44,
      "learning_rate": 4.561272151817331e-07,
      "loss": 0.1799,
      "step": 2194
    },
    {
      "epoch": 5.45,
      "learning_rate": 4.520946409608828e-07,
      "loss": 0.1462,
      "step": 2195
    },
    {
      "epoch": 5.45,
      "learning_rate": 4.480795594664433e-07,
      "loss": 0.1616,
      "step": 2196
    },
    {
      "epoch": 5.45,
      "learning_rate": 4.4408197805444587e-07,
      "loss": 0.1615,
      "step": 2197
    },
    {
      "epoch": 5.45,
      "learning_rate": 4.401019040488652e-07,
      "loss": 0.1814,
      "step": 2198
    },
    {
      "epoch": 5.46,
      "learning_rate": 4.3613934474159623e-07,
      "loss": 0.1727,
      "step": 2199
    },
    {
      "epoch": 5.46,
      "learning_rate": 4.321943073924506e-07,
      "loss": 0.16,
      "step": 2200
    },
    {
      "epoch": 5.46,
      "learning_rate": 4.282667992291334e-07,
      "loss": 0.149,
      "step": 2201
    },
    {
      "epoch": 5.46,
      "learning_rate": 4.243568274472365e-07,
      "loss": 0.1541,
      "step": 2202
    },
    {
      "epoch": 5.47,
      "learning_rate": 4.204643992102209e-07,
      "loss": 0.1511,
      "step": 2203
    },
    {
      "epoch": 5.47,
      "learning_rate": 4.16589521649412e-07,
      "loss": 0.1847,
      "step": 2204
    },
    {
      "epoch": 5.47,
      "learning_rate": 4.1273220186397457e-07,
      "loss": 0.1657,
      "step": 2205
    },
    {
      "epoch": 5.47,
      "learning_rate": 4.088924469209055e-07,
      "loss": 0.1564,
      "step": 2206
    },
    {
      "epoch": 5.48,
      "learning_rate": 4.0507026385502747e-07,
      "loss": 0.1423,
      "step": 2207
    },
    {
      "epoch": 5.48,
      "learning_rate": 4.0126565966896234e-07,
      "loss": 0.1603,
      "step": 2208
    },
    {
      "epoch": 5.48,
      "learning_rate": 3.974786413331311e-07,
      "loss": 0.1602,
      "step": 2209
    },
    {
      "epoch": 5.48,
      "learning_rate": 3.937092157857325e-07,
      "loss": 0.153,
      "step": 2210
    },
    {
      "epoch": 5.49,
      "learning_rate": 3.899573899327358e-07,
      "loss": 0.1364,
      "step": 2211
    },
    {
      "epoch": 5.49,
      "learning_rate": 3.862231706478625e-07,
      "loss": 0.1727,
      "step": 2212
    },
    {
      "epoch": 5.49,
      "learning_rate": 3.825065647725823e-07,
      "loss": 0.1663,
      "step": 2213
    },
    {
      "epoch": 5.49,
      "learning_rate": 3.788075791160917e-07,
      "loss": 0.184,
      "step": 2214
    },
    {
      "epoch": 5.5,
      "learning_rate": 3.75126220455303e-07,
      "loss": 0.1686,
      "step": 2215
    },
    {
      "epoch": 5.5,
      "learning_rate": 3.7146249553484e-07,
      "loss": 0.1438,
      "step": 2216
    },
    {
      "epoch": 5.5,
      "learning_rate": 3.678164110670135e-07,
      "loss": 0.1522,
      "step": 2217
    },
    {
      "epoch": 5.5,
      "learning_rate": 3.641879737318221e-07,
      "loss": 0.1668,
      "step": 2218
    },
    {
      "epoch": 5.51,
      "learning_rate": 3.605771901769273e-07,
      "loss": 0.1553,
      "step": 2219
    },
    {
      "epoch": 5.51,
      "learning_rate": 3.569840670176483e-07,
      "loss": 0.137,
      "step": 2220
    },
    {
      "epoch": 5.51,
      "learning_rate": 3.5340861083695053e-07,
      "loss": 0.1514,
      "step": 2221
    },
    {
      "epoch": 5.51,
      "learning_rate": 3.498508281854307e-07,
      "loss": 0.1235,
      "step": 2222
    },
    {
      "epoch": 5.52,
      "learning_rate": 3.4631072558130364e-07,
      "loss": 0.1514,
      "step": 2223
    },
    {
      "epoch": 5.52,
      "learning_rate": 3.427883095103979e-07,
      "loss": 0.1423,
      "step": 2224
    },
    {
      "epoch": 5.52,
      "learning_rate": 3.392835864261335e-07,
      "loss": 0.13,
      "step": 2225
    },
    {
      "epoch": 5.52,
      "learning_rate": 3.3579656274951743e-07,
      "loss": 0.1526,
      "step": 2226
    },
    {
      "epoch": 5.53,
      "learning_rate": 3.3232724486912946e-07,
      "loss": 0.1528,
      "step": 2227
    },
    {
      "epoch": 5.53,
      "learning_rate": 3.288756391411074e-07,
      "loss": 0.1705,
      "step": 2228
    },
    {
      "epoch": 5.53,
      "learning_rate": 3.254417518891451e-07,
      "loss": 0.1526,
      "step": 2229
    },
    {
      "epoch": 5.53,
      "learning_rate": 3.2202558940446884e-07,
      "loss": 0.1519,
      "step": 2230
    },
    {
      "epoch": 5.54,
      "learning_rate": 3.186271579458333e-07,
      "loss": 0.1411,
      "step": 2231
    },
    {
      "epoch": 5.54,
      "learning_rate": 3.152464637395058e-07,
      "loss": 0.1729,
      "step": 2232
    },
    {
      "epoch": 5.54,
      "learning_rate": 3.11883512979263e-07,
      "loss": 0.1637,
      "step": 2233
    },
    {
      "epoch": 5.54,
      "learning_rate": 3.085383118263685e-07,
      "loss": 0.1471,
      "step": 2234
    },
    {
      "epoch": 5.55,
      "learning_rate": 3.0521086640956985e-07,
      "loss": 0.1559,
      "step": 2235
    },
    {
      "epoch": 5.55,
      "learning_rate": 3.0190118282508174e-07,
      "loss": 0.1361,
      "step": 2236
    },
    {
      "epoch": 5.55,
      "learning_rate": 2.98609267136577e-07,
      "loss": 0.1512,
      "step": 2237
    },
    {
      "epoch": 5.55,
      "learning_rate": 2.9533512537518016e-07,
      "loss": 0.1532,
      "step": 2238
    },
    {
      "epoch": 5.55,
      "learning_rate": 2.920787635394484e-07,
      "loss": 0.1828,
      "step": 2239
    },
    {
      "epoch": 5.56,
      "learning_rate": 2.8884018759536503e-07,
      "loss": 0.1626,
      "step": 2240
    },
    {
      "epoch": 5.56,
      "learning_rate": 2.856194034763271e-07,
      "loss": 0.1586,
      "step": 2241
    },
    {
      "epoch": 5.56,
      "learning_rate": 2.824164170831389e-07,
      "loss": 0.156,
      "step": 2242
    },
    {
      "epoch": 5.56,
      "learning_rate": 2.792312342839931e-07,
      "loss": 0.146,
      "step": 2243
    },
    {
      "epoch": 5.57,
      "learning_rate": 2.760638609144661e-07,
      "loss": 0.1659,
      "step": 2244
    },
    {
      "epoch": 5.57,
      "learning_rate": 2.7291430277750495e-07,
      "loss": 0.1673,
      "step": 2245
    },
    {
      "epoch": 5.57,
      "learning_rate": 2.6978256564342053e-07,
      "loss": 0.1591,
      "step": 2246
    },
    {
      "epoch": 5.57,
      "learning_rate": 2.666686552498687e-07,
      "loss": 0.1838,
      "step": 2247
    },
    {
      "epoch": 5.58,
      "learning_rate": 2.6357257730185046e-07,
      "loss": 0.1593,
      "step": 2248
    },
    {
      "epoch": 5.58,
      "learning_rate": 2.604943374716906e-07,
      "loss": 0.1671,
      "step": 2249
    },
    {
      "epoch": 5.58,
      "learning_rate": 2.574339413990357e-07,
      "loss": 0.1508,
      "step": 2250
    },
    {
      "epoch": 5.58,
      "learning_rate": 2.543913946908405e-07,
      "loss": 0.1442,
      "step": 2251
    },
    {
      "epoch": 5.59,
      "learning_rate": 2.5136670292135734e-07,
      "loss": 0.1502,
      "step": 2252
    },
    {
      "epoch": 5.59,
      "learning_rate": 2.483598716321289e-07,
      "loss": 0.1587,
      "step": 2253
    },
    {
      "epoch": 5.59,
      "learning_rate": 2.4537090633197094e-07,
      "loss": 0.1726,
      "step": 2254
    },
    {
      "epoch": 5.59,
      "learning_rate": 2.4239981249697176e-07,
      "loss": 0.1527,
      "step": 2255
    },
    {
      "epoch": 5.6,
      "learning_rate": 2.394465955704739e-07,
      "loss": 0.1595,
      "step": 2256
    },
    {
      "epoch": 5.6,
      "learning_rate": 2.3651126096306932e-07,
      "loss": 0.1482,
      "step": 2257
    },
    {
      "epoch": 5.6,
      "learning_rate": 2.335938140525873e-07,
      "loss": 0.1578,
      "step": 2258
    },
    {
      "epoch": 5.6,
      "learning_rate": 2.3069426018408448e-07,
      "loss": 0.1739,
      "step": 2259
    },
    {
      "epoch": 5.61,
      "learning_rate": 2.27812604669837e-07,
      "loss": 0.1412,
      "step": 2260
    },
    {
      "epoch": 5.61,
      "learning_rate": 2.2494885278932843e-07,
      "loss": 0.1603,
      "step": 2261
    },
    {
      "epoch": 5.61,
      "learning_rate": 2.221030097892407e-07,
      "loss": 0.1613,
      "step": 2262
    },
    {
      "epoch": 5.61,
      "learning_rate": 2.192750808834465e-07,
      "loss": 0.1294,
      "step": 2263
    },
    {
      "epoch": 5.62,
      "learning_rate": 2.1646507125299587e-07,
      "loss": 0.1566,
      "step": 2264
    },
    {
      "epoch": 5.62,
      "learning_rate": 2.1367298604611063e-07,
      "loss": 0.1361,
      "step": 2265
    },
    {
      "epoch": 5.62,
      "learning_rate": 2.108988303781745e-07,
      "loss": 0.1438,
      "step": 2266
    },
    {
      "epoch": 5.62,
      "learning_rate": 2.0814260933171737e-07,
      "loss": 0.1693,
      "step": 2267
    },
    {
      "epoch": 5.63,
      "learning_rate": 2.0540432795641664e-07,
      "loss": 0.1691,
      "step": 2268
    },
    {
      "epoch": 5.63,
      "learning_rate": 2.0268399126907922e-07,
      "loss": 0.1529,
      "step": 2269
    },
    {
      "epoch": 5.63,
      "learning_rate": 1.9998160425363734e-07,
      "loss": 0.1447,
      "step": 2270
    },
    {
      "epoch": 5.63,
      "learning_rate": 1.9729717186113605e-07,
      "loss": 0.1458,
      "step": 2271
    },
    {
      "epoch": 5.64,
      "learning_rate": 1.9463069900972686e-07,
      "loss": 0.1348,
      "step": 2272
    },
    {
      "epoch": 5.64,
      "learning_rate": 1.919821905846575e-07,
      "loss": 0.1596,
      "step": 2273
    },
    {
      "epoch": 5.64,
      "learning_rate": 1.8935165143826206e-07,
      "loss": 0.1473,
      "step": 2274
    },
    {
      "epoch": 5.64,
      "learning_rate": 1.867390863899543e-07,
      "loss": 0.1459,
      "step": 2275
    },
    {
      "epoch": 5.65,
      "learning_rate": 1.8414450022621656e-07,
      "loss": 0.1592,
      "step": 2276
    },
    {
      "epoch": 5.65,
      "learning_rate": 1.8156789770059413e-07,
      "loss": 0.1499,
      "step": 2277
    },
    {
      "epoch": 5.65,
      "learning_rate": 1.7900928353368098e-07,
      "loss": 0.1543,
      "step": 2278
    },
    {
      "epoch": 5.65,
      "learning_rate": 1.7646866241311733e-07,
      "loss": 0.1749,
      "step": 2279
    },
    {
      "epoch": 5.66,
      "learning_rate": 1.7394603899357766e-07,
      "loss": 0.1537,
      "step": 2280
    },
    {
      "epoch": 5.66,
      "learning_rate": 1.714414178967627e-07,
      "loss": 0.1856,
      "step": 2281
    },
    {
      "epoch": 5.66,
      "learning_rate": 1.6895480371139083e-07,
      "loss": 0.1534,
      "step": 2282
    },
    {
      "epoch": 5.66,
      "learning_rate": 1.6648620099319003e-07,
      "loss": 0.1445,
      "step": 2283
    },
    {
      "epoch": 5.67,
      "learning_rate": 1.6403561426489023e-07,
      "loss": 0.1553,
      "step": 2284
    },
    {
      "epoch": 5.67,
      "learning_rate": 1.616030480162134e-07,
      "loss": 0.1594,
      "step": 2285
    },
    {
      "epoch": 5.67,
      "learning_rate": 1.5918850670386677e-07,
      "loss": 0.1404,
      "step": 2286
    },
    {
      "epoch": 5.67,
      "learning_rate": 1.5679199475153395e-07,
      "loss": 0.163,
      "step": 2287
    },
    {
      "epoch": 5.68,
      "learning_rate": 1.5441351654986614e-07,
      "loss": 0.1666,
      "step": 2288
    },
    {
      "epoch": 5.68,
      "learning_rate": 1.5205307645647426e-07,
      "loss": 0.1499,
      "step": 2289
    },
    {
      "epoch": 5.68,
      "learning_rate": 1.4971067879592682e-07,
      "loss": 0.1548,
      "step": 2290
    },
    {
      "epoch": 5.68,
      "learning_rate": 1.4738632785972763e-07,
      "loss": 0.1723,
      "step": 2291
    },
    {
      "epoch": 5.69,
      "learning_rate": 1.450800279063258e-07,
      "loss": 0.1774,
      "step": 2292
    },
    {
      "epoch": 5.69,
      "learning_rate": 1.4279178316109588e-07,
      "loss": 0.1584,
      "step": 2293
    },
    {
      "epoch": 5.69,
      "learning_rate": 1.40521597816331e-07,
      "loss": 0.1534,
      "step": 2294
    },
    {
      "epoch": 5.69,
      "learning_rate": 1.3826947603124086e-07,
      "loss": 0.1851,
      "step": 2295
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.360354219319393e-07,
      "loss": 0.1512,
      "step": 2296
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.3381943961144118e-07,
      "loss": 0.1534,
      "step": 2297
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.3162153312964666e-07,
      "loss": 0.1705,
      "step": 2298
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.2944170651334577e-07,
      "loss": 0.1607,
      "step": 2299
    },
    {
      "epoch": 5.71,
      "learning_rate": 1.2727996375619722e-07,
      "loss": 0.151,
      "step": 2300
    },
    {
      "epoch": 5.71,
      "learning_rate": 1.2513630881873407e-07,
      "loss": 0.1477,
      "step": 2301
    },
    {
      "epoch": 5.71,
      "learning_rate": 1.2301074562834915e-07,
      "loss": 0.176,
      "step": 2302
    },
    {
      "epoch": 5.71,
      "learning_rate": 1.2090327807928626e-07,
      "loss": 0.1462,
      "step": 2303
    },
    {
      "epoch": 5.72,
      "learning_rate": 1.1881391003264131e-07,
      "loss": 0.1623,
      "step": 2304
    },
    {
      "epoch": 5.72,
      "learning_rate": 1.167426453163456e-07,
      "loss": 0.1413,
      "step": 2305
    },
    {
      "epoch": 5.72,
      "learning_rate": 1.1468948772516696e-07,
      "loss": 0.1546,
      "step": 2306
    },
    {
      "epoch": 5.72,
      "learning_rate": 1.1265444102069534e-07,
      "loss": 0.1527,
      "step": 2307
    },
    {
      "epoch": 5.73,
      "learning_rate": 1.1063750893134273e-07,
      "loss": 0.163,
      "step": 2308
    },
    {
      "epoch": 5.73,
      "learning_rate": 1.0863869515233327e-07,
      "loss": 0.1564,
      "step": 2309
    },
    {
      "epoch": 5.73,
      "learning_rate": 1.066580033456932e-07,
      "loss": 0.154,
      "step": 2310
    },
    {
      "epoch": 5.73,
      "learning_rate": 1.046954371402531e-07,
      "loss": 0.1598,
      "step": 2311
    },
    {
      "epoch": 5.74,
      "learning_rate": 1.0275100013162898e-07,
      "loss": 0.1708,
      "step": 2312
    },
    {
      "epoch": 5.74,
      "learning_rate": 1.0082469588222788e-07,
      "loss": 0.1687,
      "step": 2313
    },
    {
      "epoch": 5.74,
      "learning_rate": 9.89165279212323e-08,
      "loss": 0.1701,
      "step": 2314
    },
    {
      "epoch": 5.74,
      "learning_rate": 9.702649974460022e-08,
      "loss": 0.1306,
      "step": 2315
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.515461481505283e-08,
      "loss": 0.1416,
      "step": 2316
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.330087656207243e-08,
      "loss": 0.1669,
      "step": 2317
    },
    {
      "epoch": 5.75,
      "learning_rate": 9.14652883818945e-08,
      "loss": 0.1487,
      "step": 2318
    },
    {
      "epoch": 5.75,
      "learning_rate": 8.964785363750228e-08,
      "loss": 0.1546,
      "step": 2319
    },
    {
      "epoch": 5.76,
      "learning_rate": 8.784857565862004e-08,
      "loss": 0.1616,
      "step": 2320
    },
    {
      "epoch": 5.76,
      "learning_rate": 8.606745774170533e-08,
      "loss": 0.1694,
      "step": 2321
    },
    {
      "epoch": 5.76,
      "learning_rate": 8.430450314994676e-08,
      "loss": 0.167,
      "step": 2322
    },
    {
      "epoch": 5.76,
      "learning_rate": 8.255971511325622e-08,
      "loss": 0.1475,
      "step": 2323
    },
    {
      "epoch": 5.77,
      "learning_rate": 8.083309682826002e-08,
      "loss": 0.1649,
      "step": 2324
    },
    {
      "epoch": 5.77,
      "learning_rate": 7.912465145829662e-08,
      "loss": 0.1545,
      "step": 2325
    },
    {
      "epoch": 5.77,
      "learning_rate": 7.743438213341225e-08,
      "loss": 0.1499,
      "step": 2326
    },
    {
      "epoch": 5.77,
      "learning_rate": 7.576229195034868e-08,
      "loss": 0.1555,
      "step": 2327
    },
    {
      "epoch": 5.78,
      "learning_rate": 7.410838397254649e-08,
      "loss": 0.1528,
      "step": 2328
    },
    {
      "epoch": 5.78,
      "learning_rate": 7.247266123012963e-08,
      "loss": 0.1569,
      "step": 2329
    },
    {
      "epoch": 5.78,
      "learning_rate": 7.08551267199098e-08,
      "loss": 0.1507,
      "step": 2330
    },
    {
      "epoch": 5.78,
      "learning_rate": 6.925578340537087e-08,
      "loss": 0.1477,
      "step": 2331
    },
    {
      "epoch": 5.79,
      "learning_rate": 6.767463421667342e-08,
      "loss": 0.1431,
      "step": 2332
    },
    {
      "epoch": 5.79,
      "learning_rate": 6.611168205064134e-08,
      "loss": 0.1513,
      "step": 2333
    },
    {
      "epoch": 5.79,
      "learning_rate": 6.456692977075962e-08,
      "loss": 0.1362,
      "step": 2334
    },
    {
      "epoch": 5.79,
      "learning_rate": 6.304038020717329e-08,
      "loss": 0.1591,
      "step": 2335
    },
    {
      "epoch": 5.8,
      "learning_rate": 6.15320361566718e-08,
      "loss": 0.1485,
      "step": 2336
    },
    {
      "epoch": 5.8,
      "learning_rate": 6.004190038269685e-08,
      "loss": 0.1551,
      "step": 2337
    },
    {
      "epoch": 5.8,
      "learning_rate": 5.856997561532574e-08,
      "loss": 0.1612,
      "step": 2338
    },
    {
      "epoch": 5.8,
      "learning_rate": 5.7116264551273505e-08,
      "loss": 0.1668,
      "step": 2339
    },
    {
      "epoch": 5.81,
      "learning_rate": 5.5680769853886373e-08,
      "loss": 0.1547,
      "step": 2340
    },
    {
      "epoch": 5.81,
      "learning_rate": 5.426349415313503e-08,
      "loss": 0.1583,
      "step": 2341
    },
    {
      "epoch": 5.81,
      "learning_rate": 5.28644400456102e-08,
      "loss": 0.1574,
      "step": 2342
    },
    {
      "epoch": 5.81,
      "learning_rate": 5.148361009452263e-08,
      "loss": 0.1447,
      "step": 2343
    },
    {
      "epoch": 5.82,
      "learning_rate": 5.01210068296909e-08,
      "loss": 0.1502,
      "step": 2344
    },
    {
      "epoch": 5.82,
      "learning_rate": 4.8776632747541406e-08,
      "loss": 0.148,
      "step": 2345
    },
    {
      "epoch": 5.82,
      "learning_rate": 4.7450490311103934e-08,
      "loss": 0.1507,
      "step": 2346
    },
    {
      "epoch": 5.82,
      "learning_rate": 4.614258195000498e-08,
      "loss": 0.1623,
      "step": 2347
    },
    {
      "epoch": 5.83,
      "learning_rate": 4.485291006046555e-08,
      "loss": 0.143,
      "step": 2348
    },
    {
      "epoch": 5.83,
      "learning_rate": 4.358147700529225e-08,
      "loss": 0.1474,
      "step": 2349
    },
    {
      "epoch": 5.83,
      "learning_rate": 4.232828511388176e-08,
      "loss": 0.1495,
      "step": 2350
    },
    {
      "epoch": 5.83,
      "learning_rate": 4.109333668220639e-08,
      "loss": 0.1741,
      "step": 2351
    },
    {
      "epoch": 5.84,
      "learning_rate": 3.987663397281627e-08,
      "loss": 0.1432,
      "step": 2352
    },
    {
      "epoch": 5.84,
      "learning_rate": 3.867817921483496e-08,
      "loss": 0.147,
      "step": 2353
    },
    {
      "epoch": 5.84,
      "learning_rate": 3.749797460395166e-08,
      "loss": 0.1652,
      "step": 2354
    },
    {
      "epoch": 5.84,
      "learning_rate": 3.633602230242006e-08,
      "loss": 0.1578,
      "step": 2355
    },
    {
      "epoch": 5.85,
      "learning_rate": 3.519232443905396e-08,
      "loss": 0.1416,
      "step": 2356
    },
    {
      "epoch": 5.85,
      "learning_rate": 3.40668831092239e-08,
      "loss": 0.1477,
      "step": 2357
    },
    {
      "epoch": 5.85,
      "learning_rate": 3.2959700374851634e-08,
      "loss": 0.1638,
      "step": 2358
    },
    {
      "epoch": 5.85,
      "learning_rate": 3.1870778264407876e-08,
      "loss": 0.1481,
      "step": 2359
    },
    {
      "epoch": 5.86,
      "learning_rate": 3.080011877290789e-08,
      "loss": 0.1439,
      "step": 2360
    },
    {
      "epoch": 5.86,
      "learning_rate": 2.9747723861908163e-08,
      "loss": 0.1484,
      "step": 2361
    },
    {
      "epoch": 5.86,
      "learning_rate": 2.871359545950192e-08,
      "loss": 0.1719,
      "step": 2362
    },
    {
      "epoch": 5.86,
      "learning_rate": 2.7697735460316954e-08,
      "loss": 0.153,
      "step": 2363
    },
    {
      "epoch": 5.87,
      "learning_rate": 2.6700145725511163e-08,
      "loss": 0.1452,
      "step": 2364
    },
    {
      "epoch": 5.87,
      "learning_rate": 2.5720828082770323e-08,
      "loss": 0.1444,
      "step": 2365
    },
    {
      "epoch": 5.87,
      "learning_rate": 2.475978432630255e-08,
      "loss": 0.1771,
      "step": 2366
    },
    {
      "epoch": 5.87,
      "learning_rate": 2.3817016216837185e-08,
      "loss": 0.1406,
      "step": 2367
    },
    {
      "epoch": 5.88,
      "learning_rate": 2.2892525481620352e-08,
      "loss": 0.1482,
      "step": 2368
    },
    {
      "epoch": 5.88,
      "learning_rate": 2.198631381441163e-08,
      "loss": 0.1445,
      "step": 2369
    },
    {
      "epoch": 5.88,
      "learning_rate": 2.109838287548516e-08,
      "loss": 0.1461,
      "step": 2370
    },
    {
      "epoch": 5.88,
      "learning_rate": 2.0228734291616314e-08,
      "loss": 0.171,
      "step": 2371
    },
    {
      "epoch": 5.88,
      "learning_rate": 1.9377369656092825e-08,
      "loss": 0.1531,
      "step": 2372
    },
    {
      "epoch": 5.89,
      "learning_rate": 1.8544290528696995e-08,
      "loss": 0.1796,
      "step": 2373
    },
    {
      "epoch": 5.89,
      "learning_rate": 1.7729498435716808e-08,
      "loss": 0.1545,
      "step": 2374
    },
    {
      "epoch": 5.89,
      "learning_rate": 1.6932994869931495e-08,
      "loss": 0.1584,
      "step": 2375
    },
    {
      "epoch": 5.89,
      "learning_rate": 1.6154781290617095e-08,
      "loss": 0.1494,
      "step": 2376
    },
    {
      "epoch": 5.9,
      "learning_rate": 1.5394859123539773e-08,
      "loss": 0.1474,
      "step": 2377
    },
    {
      "epoch": 5.9,
      "learning_rate": 1.4653229760952515e-08,
      "loss": 0.1681,
      "step": 2378
    },
    {
      "epoch": 5.9,
      "learning_rate": 1.3929894561596214e-08,
      "loss": 0.1645,
      "step": 2379
    },
    {
      "epoch": 5.9,
      "learning_rate": 1.3224854850694135e-08,
      "loss": 0.1539,
      "step": 2380
    },
    {
      "epoch": 5.91,
      "learning_rate": 1.2538111919948581e-08,
      "loss": 0.1695,
      "step": 2381
    },
    {
      "epoch": 5.91,
      "learning_rate": 1.1869667027543109e-08,
      "loss": 0.1595,
      "step": 2382
    },
    {
      "epoch": 5.91,
      "learning_rate": 1.121952139813698e-08,
      "loss": 0.139,
      "step": 2383
    },
    {
      "epoch": 5.91,
      "learning_rate": 1.058767622286072e-08,
      "loss": 0.1352,
      "step": 2384
    },
    {
      "epoch": 5.92,
      "learning_rate": 9.974132659319457e-09,
      "loss": 0.1226,
      "step": 2385
    },
    {
      "epoch": 5.92,
      "learning_rate": 9.378891831587355e-09,
      "loss": 0.1603,
      "step": 2386
    },
    {
      "epoch": 5.92,
      "learning_rate": 8.801954830205405e-09,
      "loss": 0.1454,
      "step": 2387
    },
    {
      "epoch": 5.92,
      "learning_rate": 8.243322712180312e-09,
      "loss": 0.173,
      "step": 2388
    },
    {
      "epoch": 5.93,
      "learning_rate": 7.702996500982274e-09,
      "loss": 0.183,
      "step": 2389
    },
    {
      "epoch": 5.93,
      "learning_rate": 7.180977186542759e-09,
      "loss": 0.1525,
      "step": 2390
    },
    {
      "epoch": 5.93,
      "learning_rate": 6.677265725255622e-09,
      "loss": 0.151,
      "step": 2391
    },
    {
      "epoch": 5.93,
      "learning_rate": 6.191863039970436e-09,
      "loss": 0.1561,
      "step": 2392
    },
    {
      "epoch": 5.94,
      "learning_rate": 5.724770019993608e-09,
      "loss": 0.147,
      "step": 2393
    },
    {
      "epoch": 5.94,
      "learning_rate": 5.275987521087267e-09,
      "loss": 0.1494,
      "step": 2394
    },
    {
      "epoch": 5.94,
      "learning_rate": 4.84551636546593e-09,
      "loss": 0.1813,
      "step": 2395
    },
    {
      "epoch": 5.94,
      "learning_rate": 4.433357341795397e-09,
      "loss": 0.1409,
      "step": 2396
    },
    {
      "epoch": 5.95,
      "learning_rate": 4.039511205192748e-09,
      "loss": 0.1604,
      "step": 2397
    },
    {
      "epoch": 5.95,
      "learning_rate": 3.663978677225233e-09,
      "loss": 0.1476,
      "step": 2398
    },
    {
      "epoch": 5.95,
      "learning_rate": 3.3067604459058323e-09,
      "loss": 0.1696,
      "step": 2399
    },
    {
      "epoch": 5.95,
      "learning_rate": 2.9678571656932553e-09,
      "loss": 0.1575,
      "step": 2400
    },
    {
      "epoch": 5.96,
      "learning_rate": 2.6472694574941616e-09,
      "loss": 0.1524,
      "step": 2401
    },
    {
      "epoch": 5.96,
      "learning_rate": 2.344997908656499e-09,
      "loss": 0.154,
      "step": 2402
    },
    {
      "epoch": 5.96,
      "learning_rate": 2.061043072972835e-09,
      "loss": 0.1691,
      "step": 2403
    },
    {
      "epoch": 5.96,
      "learning_rate": 1.795405470675915e-09,
      "loss": 0.165,
      "step": 2404
    },
    {
      "epoch": 5.97,
      "learning_rate": 1.5480855884419943e-09,
      "loss": 0.1725,
      "step": 2405
    },
    {
      "epoch": 5.97,
      "learning_rate": 1.3190838793863958e-09,
      "loss": 0.1485,
      "step": 2406
    },
    {
      "epoch": 5.97,
      "learning_rate": 1.1084007630612903e-09,
      "loss": 0.1753,
      "step": 2407
    },
    {
      "epoch": 5.97,
      "learning_rate": 9.160366254612474e-10,
      "loss": 0.1251,
      "step": 2408
    },
    {
      "epoch": 5.98,
      "learning_rate": 7.419918190165743e-10,
      "loss": 0.1573,
      "step": 2409
    },
    {
      "epoch": 5.98,
      "learning_rate": 5.862666625944257e-10,
      "loss": 0.1622,
      "step": 2410
    },
    {
      "epoch": 5.98,
      "learning_rate": 4.4886144149880463e-10,
      "loss": 0.1485,
      "step": 2411
    },
    {
      "epoch": 5.98,
      "learning_rate": 3.2977640747056115e-10,
      "loss": 0.1639,
      "step": 2412
    },
    {
      "epoch": 5.99,
      "learning_rate": 2.290117786851731e-10,
      "loss": 0.1473,
      "step": 2413
    },
    {
      "epoch": 5.99,
      "learning_rate": 1.4656773975385564e-10,
      "loss": 0.1467,
      "step": 2414
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.244444172245125e-11,
      "loss": 0.1527,
      "step": 2415
    },
    {
      "epoch": 5.99,
      "learning_rate": 3.6642002070319536e-11,
      "loss": 0.1643,
      "step": 2416
    },
    {
      "epoch": 6.0,
      "learning_rate": 9.160504713667805e-12,
      "loss": 0.1577,
      "step": 2417
    },
    {
      "epoch": 6.0,
      "learning_rate": 0.0,
      "loss": 0.1582,
      "step": 2418
    },
    {
      "epoch": 6.0,
      "step": 2418,
      "total_flos": 3.273136024267522e+18,
      "train_loss": 0.8313779570439613,
      "train_runtime": 35252.5907,
      "train_samples_per_second": 8.78,
      "train_steps_per_second": 0.069
    }
  ],
  "max_steps": 2418,
  "num_train_epochs": 6,
  "total_flos": 3.273136024267522e+18,
  "trial_name": null,
  "trial_params": null
}