{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.9852125693160814,
  "eval_steps": 474,
  "global_step": 3786,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 2e-05,
      "loss": 1.6827,
      "step": 1
    },
    {
      "epoch": 0.0,
      "eval_loss": 1.6403473615646362,
      "eval_runtime": 682.6339,
      "eval_samples_per_second": 9.825,
      "eval_steps_per_second": 2.457,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 4e-05,
      "loss": 1.7415,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 6e-05,
      "loss": 1.6314,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 8e-05,
      "loss": 1.7016,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001,
      "loss": 1.6104,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00012,
      "loss": 1.5584,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00014,
      "loss": 1.5481,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00016,
      "loss": 1.4103,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00018,
      "loss": 1.4447,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0002,
      "loss": 1.3613,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999996538966897,
      "loss": 1.3593,
      "step": 11
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999986155869976,
      "loss": 1.2069,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999968850716425,
      "loss": 1.2925,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0001999994462351823,
      "loss": 1.241,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999913474292154,
      "loss": 1.2604,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0001999987540305976,
      "loss": 1.2496,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999830409847403,
      "loss": 1.2056,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999778494686225,
      "loss": 1.2323,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999719657612164,
      "loss": 1.1903,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999653898665948,
      "loss": 1.1803,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999581217893095,
      "loss": 1.1948,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999501615343914,
      "loss": 1.2198,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999415091073508,
      "loss": 1.1514,
      "step": 23
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0001999932164514177,
      "loss": 1.2074,
      "step": 24
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019999221277613384,
      "loss": 1.2841,
      "step": 25
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0001999911398855782,
      "loss": 1.2213,
      "step": 26
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0001999899977804935,
      "loss": 1.1186,
      "step": 27
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019998878646167027,
      "loss": 1.1874,
      "step": 28
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019998750592994707,
      "loss": 1.114,
      "step": 29
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019998615618621018,
      "loss": 1.1579,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000199984737231394,
      "loss": 1.1493,
      "step": 31
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019998324906648067,
      "loss": 1.0677,
      "step": 32
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019998169169250035,
      "loss": 1.0954,
      "step": 33
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019998006511053108,
      "loss": 1.159,
      "step": 34
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019997836932169875,
      "loss": 1.1669,
      "step": 35
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001999766043271772,
      "loss": 1.2065,
      "step": 36
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019997477012818822,
      "loss": 1.1624,
      "step": 37
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019997286672600138,
      "loss": 1.1287,
      "step": 38
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019997089412193426,
      "loss": 1.0498,
      "step": 39
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019996885231735232,
      "loss": 1.1984,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019996674131366891,
      "loss": 1.1827,
      "step": 41
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019996456111234527,
      "loss": 1.1191,
      "step": 42
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019996231171489056,
      "loss": 1.1401,
      "step": 43
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019995999312286182,
      "loss": 1.1514,
      "step": 44
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000199957605337864,
      "loss": 1.1748,
      "step": 45
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001999551483615499,
      "loss": 1.0959,
      "step": 46
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019995262219562034,
      "loss": 1.04,
      "step": 47
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019995002684182386,
      "loss": 1.0853,
      "step": 48
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999473623019571,
      "loss": 1.0513,
      "step": 49
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019994462857786431,
      "loss": 1.1221,
      "step": 50
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999418256714379,
      "loss": 1.1019,
      "step": 51
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019993895358461807,
      "loss": 1.0554,
      "step": 52
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019993601231939283,
      "loss": 1.0659,
      "step": 53
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999330018777982,
      "loss": 1.1068,
      "step": 54
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000199929922261918,
      "loss": 1.0432,
      "step": 55
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019992677347388394,
      "loss": 0.9979,
      "step": 56
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999235555158757,
      "loss": 1.0731,
      "step": 57
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019992026839012067,
      "loss": 1.0855,
      "step": 58
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019991691209889434,
      "loss": 1.0563,
      "step": 59
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019991348664451987,
      "loss": 1.0464,
      "step": 60
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019990999202936841,
      "loss": 1.0727,
      "step": 61
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019990642825585896,
      "loss": 1.0831,
      "step": 62
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001999027953264584,
      "loss": 1.0418,
      "step": 63
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001998990932436814,
      "loss": 1.1052,
      "step": 64
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019989532201009067,
      "loss": 0.9511,
      "step": 65
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001998914816282966,
      "loss": 0.9567,
      "step": 66
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019988757210095756,
      "loss": 1.023,
      "step": 67
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019988359343077976,
      "loss": 1.0098,
      "step": 68
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019987954562051725,
      "loss": 1.0096,
      "step": 69
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019987542867297195,
      "loss": 0.9689,
      "step": 70
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001998712425909936,
      "loss": 0.975,
      "step": 71
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019986698737747992,
      "loss": 0.9841,
      "step": 72
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019986266303537634,
      "loss": 0.9613,
      "step": 73
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001998582695676762,
      "loss": 1.0895,
      "step": 74
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019985380697742068,
      "loss": 0.9522,
      "step": 75
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019984927526769884,
      "loss": 1.0116,
      "step": 76
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001998446744416476,
      "loss": 0.9006,
      "step": 77
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019984000450245156,
      "loss": 0.9101,
      "step": 78
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001998352654533434,
      "loss": 1.0029,
      "step": 79
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019983045729760345,
      "loss": 0.8863,
      "step": 80
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019982558003856,
      "loss": 0.9774,
      "step": 81
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019982063367958905,
      "loss": 0.863,
      "step": 82
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019981561822411458,
      "loss": 1.0388,
      "step": 83
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001998105336756083,
      "loss": 0.9378,
      "step": 84
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019980538003758972,
      "loss": 1.052,
      "step": 85
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019980015731362627,
      "loss": 0.9523,
      "step": 86
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019979486550733314,
      "loss": 0.929,
      "step": 87
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019978950462237337,
      "loss": 0.9387,
      "step": 88
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019978407466245777,
      "loss": 0.9565,
      "step": 89
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000199778575631345,
      "loss": 1.0259,
      "step": 90
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019977300753284158,
      "loss": 0.9605,
      "step": 91
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001997673703708017,
      "loss": 0.8683,
      "step": 92
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001997616641491275,
      "loss": 0.8937,
      "step": 93
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019975588887176883,
      "loss": 0.8495,
      "step": 94
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019975004454272338,
      "loss": 0.8717,
      "step": 95
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001997441311660367,
      "loss": 0.9684,
      "step": 96
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019973814874580198,
      "loss": 0.879,
      "step": 97
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019973209728616032,
      "loss": 0.927,
      "step": 98
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019972597679130058,
      "loss": 1.0419,
      "step": 99
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019971978726545942,
      "loss": 0.9943,
      "step": 100
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019971352871292125,
      "loss": 0.8513,
      "step": 101
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019970720113801827,
      "loss": 0.9358,
      "step": 102
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019970080454513055,
      "loss": 0.8453,
      "step": 103
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019969433893868578,
      "loss": 0.8896,
      "step": 104
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001996878043231595,
      "loss": 0.8942,
      "step": 105
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000199681200703075,
      "loss": 0.9442,
      "step": 106
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019967452808300343,
      "loss": 0.8504,
      "step": 107
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019966778646756354,
      "loss": 0.8818,
      "step": 108
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019966097586142194,
      "loss": 0.92,
      "step": 109
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000199654096269293,
      "loss": 0.9062,
      "step": 110
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019964714769593882,
      "loss": 0.9225,
      "step": 111
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001996401301461692,
      "loss": 0.8717,
      "step": 112
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019963304362484178,
      "loss": 0.8728,
      "step": 113
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019962588813686187,
      "loss": 0.8795,
      "step": 114
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001996186636871826,
      "loss": 0.8862,
      "step": 115
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001996113702808047,
      "loss": 0.9683,
      "step": 116
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019960400792277678,
      "loss": 0.876,
      "step": 117
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019959657661819508,
      "loss": 0.8434,
      "step": 118
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001995890763722036,
      "loss": 0.8244,
      "step": 119
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019958150718999408,
      "loss": 0.8203,
      "step": 120
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019957386907680594,
      "loss": 0.8332,
      "step": 121
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019956616203792635,
      "loss": 0.8765,
      "step": 122
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019955838607869014,
      "loss": 0.8099,
      "step": 123
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019955054120447994,
      "loss": 0.7839,
      "step": 124
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019954262742072592,
      "loss": 0.8986,
      "step": 125
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001995346447329062,
      "loss": 0.8446,
      "step": 126
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001995265931465463,
      "loss": 0.8019,
      "step": 127
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001995184726672197,
      "loss": 0.7754,
      "step": 128
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019951028330054737,
      "loss": 0.8713,
      "step": 129
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019950202505219808,
      "loss": 0.8419,
      "step": 130
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019949369792788825,
      "loss": 0.8456,
      "step": 131
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019948530193338195,
      "loss": 0.7685,
      "step": 132
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019947683707449094,
      "loss": 0.8374,
      "step": 133
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019946830335707468,
      "loss": 0.8055,
      "step": 134
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019945970078704024,
      "loss": 0.809,
      "step": 135
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019945102937034238,
      "loss": 0.8346,
      "step": 136
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019944228911298353,
      "loss": 0.8243,
      "step": 137
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019943348002101371,
      "loss": 0.6564,
      "step": 138
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001994246021005307,
      "loss": 0.8963,
      "step": 139
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019941565535767981,
      "loss": 0.7951,
      "step": 140
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000199406639798654,
      "loss": 0.6793,
      "step": 141
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000199397555429694,
      "loss": 0.8094,
      "step": 142
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000199388402257088,
      "loss": 0.9244,
      "step": 143
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019937918028717187,
      "loss": 0.836,
      "step": 144
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019936988952632916,
      "loss": 0.7727,
      "step": 145
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000199360529980991,
      "loss": 0.7911,
      "step": 146
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001993511016576361,
      "loss": 0.8511,
      "step": 147
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001993416045627908,
      "loss": 0.8016,
      "step": 148
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019933203870302907,
      "loss": 0.8169,
      "step": 149
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001993224040849725,
      "loss": 0.7972,
      "step": 150
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019931270071529017,
      "loss": 0.7496,
      "step": 151
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019930292860069888,
      "loss": 0.8504,
      "step": 152
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001992930877479629,
      "loss": 0.7957,
      "step": 153
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019928317816389417,
      "loss": 0.8374,
      "step": 154
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019927319985535213,
      "loss": 0.7379,
      "step": 155
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019926315282924387,
      "loss": 0.7342,
      "step": 156
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000199253037092524,
      "loss": 0.6978,
      "step": 157
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019924285265219467,
      "loss": 0.7789,
      "step": 158
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019923259951530565,
      "loss": 0.7688,
      "step": 159
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001992222776889542,
      "loss": 0.8388,
      "step": 160
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019921188718028518,
      "loss": 0.7624,
      "step": 161
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019920142799649098,
      "loss": 0.8582,
      "step": 162
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001991909001448115,
      "loss": 0.9053,
      "step": 163
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019918030363253418,
      "loss": 0.6999,
      "step": 164
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019916963846699403,
      "loss": 0.7555,
      "step": 165
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019915890465557347,
      "loss": 0.783,
      "step": 166
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001991481022057026,
      "loss": 0.7885,
      "step": 167
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001991372311248589,
      "loss": 0.6863,
      "step": 168
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019912629142056744,
      "loss": 0.8085,
      "step": 169
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019911528310040074,
      "loss": 0.769,
      "step": 170
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001991042061719788,
      "loss": 0.7847,
      "step": 171
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001990930606429692,
      "loss": 0.7799,
      "step": 172
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001990818465210869,
      "loss": 0.7755,
      "step": 173
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001990705638140944,
      "loss": 0.918,
      "step": 174
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001990592125298017,
      "loss": 0.7651,
      "step": 175
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001990477926760662,
      "loss": 0.6945,
      "step": 176
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019903630426079279,
      "loss": 0.7033,
      "step": 177
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00019902474729193386,
      "loss": 0.7857,
      "step": 178
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001990131217774892,
      "loss": 0.7461,
      "step": 179
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019900142772550603,
      "loss": 0.7372,
      "step": 180
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019898966514407914,
      "loss": 0.8455,
      "step": 181
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001989778340413506,
      "loss": 0.7223,
      "step": 182
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019896593442551,
      "loss": 0.6334,
      "step": 183
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019895396630479435,
      "loss": 0.6756,
      "step": 184
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019894192968748802,
      "loss": 0.7634,
      "step": 185
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019892982458192288,
      "loss": 0.6425,
      "step": 186
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001989176509964781,
      "loss": 0.7796,
      "step": 187
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001989054089395804,
      "loss": 0.7484,
      "step": 188
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019889309841970374,
      "loss": 0.7369,
      "step": 189
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001988807194453696,
      "loss": 0.8663,
      "step": 190
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019886827202514673,
      "loss": 0.6501,
      "step": 191
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019885575616765137,
      "loss": 0.6236,
      "step": 192
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019884317188154706,
      "loss": 0.7812,
      "step": 193
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019883051917554472,
      "loss": 0.6338,
      "step": 194
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019881779805840265,
      "loss": 0.6472,
      "step": 195
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019880500853892646,
      "loss": 0.7427,
      "step": 196
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001987921506259692,
      "loss": 0.7296,
      "step": 197
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019877922432843116,
      "loss": 0.6716,
      "step": 198
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019876622965525997,
      "loss": 0.658,
      "step": 199
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001987531666154507,
      "loss": 0.676,
      "step": 200
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019874003521804564,
      "loss": 0.6148,
      "step": 201
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019872683547213446,
      "loss": 0.6585,
      "step": 202
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001987135673868541,
      "loss": 0.6368,
      "step": 203
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019870023097138875,
      "loss": 0.716,
      "step": 204
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019868682623497006,
      "loss": 0.7773,
      "step": 205
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001986733531868768,
      "loss": 0.737,
      "step": 206
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019865981183643516,
      "loss": 0.7895,
      "step": 207
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019864620219301856,
      "loss": 0.7056,
      "step": 208
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001986325242660476,
      "loss": 0.798,
      "step": 209
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019861877806499035,
      "loss": 0.675,
      "step": 210
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001986049635993619,
      "loss": 0.7328,
      "step": 211
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019859108087872483,
      "loss": 0.5785,
      "step": 212
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019857712991268875,
      "loss": 0.6947,
      "step": 213
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001985631107109107,
      "loss": 0.6767,
      "step": 214
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019854902328309478,
      "loss": 0.6632,
      "step": 215
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019853486763899247,
      "loss": 0.6893,
      "step": 216
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019852064378840235,
      "loss": 0.7121,
      "step": 217
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019850635174117033,
      "loss": 0.6771,
      "step": 218
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019849199150718936,
      "loss": 0.6797,
      "step": 219
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001984775630963998,
      "loss": 0.7419,
      "step": 220
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000198463066518789,
      "loss": 0.8314,
      "step": 221
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019844850178439162,
      "loss": 0.6751,
      "step": 222
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001984338689032895,
      "loss": 0.6672,
      "step": 223
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019841916788561155,
      "loss": 0.583,
      "step": 224
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019840439874153397,
      "loss": 0.6423,
      "step": 225
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019838956148128002,
      "loss": 0.6227,
      "step": 226
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019837465611512018,
      "loss": 0.6453,
      "step": 227
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019835968265337204,
      "loss": 0.7909,
      "step": 228
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019834464110640032,
      "loss": 0.6484,
      "step": 229
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019832953148461683,
      "loss": 0.6466,
      "step": 230
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019831435379848066,
      "loss": 0.6341,
      "step": 231
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001982991080584978,
      "loss": 0.7028,
      "step": 232
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019828379427522155,
      "loss": 0.6312,
      "step": 233
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019826841245925212,
      "loss": 0.6302,
      "step": 234
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00019825296262123692,
      "loss": 0.5905,
      "step": 235
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001982374447718705,
      "loss": 0.5553,
      "step": 236
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019822185892189433,
      "loss": 0.7098,
      "step": 237
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019820620508209712,
      "loss": 0.5711,
      "step": 238
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019819048326331448,
      "loss": 0.5917,
      "step": 239
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001981746934764292,
      "loss": 0.7259,
      "step": 240
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001981588357323711,
      "loss": 0.6311,
      "step": 241
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019814291004211697,
      "loss": 0.5428,
      "step": 242
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001981269164166907,
      "loss": 0.6518,
      "step": 243
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019811085486716318,
      "loss": 0.5686,
      "step": 244
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019809472540465234,
      "loss": 0.652,
      "step": 245
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019807852804032305,
      "loss": 0.6471,
      "step": 246
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001980622627853873,
      "loss": 0.5876,
      "step": 247
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019804592965110395,
      "loss": 0.4256,
      "step": 248
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001980295286487789,
      "loss": 0.4416,
      "step": 249
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001980130597897651,
      "loss": 0.6053,
      "step": 250
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019799652308546235,
      "loss": 0.6321,
      "step": 251
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019797991854731745,
      "loss": 0.6391,
      "step": 252
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019796324618682424,
      "loss": 0.5523,
      "step": 253
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019794650601552335,
      "loss": 0.551,
      "step": 254
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019792969804500253,
      "loss": 0.6906,
      "step": 255
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001979128222868963,
      "loss": 0.622,
      "step": 256
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019789587875288617,
      "loss": 0.5587,
      "step": 257
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019787886745470065,
      "loss": 0.5844,
      "step": 258
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019786178840411496,
      "loss": 0.511,
      "step": 259
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019784464161295143,
      "loss": 0.4971,
      "step": 260
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001978274270930791,
      "loss": 0.4905,
      "step": 261
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019781014485641403,
      "loss": 0.678,
      "step": 262
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019779279491491907,
      "loss": 0.5679,
      "step": 263
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000197775377280604,
      "loss": 0.5433,
      "step": 264
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019775789196552538,
      "loss": 0.6029,
      "step": 265
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019774033898178667,
      "loss": 0.4002,
      "step": 266
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019772271834153816,
      "loss": 0.571,
      "step": 267
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000197705030056977,
      "loss": 0.5115,
      "step": 268
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019768727414034716,
      "loss": 0.512,
      "step": 269
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019766945060393934,
      "loss": 0.4941,
      "step": 270
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001976515594600911,
      "loss": 0.6701,
      "step": 271
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001976336007211869,
      "loss": 0.6203,
      "step": 272
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019761557439965777,
      "loss": 0.5377,
      "step": 273
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00019759748050798178,
      "loss": 0.398,
      "step": 274
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019757931905868355,
      "loss": 0.5649,
      "step": 275
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019756109006433459,
      "loss": 0.566,
      "step": 276
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019754279353755312,
      "loss": 0.4631,
      "step": 277
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019752442949100415,
      "loss": 0.5042,
      "step": 278
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019750599793739936,
      "loss": 0.6321,
      "step": 279
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001974874988894972,
      "loss": 0.6155,
      "step": 280
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019746893236010282,
      "loss": 0.6981,
      "step": 281
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019745029836206813,
      "loss": 0.5124,
      "step": 282
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001974315969082917,
      "loss": 0.5185,
      "step": 283
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019741282801171873,
      "loss": 0.6237,
      "step": 284
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001973939916853413,
      "loss": 0.6814,
      "step": 285
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019737508794219792,
      "loss": 0.7321,
      "step": 286
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019735611679537395,
      "loss": 0.4718,
      "step": 287
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019733707825800135,
      "loss": 0.5707,
      "step": 288
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001973179723432587,
      "loss": 0.6378,
      "step": 289
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019729879906437122,
      "loss": 0.6632,
      "step": 290
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00019727955843461086,
      "loss": 0.5456,
      "step": 291
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000197260250467296,
      "loss": 0.6076,
      "step": 292
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001972408751757918,
      "loss": 0.5761,
      "step": 293
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019722143257351,
      "loss": 0.3927,
      "step": 294
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019720192267390883,
      "loss": 0.6148,
      "step": 295
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001971823454904932,
      "loss": 0.5988,
      "step": 296
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001971627010368146,
      "loss": 0.4857,
      "step": 297
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019714298932647098,
      "loss": 0.6357,
      "step": 298
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000197123210373107,
      "loss": 0.5416,
      "step": 299
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019710336419041373,
      "loss": 0.4784,
      "step": 300
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019708345079212883,
      "loss": 0.5227,
      "step": 301
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001970634701920365,
      "loss": 0.6115,
      "step": 302
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001970434224039674,
      "loss": 0.4916,
      "step": 303
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019702330744179883,
      "loss": 0.4358,
      "step": 304
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019700312531945442,
      "loss": 0.7045,
      "step": 305
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001969828760509044,
      "loss": 0.4024,
      "step": 306
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019696255965016545,
      "loss": 0.6041,
      "step": 307
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001969421761313007,
      "loss": 0.6045,
      "step": 308
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019692172550841977,
      "loss": 0.5376,
      "step": 309
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019690120779567875,
      "loss": 0.5746,
      "step": 310
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00019688062300728007,
      "loss": 0.4804,
      "step": 311
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001968599711574727,
      "loss": 0.7026,
      "step": 312
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000196839252260552,
      "loss": 0.5828,
      "step": 313
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019681846633085967,
      "loss": 0.6048,
      "step": 314
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019679761338278393,
      "loss": 0.4623,
      "step": 315
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019677669343075928,
      "loss": 0.6218,
      "step": 316
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019675570648926667,
      "loss": 0.5433,
      "step": 317
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019673465257283342,
      "loss": 0.6229,
      "step": 318
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019671353169603315,
      "loss": 0.3385,
      "step": 319
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001966923438734859,
      "loss": 0.4531,
      "step": 320
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019667108911985802,
      "loss": 0.4682,
      "step": 321
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001966497674498622,
      "loss": 0.4215,
      "step": 322
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019662837887825737,
      "loss": 0.4271,
      "step": 323
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001966069234198489,
      "loss": 0.5211,
      "step": 324
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019658540108948843,
      "loss": 0.6122,
      "step": 325
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019656381190207386,
      "loss": 0.5395,
      "step": 326
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019654215587254926,
      "loss": 0.4182,
      "step": 327
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019652043301590522,
      "loss": 0.4397,
      "step": 328
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019649864334717835,
      "loss": 0.5954,
      "step": 329
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001964767868814516,
      "loss": 0.5507,
      "step": 330
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00019645486363385422,
      "loss": 0.5757,
      "step": 331
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019643287361956159,
      "loss": 0.4473,
      "step": 332
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019641081685379535,
      "loss": 0.4888,
      "step": 333
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019638869335182336,
      "loss": 0.5083,
      "step": 334
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001963665031289596,
      "loss": 0.5729,
      "step": 335
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019634424620056432,
      "loss": 0.4122,
      "step": 336
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019632192258204394,
      "loss": 0.4328,
      "step": 337
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019629953228885098,
      "loss": 0.3844,
      "step": 338
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019627707533648417,
      "loss": 0.5123,
      "step": 339
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019625455174048835,
      "loss": 0.3861,
      "step": 340
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001962319615164545,
      "loss": 0.5526,
      "step": 341
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019620930468001974,
      "loss": 0.5327,
      "step": 342
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019618658124686728,
      "loss": 0.4724,
      "step": 343
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001961637912327264,
      "loss": 0.4352,
      "step": 344
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019614093465337255,
      "loss": 0.5487,
      "step": 345
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019611801152462715,
      "loss": 0.517,
      "step": 346
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019609502186235776,
      "loss": 0.6051,
      "step": 347
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019607196568247802,
      "loss": 0.4616,
      "step": 348
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001960488430009475,
      "loss": 0.3068,
      "step": 349
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001960256538337719,
      "loss": 0.615,
      "step": 350
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019600239819700296,
      "loss": 0.4958,
      "step": 351
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019597907610673833,
      "loss": 0.6533,
      "step": 352
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019595568757912172,
      "loss": 0.6606,
      "step": 353
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019593223263034287,
      "loss": 0.4331,
      "step": 354
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001959087112766374,
      "loss": 0.4732,
      "step": 355
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019588512353428694,
      "loss": 0.5235,
      "step": 356
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019586146941961912,
      "loss": 0.4875,
      "step": 357
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019583774894900747,
      "loss": 0.4619,
      "step": 358
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019581396213887145,
      "loss": 0.4515,
      "step": 359
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019579010900567644,
      "loss": 0.4014,
      "step": 360
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019576618956593372,
      "loss": 0.6054,
      "step": 361
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019574220383620055,
      "loss": 0.463,
      "step": 362
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019571815183307996,
      "loss": 0.5436,
      "step": 363
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001956940335732209,
      "loss": 0.4796,
      "step": 364
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001956698490733182,
      "loss": 0.5448,
      "step": 365
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019564559835011253,
      "loss": 0.4491,
      "step": 366
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019562128142039043,
      "loss": 0.4948,
      "step": 367
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001955968983009842,
      "loss": 0.5413,
      "step": 368
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019557244900877198,
      "loss": 0.5024,
      "step": 369
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001955479335606778,
      "loss": 0.5393,
      "step": 370
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019552335197367136,
      "loss": 0.4885,
      "step": 371
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019549870426476825,
      "loss": 0.4767,
      "step": 372
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001954739904510297,
      "loss": 0.5413,
      "step": 373
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019544921054956286,
      "loss": 0.4429,
      "step": 374
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019542436457752044,
      "loss": 0.5065,
      "step": 375
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001953994525521011,
      "loss": 0.4118,
      "step": 376
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019537447449054901,
      "loss": 0.4231,
      "step": 377
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019534943041015423,
      "loss": 0.5443,
      "step": 378
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001953243203282524,
      "loss": 0.4195,
      "step": 379
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001952991442622249,
      "loss": 0.4849,
      "step": 380
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019527390222949875,
      "loss": 0.3906,
      "step": 381
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019524859424754665,
      "loss": 0.5446,
      "step": 382
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019522322033388697,
      "loss": 0.536,
      "step": 383
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019519778050608368,
      "loss": 0.5869,
      "step": 384
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019517227478174642,
      "loss": 0.339,
      "step": 385
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019514670317853042,
      "loss": 0.3948,
      "step": 386
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001951210657141365,
      "loss": 0.3832,
      "step": 387
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00019509536240631107,
      "loss": 0.4871,
      "step": 388
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019506959327284619,
      "loss": 0.5002,
      "step": 389
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019504375833157936,
      "loss": 0.4643,
      "step": 390
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019501785760039371,
      "loss": 0.4528,
      "step": 391
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019499189109721795,
      "loss": 0.5932,
      "step": 392
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019496585884002616,
      "loss": 0.3793,
      "step": 393
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019493976084683813,
      "loss": 0.3792,
      "step": 394
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019491359713571903,
      "loss": 0.4917,
      "step": 395
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019488736772477961,
      "loss": 0.3786,
      "step": 396
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019486107263217594,
      "loss": 0.3366,
      "step": 397
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019483471187610972,
      "loss": 0.4563,
      "step": 398
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019480828547482806,
      "loss": 0.4573,
      "step": 399
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019478179344662344,
      "loss": 0.3949,
      "step": 400
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019475523580983383,
      "loss": 0.5698,
      "step": 401
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001947286125828426,
      "loss": 0.3103,
      "step": 402
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001947019237840786,
      "loss": 0.4051,
      "step": 403
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019467516943201586,
      "loss": 0.4425,
      "step": 404
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019464834954517398,
      "loss": 0.4839,
      "step": 405
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019462146414211793,
      "loss": 0.491,
      "step": 406
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019459451324145785,
      "loss": 0.5769,
      "step": 407
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019456749686184944,
      "loss": 0.4706,
      "step": 408
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019454041502199351,
      "loss": 0.3509,
      "step": 409
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019451326774063636,
      "loss": 0.3738,
      "step": 410
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019448605503656952,
      "loss": 0.4354,
      "step": 411
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019445877692862975,
      "loss": 0.4723,
      "step": 412
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019443143343569918,
      "loss": 0.313,
      "step": 413
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019440402457670514,
      "loss": 0.4611,
      "step": 414
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019437655037062026,
      "loss": 0.2965,
      "step": 415
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019434901083646233,
      "loss": 0.4249,
      "step": 416
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019432140599329442,
      "loss": 0.3893,
      "step": 417
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019429373586022473,
      "loss": 0.5318,
      "step": 418
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019426600045640676,
      "loss": 0.4963,
      "step": 419
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019423819980103914,
      "loss": 0.4694,
      "step": 420
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019421033391336568,
      "loss": 0.414,
      "step": 421
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001941824028126753,
      "loss": 0.5468,
      "step": 422
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019415440651830208,
      "loss": 0.4955,
      "step": 423
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001941263450496253,
      "loss": 0.4204,
      "step": 424
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019409821842606922,
      "loss": 0.3463,
      "step": 425
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019407002666710336,
      "loss": 0.4911,
      "step": 426
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019404176979224212,
      "loss": 0.393,
      "step": 427
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019401344782104523,
      "loss": 0.5595,
      "step": 428
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019398506077311729,
      "loss": 0.3889,
      "step": 429
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019395660866810798,
      "loss": 0.3107,
      "step": 430
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019392809152571206,
      "loss": 0.3679,
      "step": 431
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001938995093656693,
      "loss": 0.4218,
      "step": 432
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019387086220776443,
      "loss": 0.4102,
      "step": 433
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019384215007182718,
      "loss": 0.4553,
      "step": 434
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019381337297773234,
      "loss": 0.3795,
      "step": 435
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019378453094539956,
      "loss": 0.2866,
      "step": 436
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019375562399479354,
      "loss": 0.4198,
      "step": 437
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019372665214592382,
      "loss": 0.3574,
      "step": 438
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019369761541884487,
      "loss": 0.3776,
      "step": 439
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001936685138336562,
      "loss": 0.3728,
      "step": 440
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019363934741050204,
      "loss": 0.5994,
      "step": 441
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019361011616957164,
      "loss": 0.4657,
      "step": 442
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000193580820131099,
      "loss": 0.5495,
      "step": 443
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00019355145931536304,
      "loss": 0.4567,
      "step": 444
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019352203374268755,
      "loss": 0.4354,
      "step": 445
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001934925434334411,
      "loss": 0.4938,
      "step": 446
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019346298840803704,
      "loss": 0.3425,
      "step": 447
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001934333686869336,
      "loss": 0.3021,
      "step": 448
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001934036842906337,
      "loss": 0.2868,
      "step": 449
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001933739352396851,
      "loss": 0.5395,
      "step": 450
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001933441215546803,
      "loss": 0.3963,
      "step": 451
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019331424325625657,
      "loss": 0.4274,
      "step": 452
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019328430036509575,
      "loss": 0.4338,
      "step": 453
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019325429290192457,
      "loss": 0.4878,
      "step": 454
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019322422088751445,
      "loss": 0.3266,
      "step": 455
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019319408434268132,
      "loss": 0.2842,
      "step": 456
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019316388328828604,
      "loss": 0.2113,
      "step": 457
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019313361774523385,
      "loss": 0.4431,
      "step": 458
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001931032877344748,
      "loss": 0.4374,
      "step": 459
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019307289327700358,
      "loss": 0.3236,
      "step": 460
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019304243439385938,
      "loss": 0.5163,
      "step": 461
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019301191110612605,
      "loss": 0.4897,
      "step": 462
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000192981323434932,
      "loss": 0.3545,
      "step": 463
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019295067140145022,
      "loss": 0.3728,
      "step": 464
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001929199550268983,
      "loss": 0.2153,
      "step": 465
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019288917433253825,
      "loss": 0.258,
      "step": 466
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019285832933967672,
      "loss": 0.3084,
      "step": 467
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019282742006966478,
      "loss": 0.4221,
      "step": 468
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019279644654389807,
      "loss": 0.4629,
      "step": 469
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019276540878381663,
      "loss": 0.4955,
      "step": 470
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000192734306810905,
      "loss": 0.2758,
      "step": 471
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019270314064669222,
      "loss": 0.2728,
      "step": 472
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019267191031275172,
      "loss": 0.3211,
      "step": 473
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019264061583070127,
      "loss": 0.4101,
      "step": 474
    },
    {
      "epoch": 0.25,
      "eval_loss": 0.37793055176734924,
      "eval_runtime": 683.4711,
      "eval_samples_per_second": 9.813,
      "eval_steps_per_second": 2.454,
      "step": 474
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019260925722220317,
      "loss": 0.3679,
      "step": 475
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019257783450896406,
      "loss": 0.4074,
      "step": 476
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019254634771273492,
      "loss": 0.4984,
      "step": 477
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019251479685531114,
      "loss": 0.3992,
      "step": 478
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019248318195853244,
      "loss": 0.4017,
      "step": 479
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00019245150304428283,
      "loss": 0.3794,
      "step": 480
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001924197601344907,
      "loss": 0.369,
      "step": 481
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001923879532511287,
      "loss": 0.4137,
      "step": 482
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001923560824162137,
      "loss": 0.4004,
      "step": 483
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000192324147651807,
      "loss": 0.2607,
      "step": 484
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000192292148980014,
      "loss": 0.4933,
      "step": 485
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019226008642298442,
      "loss": 0.3053,
      "step": 486
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019222796000291213,
      "loss": 0.4018,
      "step": 487
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019219576974203526,
      "loss": 0.4102,
      "step": 488
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019216351566263618,
      "loss": 0.4241,
      "step": 489
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019213119778704128,
      "loss": 0.4772,
      "step": 490
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019209881613762133,
      "loss": 0.3709,
      "step": 491
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019206637073679103,
      "loss": 0.2739,
      "step": 492
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001920338616070093,
      "loss": 0.4537,
      "step": 493
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019200128877077925,
      "loss": 0.3589,
      "step": 494
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019196865225064793,
      "loss": 0.3767,
      "step": 495
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001919359520692066,
      "loss": 0.3263,
      "step": 496
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019190318824909052,
      "loss": 0.2835,
      "step": 497
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019187036081297907,
      "loss": 0.4297,
      "step": 498
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019183746978359553,
      "loss": 0.4173,
      "step": 499
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001918045151837074,
      "loss": 0.4225,
      "step": 500
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000191771497036126,
      "loss": 0.5362,
      "step": 501
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001917384153637067,
      "loss": 0.341,
      "step": 502
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019170527018934886,
      "loss": 0.3225,
      "step": 503
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019167206153599585,
      "loss": 0.4146,
      "step": 504
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001916387894266349,
      "loss": 0.3541,
      "step": 505
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019160545388429708,
      "loss": 0.432,
      "step": 506
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019157205493205762,
      "loss": 0.3055,
      "step": 507
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001915385925930354,
      "loss": 0.3346,
      "step": 508
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001915050668903933,
      "loss": 0.5637,
      "step": 509
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019147147784733806,
      "loss": 0.2844,
      "step": 510
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001914378254871202,
      "loss": 0.5342,
      "step": 511
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001914041098330341,
      "loss": 0.3227,
      "step": 512
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000191370330908418,
      "loss": 0.5065,
      "step": 513
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019133648873665392,
      "loss": 0.5285,
      "step": 514
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019130258334116752,
      "loss": 0.308,
      "step": 515
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019126861474542844,
      "loss": 0.32,
      "step": 516
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019123458297294992,
      "loss": 0.3941,
      "step": 517
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000191200488047289,
      "loss": 0.3382,
      "step": 518
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001911663299920464,
      "loss": 0.3564,
      "step": 519
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019113210883086657,
      "loss": 0.3528,
      "step": 520
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019109782458743763,
      "loss": 0.3915,
      "step": 521
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019106347728549135,
      "loss": 0.3346,
      "step": 522
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001910290669488031,
      "loss": 0.2841,
      "step": 523
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019099459360119204,
      "loss": 0.465,
      "step": 524
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001909600572665208,
      "loss": 0.3728,
      "step": 525
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019092545796869565,
      "loss": 0.4835,
      "step": 526
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001908907957316665,
      "loss": 0.4034,
      "step": 527
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001908560705794267,
      "loss": 0.3452,
      "step": 528
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019082128253601334,
      "loss": 0.2769,
      "step": 529
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019078643162550685,
      "loss": 0.3648,
      "step": 530
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019075151787203125,
      "loss": 0.3377,
      "step": 531
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019071654129975416,
      "loss": 0.4024,
      "step": 532
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001906815019328865,
      "loss": 0.3533,
      "step": 533
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001906463997956828,
      "loss": 0.453,
      "step": 534
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000190611234912441,
      "loss": 0.3315,
      "step": 535
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019057600730750242,
      "loss": 0.3055,
      "step": 536
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001905407170052519,
      "loss": 0.293,
      "step": 537
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001905053640301176,
      "loss": 0.3877,
      "step": 538
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00019046994840657105,
      "loss": 0.4629,
      "step": 539
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019043447015912722,
      "loss": 0.4711,
      "step": 540
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019039892931234435,
      "loss": 0.3827,
      "step": 541
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019036332589082407,
      "loss": 0.2052,
      "step": 542
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019032765991921133,
      "loss": 0.3286,
      "step": 543
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001902919314221943,
      "loss": 0.3437,
      "step": 544
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019025614042450455,
      "loss": 0.4702,
      "step": 545
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019022028695091675,
      "loss": 0.3217,
      "step": 546
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000190184371026249,
      "loss": 0.3117,
      "step": 547
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001901483926753625,
      "loss": 0.4213,
      "step": 548
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001901123519231617,
      "loss": 0.2494,
      "step": 549
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001900762487945943,
      "loss": 0.2425,
      "step": 550
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019004008331465104,
      "loss": 0.1523,
      "step": 551
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00019000385550836595,
      "loss": 0.2795,
      "step": 552
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00018996756540081618,
      "loss": 0.3282,
      "step": 553
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00018993121301712193,
      "loss": 0.3516,
      "step": 554
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001898947983824466,
      "loss": 0.4524,
      "step": 555
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00018985832152199663,
      "loss": 0.3308,
      "step": 556
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00018982178246102152,
      "loss": 0.3367,
      "step": 557
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001897851812248139,
      "loss": 0.465,
      "step": 558
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018974851783870936,
      "loss": 0.2814,
      "step": 559
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018971179232808654,
      "loss": 0.2469,
      "step": 560
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018967500471836707,
      "loss": 0.3493,
      "step": 561
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001896381550350156,
      "loss": 0.2774,
      "step": 562
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018960124330353969,
      "loss": 0.4549,
      "step": 563
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001895642695494899,
      "loss": 0.1397,
      "step": 564
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018952723379845975,
      "loss": 0.4118,
      "step": 565
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001894901360760856,
      "loss": 0.5004,
      "step": 566
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018945297640804673,
      "loss": 0.3488,
      "step": 567
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018941575482006527,
      "loss": 0.3371,
      "step": 568
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018937847133790637,
      "loss": 0.2995,
      "step": 569
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018934112598737777,
      "loss": 0.3038,
      "step": 570
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018930371879433025,
      "loss": 0.3156,
      "step": 571
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018926624978465731,
      "loss": 0.3767,
      "step": 572
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001892287189842952,
      "loss": 0.2775,
      "step": 573
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018919112641922303,
      "loss": 0.3759,
      "step": 574
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001891534721154626,
      "loss": 0.3289,
      "step": 575
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018911575609907848,
      "loss": 0.2017,
      "step": 576
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00018907797839617794,
      "loss": 0.0915,
      "step": 577
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000189040139032911,
      "loss": 0.4022,
      "step": 578
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018900223803547023,
      "loss": 0.4652,
      "step": 579
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018896427543009102,
      "loss": 0.2639,
      "step": 580
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018892625124305133,
      "loss": 0.4154,
      "step": 581
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018888816550067175,
      "loss": 0.3917,
      "step": 582
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018885001822931545,
      "loss": 0.1619,
      "step": 583
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018881180945538826,
      "loss": 0.2735,
      "step": 584
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018877353920533852,
      "loss": 0.3536,
      "step": 585
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018873520750565718,
      "loss": 0.3287,
      "step": 586
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018869681438287765,
      "loss": 0.3321,
      "step": 587
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018865835986357597,
      "loss": 0.3887,
      "step": 588
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001886198439743706,
      "loss": 0.2031,
      "step": 589
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001885812667419224,
      "loss": 0.3277,
      "step": 590
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001885426281929349,
      "loss": 0.3978,
      "step": 591
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001885039283541539,
      "loss": 0.3808,
      "step": 592
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018846516725236776,
      "loss": 0.431,
      "step": 593
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018842634491440704,
      "loss": 0.2437,
      "step": 594
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018838746136714494,
      "loss": 0.364,
      "step": 595
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00018834851663749683,
      "loss": 0.2402,
      "step": 596
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018830951075242056,
      "loss": 0.3718,
      "step": 597
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018827044373891623,
      "loss": 0.3139,
      "step": 598
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018823131562402628,
      "loss": 0.3432,
      "step": 599
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001881921264348355,
      "loss": 0.3933,
      "step": 600
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018815287619847087,
      "loss": 0.4458,
      "step": 601
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018811356494210165,
      "loss": 0.3793,
      "step": 602
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001880741926929394,
      "loss": 0.3546,
      "step": 603
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018803475947823782,
      "loss": 0.2399,
      "step": 604
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018799526532529282,
      "loss": 0.3304,
      "step": 605
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018795571026144255,
      "loss": 0.2823,
      "step": 606
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018791609431406724,
      "loss": 0.238,
      "step": 607
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001878764175105894,
      "loss": 0.242,
      "step": 608
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018783667987847352,
      "loss": 0.413,
      "step": 609
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018779688144522622,
      "loss": 0.3367,
      "step": 610
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018775702223839628,
      "loss": 0.222,
      "step": 611
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018771710228557452,
      "loss": 0.3536,
      "step": 612
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018767712161439374,
      "loss": 0.2399,
      "step": 613
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018763708025252887,
      "loss": 0.2757,
      "step": 614
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018759697822769681,
      "loss": 0.3152,
      "step": 615
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018755681556765645,
      "loss": 0.3451,
      "step": 616
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018751659230020862,
      "loss": 0.4001,
      "step": 617
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018747630845319612,
      "loss": 0.3261,
      "step": 618
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018743596405450377,
      "loss": 0.2765,
      "step": 619
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018739555913205815,
      "loss": 0.3791,
      "step": 620
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018735509371382785,
      "loss": 0.2295,
      "step": 621
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018731456782782328,
      "loss": 0.5221,
      "step": 622
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018727398150209675,
      "loss": 0.2809,
      "step": 623
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001872333347647424,
      "loss": 0.3125,
      "step": 624
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018719262764389615,
      "loss": 0.3124,
      "step": 625
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001871518601677357,
      "loss": 0.2489,
      "step": 626
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001871110323644806,
      "loss": 0.3399,
      "step": 627
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018707014426239216,
      "loss": 0.2163,
      "step": 628
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018702919588977336,
      "loss": 0.3024,
      "step": 629
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001869881872749689,
      "loss": 0.3693,
      "step": 630
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001869471184463653,
      "loss": 0.4264,
      "step": 631
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001869059894323906,
      "loss": 0.1319,
      "step": 632
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018686480026151457,
      "loss": 0.2647,
      "step": 633
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00018682355096224872,
      "loss": 0.3384,
      "step": 634
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000186782241563146,
      "loss": 0.3035,
      "step": 635
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001867408720928011,
      "loss": 0.35,
      "step": 636
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001866994425798502,
      "loss": 0.3296,
      "step": 637
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001866579530529711,
      "loss": 0.2873,
      "step": 638
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018661640354088315,
      "loss": 0.3217,
      "step": 639
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018657479407234714,
      "loss": 0.3347,
      "step": 640
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018653312467616545,
      "loss": 0.4092,
      "step": 641
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018649139538118195,
      "loss": 0.3355,
      "step": 642
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001864496062162819,
      "loss": 0.3617,
      "step": 643
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018640775721039202,
      "loss": 0.165,
      "step": 644
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018636584839248048,
      "loss": 0.3588,
      "step": 645
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018632387979155687,
      "loss": 0.2706,
      "step": 646
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001862818514366721,
      "loss": 0.3457,
      "step": 647
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001862397633569185,
      "loss": 0.1984,
      "step": 648
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018619761558142964,
      "loss": 0.317,
      "step": 649
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001861554081393806,
      "loss": 0.432,
      "step": 650
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018611314105998762,
      "loss": 0.2937,
      "step": 651
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018607081437250824,
      "loss": 0.3076,
      "step": 652
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018602842810624124,
      "loss": 0.3591,
      "step": 653
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001859859822905267,
      "loss": 0.2956,
      "step": 654
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018594347695474593,
      "loss": 0.3581,
      "step": 655
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018590091212832135,
      "loss": 0.268,
      "step": 656
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018585828784071663,
      "loss": 0.2474,
      "step": 657
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018581560412143661,
      "loss": 0.3519,
      "step": 658
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018577286100002723,
      "loss": 0.4074,
      "step": 659
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018573005850607555,
      "loss": 0.3848,
      "step": 660
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018568719666920973,
      "loss": 0.3465,
      "step": 661
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018564427551909902,
      "loss": 0.3662,
      "step": 662
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001856012950854538,
      "loss": 0.2099,
      "step": 663
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001855582553980253,
      "loss": 0.3098,
      "step": 664
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001855151564866059,
      "loss": 0.3158,
      "step": 665
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018547199838102904,
      "loss": 0.2888,
      "step": 666
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018542878111116898,
      "loss": 0.3291,
      "step": 667
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018538550470694097,
      "loss": 0.3434,
      "step": 668
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018534216919830125,
      "loss": 0.3885,
      "step": 669
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018529877461524697,
      "loss": 0.263,
      "step": 670
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018525532098781616,
      "loss": 0.3676,
      "step": 671
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018521180834608762,
      "loss": 0.3249,
      "step": 672
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001851682367201812,
      "loss": 0.249,
      "step": 673
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001851246061402574,
      "loss": 0.2799,
      "step": 674
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001850809166365176,
      "loss": 0.2363,
      "step": 675
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000185037168239204,
      "loss": 0.3668,
      "step": 676
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018499336097859947,
      "loss": 0.2529,
      "step": 677
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018494949488502774,
      "loss": 0.2634,
      "step": 678
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001849055699888532,
      "loss": 0.2206,
      "step": 679
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018486158632048097,
      "loss": 0.291,
      "step": 680
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018481754391035677,
      "loss": 0.2579,
      "step": 681
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001847734427889671,
      "loss": 0.3266,
      "step": 682
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018472928298683905,
      "loss": 0.3593,
      "step": 683
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018468506453454033,
      "loss": 0.2697,
      "step": 684
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001846407874626792,
      "loss": 0.2856,
      "step": 685
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001845964518019046,
      "loss": 0.2812,
      "step": 686
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018455205758290594,
      "loss": 0.3632,
      "step": 687
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018450760483641317,
      "loss": 0.3009,
      "step": 688
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018446309359319683,
      "loss": 0.4158,
      "step": 689
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018441852388406785,
      "loss": 0.3867,
      "step": 690
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00018437389573987772,
      "loss": 0.2218,
      "step": 691
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001843292091915183,
      "loss": 0.1111,
      "step": 692
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018428446426992193,
      "loss": 0.229,
      "step": 693
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018423966100606132,
      "loss": 0.2935,
      "step": 694
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001841947994309496,
      "loss": 0.2831,
      "step": 695
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018414987957564028,
      "loss": 0.2941,
      "step": 696
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018410490147122713,
      "loss": 0.227,
      "step": 697
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018405986514884434,
      "loss": 0.2881,
      "step": 698
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001840147706396663,
      "loss": 0.1922,
      "step": 699
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018396961797490776,
      "loss": 0.2996,
      "step": 700
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001839244071858237,
      "loss": 0.2432,
      "step": 701
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001838791383037093,
      "loss": 0.4229,
      "step": 702
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018383381135990003,
      "loss": 0.3569,
      "step": 703
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018378842638577144,
      "loss": 0.3698,
      "step": 704
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018374298341273932,
      "loss": 0.272,
      "step": 705
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018369748247225965,
      "loss": 0.1821,
      "step": 706
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001836519235958284,
      "loss": 0.2802,
      "step": 707
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018360630681498182,
      "loss": 0.2001,
      "step": 708
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018356063216129606,
      "loss": 0.243,
      "step": 709
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001835148996663875,
      "loss": 0.3062,
      "step": 710
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001834691093619124,
      "loss": 0.2812,
      "step": 711
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018342326127956712,
      "loss": 0.2993,
      "step": 712
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018337735545108808,
      "loss": 0.1991,
      "step": 713
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001833313919082515,
      "loss": 0.452,
      "step": 714
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018328537068287374,
      "loss": 0.326,
      "step": 715
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018323929180681092,
      "loss": 0.4507,
      "step": 716
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001831931553119592,
      "loss": 0.3643,
      "step": 717
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018314696123025454,
      "loss": 0.1835,
      "step": 718
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018310070959367278,
      "loss": 0.1948,
      "step": 719
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018305440043422962,
      "loss": 0.1967,
      "step": 720
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001830080337839806,
      "loss": 0.3344,
      "step": 721
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018296160967502098,
      "loss": 0.3147,
      "step": 722
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018291512813948583,
      "loss": 0.3631,
      "step": 723
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018286858920954998,
      "loss": 0.3138,
      "step": 724
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018282199291742805,
      "loss": 0.4408,
      "step": 725
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018277533929537422,
      "loss": 0.1843,
      "step": 726
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018272862837568246,
      "loss": 0.2158,
      "step": 727
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018268186019068642,
      "loss": 0.347,
      "step": 728
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018263503477275927,
      "loss": 0.2315,
      "step": 729
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018258815215431396,
      "loss": 0.329,
      "step": 730
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018254121236780288,
      "loss": 0.2324,
      "step": 731
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001824942154457181,
      "loss": 0.1822,
      "step": 732
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018244716142059114,
      "loss": 0.2761,
      "step": 733
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001824000503249932,
      "loss": 0.2391,
      "step": 734
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018235288219153485,
      "loss": 0.2526,
      "step": 735
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018230565705286617,
      "loss": 0.2403,
      "step": 736
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018225837494167672,
      "loss": 0.2416,
      "step": 737
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018221103589069552,
      "loss": 0.2628,
      "step": 738
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018216363993269093,
      "loss": 0.1825,
      "step": 739
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018211618710047077,
      "loss": 0.1543,
      "step": 740
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001820686774268822,
      "loss": 0.1856,
      "step": 741
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018202111094481173,
      "loss": 0.2236,
      "step": 742
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018197348768718522,
      "loss": 0.2864,
      "step": 743
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018192580768696775,
      "loss": 0.1939,
      "step": 744
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018187807097716374,
      "loss": 0.2532,
      "step": 745
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001818302775908169,
      "loss": 0.3025,
      "step": 746
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00018178242756101013,
      "loss": 0.1924,
      "step": 747
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001817345209208655,
      "loss": 0.2351,
      "step": 748
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001816865577035443,
      "loss": 0.3442,
      "step": 749
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018163853794224703,
      "loss": 0.2365,
      "step": 750
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018159046167021326,
      "loss": 0.3609,
      "step": 751
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001815423289207217,
      "loss": 0.2266,
      "step": 752
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018149413972709012,
      "loss": 0.1991,
      "step": 753
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001814458941226755,
      "loss": 0.3051,
      "step": 754
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018139759214087369,
      "loss": 0.2871,
      "step": 755
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018134923381511966,
      "loss": 0.4585,
      "step": 756
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018130081917888734,
      "loss": 0.2535,
      "step": 757
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018125234826568972,
      "loss": 0.2987,
      "step": 758
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001812038211090786,
      "loss": 0.3086,
      "step": 759
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018115523774264492,
      "loss": 0.2005,
      "step": 760
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001811065982000183,
      "loss": 0.2692,
      "step": 761
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001810579025148674,
      "loss": 0.4413,
      "step": 762
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018100915072089967,
      "loss": 0.2543,
      "step": 763
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001809603428518615,
      "loss": 0.2101,
      "step": 764
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001809114789415379,
      "loss": 0.2243,
      "step": 765
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018086255902375286,
      "loss": 0.2421,
      "step": 766
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001808135831323691,
      "loss": 0.1715,
      "step": 767
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000180764551301288,
      "loss": 0.2534,
      "step": 768
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018071546356444973,
      "loss": 0.3424,
      "step": 769
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018066631995583318,
      "loss": 0.2321,
      "step": 770
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018061712050945584,
      "loss": 0.2226,
      "step": 771
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018056786525937394,
      "loss": 0.2603,
      "step": 772
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018051855423968223,
      "loss": 0.2871,
      "step": 773
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018046918748451418,
      "loss": 0.3288,
      "step": 774
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018041976502804175,
      "loss": 0.3684,
      "step": 775
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018037028690447547,
      "loss": 0.4459,
      "step": 776
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018032075314806448,
      "loss": 0.2651,
      "step": 777
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018027116379309638,
      "loss": 0.204,
      "step": 778
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001802215188738972,
      "loss": 0.1945,
      "step": 779
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018017181842483148,
      "loss": 0.2372,
      "step": 780
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018012206248030224,
      "loss": 0.2937,
      "step": 781
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018007225107475087,
      "loss": 0.3691,
      "step": 782
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00018002238424265712,
      "loss": 0.2545,
      "step": 783
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00017997246201853918,
      "loss": 0.2494,
      "step": 784
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001799224844369535,
      "loss": 0.1958,
      "step": 785
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017987245153249494,
      "loss": 0.2434,
      "step": 786
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001798223633397966,
      "loss": 0.1881,
      "step": 787
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017977221989352982,
      "loss": 0.1827,
      "step": 788
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017972202122840426,
      "loss": 0.2691,
      "step": 789
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017967176737916776,
      "loss": 0.236,
      "step": 790
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017962145838060635,
      "loss": 0.3316,
      "step": 791
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017957109426754427,
      "loss": 0.2439,
      "step": 792
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017952067507484394,
      "loss": 0.2275,
      "step": 793
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017947020083740575,
      "loss": 0.316,
      "step": 794
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001794196715901684,
      "loss": 0.1572,
      "step": 795
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001793690873681085,
      "loss": 0.3193,
      "step": 796
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017931844820624085,
      "loss": 0.327,
      "step": 797
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017926775413961815,
      "loss": 0.2915,
      "step": 798
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001792170052033312,
      "loss": 0.1687,
      "step": 799
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017916620143250877,
      "loss": 0.1305,
      "step": 800
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017911534286231754,
      "loss": 0.2037,
      "step": 801
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00017906442952796213,
      "loss": 0.1302,
      "step": 802
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001790134614646851,
      "loss": 0.3322,
      "step": 803
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001789624387077669,
      "loss": 0.2175,
      "step": 804
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001789113612925258,
      "loss": 0.2276,
      "step": 805
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017886022925431795,
      "loss": 0.2507,
      "step": 806
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017880904262853727,
      "loss": 0.1961,
      "step": 807
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001787578014506155,
      "loss": 0.1695,
      "step": 808
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017870650575602207,
      "loss": 0.2346,
      "step": 809
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017865515558026428,
      "loss": 0.2358,
      "step": 810
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017860375095888698,
      "loss": 0.1687,
      "step": 811
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001785522919274728,
      "loss": 0.1535,
      "step": 812
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017850077852164207,
      "loss": 0.237,
      "step": 813
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001784492107770527,
      "loss": 0.3194,
      "step": 814
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001783975887294002,
      "loss": 0.1756,
      "step": 815
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001783459124144177,
      "loss": 0.1172,
      "step": 816
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017829418186787589,
      "loss": 0.3665,
      "step": 817
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017824239712558302,
      "loss": 0.2561,
      "step": 818
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001781905582233848,
      "loss": 0.2071,
      "step": 819
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017813866519716446,
      "loss": 0.372,
      "step": 820
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017808671808284272,
      "loss": 0.2652,
      "step": 821
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001780347169163777,
      "loss": 0.2799,
      "step": 822
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00017798266173376496,
      "loss": 0.3879,
      "step": 823
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017793055257103743,
      "loss": 0.2589,
      "step": 824
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017787838946426544,
      "loss": 0.1873,
      "step": 825
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001778261724495566,
      "loss": 0.2849,
      "step": 826
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017777390156305592,
      "loss": 0.2214,
      "step": 827
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001777215768409456,
      "loss": 0.3025,
      "step": 828
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001776691983194452,
      "loss": 0.1666,
      "step": 829
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001776167660348115,
      "loss": 0.2567,
      "step": 830
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017756428002333843,
      "loss": 0.1155,
      "step": 831
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017751174032135716,
      "loss": 0.214,
      "step": 832
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017745914696523602,
      "loss": 0.3534,
      "step": 833
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001774064999913805,
      "loss": 0.292,
      "step": 834
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001773537994362332,
      "loss": 0.3356,
      "step": 835
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001773010453362737,
      "loss": 0.2953,
      "step": 836
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017724823772801885,
      "loss": 0.1922,
      "step": 837
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017719537664802234,
      "loss": 0.2613,
      "step": 838
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017714246213287505,
      "loss": 0.1698,
      "step": 839
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017708949421920463,
      "loss": 0.2185,
      "step": 840
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017703647294367597,
      "loss": 0.1696,
      "step": 841
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00017698339834299061,
      "loss": 0.2894,
      "step": 842
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017693027045388726,
      "loss": 0.2221,
      "step": 843
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017687708931314134,
      "loss": 0.3483,
      "step": 844
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017682385495756514,
      "loss": 0.1336,
      "step": 845
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001767705674240079,
      "loss": 0.2697,
      "step": 846
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017671722674935557,
      "loss": 0.2868,
      "step": 847
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017666383297053098,
      "loss": 0.267,
      "step": 848
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017661038612449358,
      "loss": 0.295,
      "step": 849
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017655688624823967,
      "loss": 0.3896,
      "step": 850
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001765033333788022,
      "loss": 0.3024,
      "step": 851
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017644972755325084,
      "loss": 0.1977,
      "step": 852
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017639606880869187,
      "loss": 0.249,
      "step": 853
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017634235718226826,
      "loss": 0.2258,
      "step": 854
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017628859271115953,
      "loss": 0.2794,
      "step": 855
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017623477543258183,
      "loss": 0.1857,
      "step": 856
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001761809053837878,
      "loss": 0.2408,
      "step": 857
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017612698260206666,
      "loss": 0.2486,
      "step": 858
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017607300712474416,
      "loss": 0.1848,
      "step": 859
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001760189789891824,
      "loss": 0.321,
      "step": 860
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00017596489823278006,
      "loss": 0.2427,
      "step": 861
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001759107648929722,
      "loss": 0.1817,
      "step": 862
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017585657900723027,
      "loss": 0.2924,
      "step": 863
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017580234061306208,
      "loss": 0.3916,
      "step": 864
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017574804974801187,
      "loss": 0.1743,
      "step": 865
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017569370644966007,
      "loss": 0.1338,
      "step": 866
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017563931075562347,
      "loss": 0.2866,
      "step": 867
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001755848627035552,
      "loss": 0.3199,
      "step": 868
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017553036233114447,
      "loss": 0.234,
      "step": 869
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017547580967611684,
      "loss": 0.2246,
      "step": 870
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017542120477623401,
      "loss": 0.1688,
      "step": 871
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001753665476692939,
      "loss": 0.4161,
      "step": 872
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001753118383931305,
      "loss": 0.1771,
      "step": 873
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017525707698561385,
      "loss": 0.2802,
      "step": 874
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017520226348465028,
      "loss": 0.226,
      "step": 875
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000175147397928182,
      "loss": 0.3809,
      "step": 876
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017509248035418732,
      "loss": 0.2412,
      "step": 877
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017503751080068052,
      "loss": 0.2435,
      "step": 878
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017498248930571193,
      "loss": 0.1925,
      "step": 879
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00017492741590736778,
      "loss": 0.2112,
      "step": 880
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017487229064377022,
      "loss": 0.27,
      "step": 881
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017481711355307736,
      "loss": 0.3055,
      "step": 882
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001747618846734831,
      "loss": 0.2365,
      "step": 883
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001747066040432173,
      "loss": 0.316,
      "step": 884
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001746512717005455,
      "loss": 0.3422,
      "step": 885
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017459588768376923,
      "loss": 0.1506,
      "step": 886
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017454045203122554,
      "loss": 0.2084,
      "step": 887
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017448496478128742,
      "loss": 0.1769,
      "step": 888
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001744294259723635,
      "loss": 0.2949,
      "step": 889
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017437383564289816,
      "loss": 0.2592,
      "step": 890
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017431819383137133,
      "loss": 0.3036,
      "step": 891
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017426250057629865,
      "loss": 0.3226,
      "step": 892
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001742067559162314,
      "loss": 0.1813,
      "step": 893
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017415095988975636,
      "loss": 0.201,
      "step": 894
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017409511253549593,
      "loss": 0.1749,
      "step": 895
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000174039213892108,
      "loss": 0.2207,
      "step": 896
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000173983263998286,
      "loss": 0.2913,
      "step": 897
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001739272628927588,
      "loss": 0.2617,
      "step": 898
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00017387121061429077,
      "loss": 0.2061,
      "step": 899
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017381510720168163,
      "loss": 0.193,
      "step": 900
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017375895269376655,
      "loss": 0.2258,
      "step": 901
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017370274712941606,
      "loss": 0.2811,
      "step": 902
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017364649054753603,
      "loss": 0.2453,
      "step": 903
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001735901829870676,
      "loss": 0.1704,
      "step": 904
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017353382448698725,
      "loss": 0.1792,
      "step": 905
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017347741508630672,
      "loss": 0.2441,
      "step": 906
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000173420954824073,
      "loss": 0.2107,
      "step": 907
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001733644437393682,
      "loss": 0.1426,
      "step": 908
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001733078818713097,
      "loss": 0.2933,
      "step": 909
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017325126925904998,
      "loss": 0.2136,
      "step": 910
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017319460594177668,
      "loss": 0.2711,
      "step": 911
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001731378919587125,
      "loss": 0.1356,
      "step": 912
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001730811273491153,
      "loss": 0.3089,
      "step": 913
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017302431215227782,
      "loss": 0.2504,
      "step": 914
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017296744640752802,
      "loss": 0.1675,
      "step": 915
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017291053015422867,
      "loss": 0.1151,
      "step": 916
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001728535634317776,
      "loss": 0.3119,
      "step": 917
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00017279654627960756,
      "loss": 0.3784,
      "step": 918
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017273947873718617,
      "loss": 0.2807,
      "step": 919
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000172682360844016,
      "loss": 0.1607,
      "step": 920
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017262519263963442,
      "loss": 0.2942,
      "step": 921
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017256797416361362,
      "loss": 0.2008,
      "step": 922
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017251070545556063,
      "loss": 0.1764,
      "step": 923
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017245338655511724,
      "loss": 0.1843,
      "step": 924
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017239601750195999,
      "loss": 0.2535,
      "step": 925
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017233859833580003,
      "loss": 0.2943,
      "step": 926
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017228112909638334,
      "loss": 0.2516,
      "step": 927
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017222360982349053,
      "loss": 0.2925,
      "step": 928
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001721660405569368,
      "loss": 0.3369,
      "step": 929
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017210842133657197,
      "loss": 0.2365,
      "step": 930
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017205075220228045,
      "loss": 0.2441,
      "step": 931
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017199303319398123,
      "loss": 0.2055,
      "step": 932
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017193526435162773,
      "loss": 0.2948,
      "step": 933
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000171877445715208,
      "loss": 0.1295,
      "step": 934
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001718195773247444,
      "loss": 0.1554,
      "step": 935
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00017176165922029387,
      "loss": 0.1863,
      "step": 936
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001717036914419477,
      "loss": 0.2909,
      "step": 937
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017164567402983152,
      "loss": 0.2992,
      "step": 938
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017158760702410542,
      "loss": 0.1382,
      "step": 939
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017152949046496377,
      "loss": 0.1389,
      "step": 940
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017147132439263518,
      "loss": 0.2193,
      "step": 941
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017141310884738267,
      "loss": 0.3325,
      "step": 942
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017135484386950335,
      "loss": 0.2249,
      "step": 943
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001712965294993287,
      "loss": 0.2402,
      "step": 944
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001712381657772242,
      "loss": 0.2994,
      "step": 945
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017117975274358975,
      "loss": 0.2955,
      "step": 946
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017112129043885915,
      "loss": 0.1845,
      "step": 947
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017106277890350038,
      "loss": 0.0992,
      "step": 948
    },
    {
      "epoch": 0.5,
      "eval_loss": 0.2248118370771408,
      "eval_runtime": 683.3704,
      "eval_samples_per_second": 9.815,
      "eval_steps_per_second": 2.454,
      "step": 948
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017100421817801552,
      "loss": 0.2319,
      "step": 949
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017094560830294074,
      "loss": 0.199,
      "step": 950
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017088694931884617,
      "loss": 0.2376,
      "step": 951
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017082824126633591,
      "loss": 0.2192,
      "step": 952
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001707694841860481,
      "loss": 0.2483,
      "step": 953
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017071067811865476,
      "loss": 0.3265,
      "step": 954
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00017065182310486185,
      "loss": 0.1277,
      "step": 955
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001705929191854092,
      "loss": 0.1946,
      "step": 956
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00017053396640107047,
      "loss": 0.1577,
      "step": 957
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00017047496479265318,
      "loss": 0.238,
      "step": 958
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00017041591440099867,
      "loss": 0.1685,
      "step": 959
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00017035681526698197,
      "loss": 0.1262,
      "step": 960
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00017029766743151189,
      "loss": 0.2205,
      "step": 961
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00017023847093553102,
      "loss": 0.2684,
      "step": 962
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00017017922582001545,
      "loss": 0.2532,
      "step": 963
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00017011993212597514,
      "loss": 0.2328,
      "step": 964
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00017006058989445352,
      "loss": 0.2268,
      "step": 965
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001700011991665277,
      "loss": 0.2321,
      "step": 966
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00016994175998330836,
      "loss": 0.1233,
      "step": 967
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00016988227238593962,
      "loss": 0.2376,
      "step": 968
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001698227364155993,
      "loss": 0.2735,
      "step": 969
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001697631521134985,
      "loss": 0.1292,
      "step": 970
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00016970351952088188,
      "loss": 0.1159,
      "step": 971
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00016964383867902757,
      "loss": 0.1861,
      "step": 972
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00016958410962924697,
      "loss": 0.1998,
      "step": 973
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000169524332412885,
      "loss": 0.2399,
      "step": 974
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00016946450707131978,
      "loss": 0.2197,
      "step": 975
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016940463364596283,
      "loss": 0.2019,
      "step": 976
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016934471217825898,
      "loss": 0.2025,
      "step": 977
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001692847427096862,
      "loss": 0.2717,
      "step": 978
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016922472528175576,
      "loss": 0.1484,
      "step": 979
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016916465993601216,
      "loss": 0.1891,
      "step": 980
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016910454671403299,
      "loss": 0.318,
      "step": 981
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016904438565742904,
      "loss": 0.2791,
      "step": 982
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016898417680784422,
      "loss": 0.1819,
      "step": 983
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016892392020695542,
      "loss": 0.3261,
      "step": 984
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001688636158964727,
      "loss": 0.304,
      "step": 985
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016880326391813916,
      "loss": 0.1509,
      "step": 986
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001687428643137307,
      "loss": 0.1823,
      "step": 987
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001686824171250565,
      "loss": 0.2557,
      "step": 988
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016862192239395837,
      "loss": 0.2367,
      "step": 989
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0001685613801623112,
      "loss": 0.1881,
      "step": 990
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016850079047202275,
      "loss": 0.2442,
      "step": 991
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016844015336503357,
      "loss": 0.1624,
      "step": 992
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016837946888331707,
      "loss": 0.2578,
      "step": 993
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00016831873706887952,
      "loss": 0.1803,
      "step": 994
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016825795796375978,
      "loss": 0.159,
      "step": 995
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0001681971316100296,
      "loss": 0.2785,
      "step": 996
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016813625804979338,
      "loss": 0.2227,
      "step": 997
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016807533732518821,
      "loss": 0.1191,
      "step": 998
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016801436947838384,
      "loss": 0.3066,
      "step": 999
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016795335455158252,
      "loss": 0.1774,
      "step": 1000
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0001678922925870193,
      "loss": 0.1843,
      "step": 1001
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016783118362696163,
      "loss": 0.1607,
      "step": 1002
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016777002771370953,
      "loss": 0.2425,
      "step": 1003
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016770882488959554,
      "loss": 0.1218,
      "step": 1004
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016764757519698465,
      "loss": 0.2003,
      "step": 1005
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0001675862786782743,
      "loss": 0.1496,
      "step": 1006
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016752493537589437,
      "loss": 0.3887,
      "step": 1007
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016746354533230705,
      "loss": 0.1598,
      "step": 1008
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000167402108590007,
      "loss": 0.1702,
      "step": 1009
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016734062519152112,
      "loss": 0.2758,
      "step": 1010
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0001672790951794086,
      "loss": 0.2565,
      "step": 1011
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016721751859626095,
      "loss": 0.2822,
      "step": 1012
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00016715589548470185,
      "loss": 0.2398,
      "step": 1013
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016709422588738726,
      "loss": 0.2,
      "step": 1014
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016703250984700528,
      "loss": 0.2886,
      "step": 1015
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016697074740627615,
      "loss": 0.2014,
      "step": 1016
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016690893860795227,
      "loss": 0.2151,
      "step": 1017
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016684708349481804,
      "loss": 0.2235,
      "step": 1018
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016678518210969,
      "loss": 0.1507,
      "step": 1019
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0001667232344954168,
      "loss": 0.281,
      "step": 1020
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016666124069487882,
      "loss": 0.2057,
      "step": 1021
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016659920075098866,
      "loss": 0.2804,
      "step": 1022
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016653711470669084,
      "loss": 0.103,
      "step": 1023
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016647498260496165,
      "loss": 0.2476,
      "step": 1024
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016641280448880934,
      "loss": 0.2261,
      "step": 1025
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016635058040127406,
      "loss": 0.2407,
      "step": 1026
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0001662883103854277,
      "loss": 0.282,
      "step": 1027
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000166225994484374,
      "loss": 0.2101,
      "step": 1028
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016616363274124844,
      "loss": 0.2921,
      "step": 1029
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00016610122519921817,
      "loss": 0.1365,
      "step": 1030
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0001660387719014822,
      "loss": 0.2909,
      "step": 1031
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001659762728912711,
      "loss": 0.2805,
      "step": 1032
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016591372821184703,
      "loss": 0.2616,
      "step": 1033
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016585113790650388,
      "loss": 0.1887,
      "step": 1034
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016578850201856704,
      "loss": 0.0801,
      "step": 1035
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016572582059139353,
      "loss": 0.2332,
      "step": 1036
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016566309366837186,
      "loss": 0.1621,
      "step": 1037
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016560032129292197,
      "loss": 0.1919,
      "step": 1038
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016553750350849534,
      "loss": 0.3352,
      "step": 1039
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016547464035857482,
      "loss": 0.2112,
      "step": 1040
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001654117318866748,
      "loss": 0.0951,
      "step": 1041
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001653487781363408,
      "loss": 0.2601,
      "step": 1042
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016528577915114992,
      "loss": 0.3096,
      "step": 1043
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016522273497471046,
      "loss": 0.1938,
      "step": 1044
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016515964565066199,
      "loss": 0.3066,
      "step": 1045
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001650965112226754,
      "loss": 0.1719,
      "step": 1046
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001650333317344527,
      "loss": 0.1368,
      "step": 1047
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016497010722972717,
      "loss": 0.2769,
      "step": 1048
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001649068377522633,
      "loss": 0.2058,
      "step": 1049
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00016484352334585653,
      "loss": 0.1596,
      "step": 1050
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001647801640543336,
      "loss": 0.1181,
      "step": 1051
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016471675992155217,
      "loss": 0.1843,
      "step": 1052
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016465331099140102,
      "loss": 0.2563,
      "step": 1053
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016458981730779998,
      "loss": 0.1575,
      "step": 1054
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001645262789146997,
      "loss": 0.1867,
      "step": 1055
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016446269585608194,
      "loss": 0.2861,
      "step": 1056
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016439906817595929,
      "loss": 0.2191,
      "step": 1057
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016433539591837525,
      "loss": 0.1784,
      "step": 1058
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001642716791274042,
      "loss": 0.3081,
      "step": 1059
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001642079178471513,
      "loss": 0.274,
      "step": 1060
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016414411212175255,
      "loss": 0.1619,
      "step": 1061
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016408026199537468,
      "loss": 0.1497,
      "step": 1062
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016401636751221517,
      "loss": 0.13,
      "step": 1063
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001639524287165022,
      "loss": 0.2481,
      "step": 1064
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016388844565249464,
      "loss": 0.2375,
      "step": 1065
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016382441836448202,
      "loss": 0.3027,
      "step": 1066
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001637603468967844,
      "loss": 0.2743,
      "step": 1067
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0001636962312937525,
      "loss": 0.1273,
      "step": 1068
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00016363207159976755,
      "loss": 0.0966,
      "step": 1069
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016356786785924135,
      "loss": 0.1976,
      "step": 1070
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001635036201166161,
      "loss": 0.2194,
      "step": 1071
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016343932841636456,
      "loss": 0.1342,
      "step": 1072
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016337499280298984,
      "loss": 0.1649,
      "step": 1073
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001633106133210255,
      "loss": 0.1314,
      "step": 1074
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016324619001503545,
      "loss": 0.1961,
      "step": 1075
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016318172292961389,
      "loss": 0.2579,
      "step": 1076
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001631172121093854,
      "loss": 0.2228,
      "step": 1077
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016305265759900477,
      "loss": 0.2131,
      "step": 1078
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016298805944315708,
      "loss": 0.2774,
      "step": 1079
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016292341768655758,
      "loss": 0.2261,
      "step": 1080
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016285873237395176,
      "loss": 0.2141,
      "step": 1081
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001627940035501152,
      "loss": 0.1791,
      "step": 1082
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016272923125985363,
      "loss": 0.2808,
      "step": 1083
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0001626644155480028,
      "loss": 0.1548,
      "step": 1084
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016259955645942864,
      "loss": 0.2639,
      "step": 1085
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016253465403902702,
      "loss": 0.1705,
      "step": 1086
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016246970833172383,
      "loss": 0.2487,
      "step": 1087
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00016240471938247492,
      "loss": 0.2088,
      "step": 1088
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016233968723626608,
      "loss": 0.212,
      "step": 1089
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000162274611938113,
      "loss": 0.195,
      "step": 1090
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016220949353306114,
      "loss": 0.2121,
      "step": 1091
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000162144332066186,
      "loss": 0.2462,
      "step": 1092
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016207912758259272,
      "loss": 0.2408,
      "step": 1093
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016201388012741634,
      "loss": 0.2541,
      "step": 1094
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001619485897458215,
      "loss": 0.2374,
      "step": 1095
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016188325648300268,
      "loss": 0.231,
      "step": 1096
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016181788038418398,
      "loss": 0.175,
      "step": 1097
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001617524614946192,
      "loss": 0.1361,
      "step": 1098
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001616869998595917,
      "loss": 0.1256,
      "step": 1099
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016162149552441445,
      "loss": 0.2966,
      "step": 1100
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016155594853443,
      "loss": 0.1651,
      "step": 1101
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001614903589350104,
      "loss": 0.3146,
      "step": 1102
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016142472677155723,
      "loss": 0.2281,
      "step": 1103
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001613590520895015,
      "loss": 0.1834,
      "step": 1104
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016129333493430364,
      "loss": 0.1806,
      "step": 1105
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00016122757535145347,
      "loss": 0.2328,
      "step": 1106
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0001611617733864703,
      "loss": 0.2483,
      "step": 1107
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00016109592908490261,
      "loss": 0.2097,
      "step": 1108
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00016103004249232828,
      "loss": 0.1533,
      "step": 1109
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001609641136543545,
      "loss": 0.1207,
      "step": 1110
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00016089814261661752,
      "loss": 0.1884,
      "step": 1111
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001608321294247831,
      "loss": 0.1436,
      "step": 1112
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00016076607412454584,
      "loss": 0.1678,
      "step": 1113
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001606999767616298,
      "loss": 0.0973,
      "step": 1114
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001606338373817879,
      "loss": 0.2954,
      "step": 1115
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001605676560308023,
      "loss": 0.1868,
      "step": 1116
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001605014327544842,
      "loss": 0.117,
      "step": 1117
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00016043516759867374,
      "loss": 0.1519,
      "step": 1118
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00016036886060924013,
      "loss": 0.2882,
      "step": 1119
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001603025118320815,
      "loss": 0.2382,
      "step": 1120
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001602361213131249,
      "loss": 0.107,
      "step": 1121
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00016016968909832632,
      "loss": 0.1723,
      "step": 1122
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00016010321523367055,
      "loss": 0.2484,
      "step": 1123
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00016003669976517125,
      "loss": 0.2224,
      "step": 1124
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00015997014273887085,
      "loss": 0.3619,
      "step": 1125
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00015990354420084056,
      "loss": 0.1407,
      "step": 1126
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015983690419718038,
      "loss": 0.248,
      "step": 1127
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015977022277401893,
      "loss": 0.1617,
      "step": 1128
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015970349997751355,
      "loss": 0.1832,
      "step": 1129
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015963673585385016,
      "loss": 0.1627,
      "step": 1130
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015956993044924334,
      "loss": 0.2363,
      "step": 1131
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015950308380993625,
      "loss": 0.1417,
      "step": 1132
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001594361959822006,
      "loss": 0.2745,
      "step": 1133
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015936926701233653,
      "loss": 0.289,
      "step": 1134
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015930229694667276,
      "loss": 0.1142,
      "step": 1135
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001592352858315664,
      "loss": 0.2093,
      "step": 1136
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015916823371340296,
      "loss": 0.1656,
      "step": 1137
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001591011406385964,
      "loss": 0.1432,
      "step": 1138
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000159034006653589,
      "loss": 0.2203,
      "step": 1139
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001589668318048513,
      "loss": 0.1496,
      "step": 1140
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001588996161388822,
      "loss": 0.2327,
      "step": 1141
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015883235970220886,
      "loss": 0.1868,
      "step": 1142
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015876506254138657,
      "loss": 0.1784,
      "step": 1143
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0001586977247029989,
      "loss": 0.1543,
      "step": 1144
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00015863034623365754,
      "loss": 0.2577,
      "step": 1145
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015856292718000235,
      "loss": 0.1713,
      "step": 1146
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001584954675887012,
      "loss": 0.1933,
      "step": 1147
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015842796750645008,
      "loss": 0.251,
      "step": 1148
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000158360426979973,
      "loss": 0.2689,
      "step": 1149
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015829284605602193,
      "loss": 0.1312,
      "step": 1150
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015822522478137693,
      "loss": 0.1173,
      "step": 1151
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015815756320284577,
      "loss": 0.1227,
      "step": 1152
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015808986136726428,
      "loss": 0.2088,
      "step": 1153
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015802211932149614,
      "loss": 0.2212,
      "step": 1154
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015795433711243286,
      "loss": 0.2685,
      "step": 1155
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015788651478699367,
      "loss": 0.1998,
      "step": 1156
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015781865239212573,
      "loss": 0.274,
      "step": 1157
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015775074997480374,
      "loss": 0.1282,
      "step": 1158
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001576828075820303,
      "loss": 0.3618,
      "step": 1159
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015761482526083546,
      "loss": 0.2845,
      "step": 1160
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015754680305827713,
      "loss": 0.1266,
      "step": 1161
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0001574787410214407,
      "loss": 0.1431,
      "step": 1162
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015741063919743917,
      "loss": 0.1556,
      "step": 1163
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00015734249763341306,
      "loss": 0.2614,
      "step": 1164
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015727431637653041,
      "loss": 0.238,
      "step": 1165
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015720609547398677,
      "loss": 0.2459,
      "step": 1166
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015713783497300505,
      "loss": 0.1817,
      "step": 1167
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015706953492083568,
      "loss": 0.0959,
      "step": 1168
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015700119536475634,
      "loss": 0.1364,
      "step": 1169
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015693281635207216,
      "loss": 0.117,
      "step": 1170
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015686439793011556,
      "loss": 0.2701,
      "step": 1171
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015679594014624622,
      "loss": 0.207,
      "step": 1172
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015672744304785104,
      "loss": 0.1699,
      "step": 1173
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001566589066823442,
      "loss": 0.2169,
      "step": 1174
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015659033109716698,
      "loss": 0.0678,
      "step": 1175
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015652171633978793,
      "loss": 0.2328,
      "step": 1176
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015645306245770256,
      "loss": 0.2603,
      "step": 1177
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001563843694984336,
      "loss": 0.2114,
      "step": 1178
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015631563750953076,
      "loss": 0.2167,
      "step": 1179
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015624686653857074,
      "loss": 0.3016,
      "step": 1180
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0001561780566331573,
      "loss": 0.1622,
      "step": 1181
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015610920784092107,
      "loss": 0.1814,
      "step": 1182
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00015604032020951968,
      "loss": 0.182,
      "step": 1183
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0001559713937866376,
      "loss": 0.168,
      "step": 1184
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015590242861998616,
      "loss": 0.1222,
      "step": 1185
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0001558334247573035,
      "loss": 0.1175,
      "step": 1186
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0001557643822463545,
      "loss": 0.2246,
      "step": 1187
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015569530113493095,
      "loss": 0.1939,
      "step": 1188
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015562618147085114,
      "loss": 0.2485,
      "step": 1189
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015555702330196023,
      "loss": 0.333,
      "step": 1190
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015548782667612995,
      "loss": 0.1979,
      "step": 1191
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015541859164125863,
      "loss": 0.1355,
      "step": 1192
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015534931824527127,
      "loss": 0.18,
      "step": 1193
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015528000653611935,
      "loss": 0.2431,
      "step": 1194
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015521065656178088,
      "loss": 0.1738,
      "step": 1195
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0001551412683702604,
      "loss": 0.3208,
      "step": 1196
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015507184200958885,
      "loss": 0.2392,
      "step": 1197
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015500237752782362,
      "loss": 0.2441,
      "step": 1198
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015493287497304848,
      "loss": 0.2075,
      "step": 1199
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015486333439337357,
      "loss": 0.0869,
      "step": 1200
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015479375583693535,
      "loss": 0.1392,
      "step": 1201
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00015472413935189655,
      "loss": 0.2814,
      "step": 1202
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015465448498644616,
      "loss": 0.2047,
      "step": 1203
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015458479278879938,
      "loss": 0.1823,
      "step": 1204
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015451506280719763,
      "loss": 0.1552,
      "step": 1205
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015444529508990844,
      "loss": 0.2431,
      "step": 1206
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001543754896852255,
      "loss": 0.1104,
      "step": 1207
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001543056466414686,
      "loss": 0.2062,
      "step": 1208
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015423576600698353,
      "loss": 0.1686,
      "step": 1209
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001541658478301421,
      "loss": 0.2138,
      "step": 1210
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015409589215934218,
      "loss": 0.2286,
      "step": 1211
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015402589904300754,
      "loss": 0.1782,
      "step": 1212
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015395586852958788,
      "loss": 0.2035,
      "step": 1213
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015388580066755876,
      "loss": 0.1019,
      "step": 1214
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015381569550542166,
      "loss": 0.2119,
      "step": 1215
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001537455530917038,
      "loss": 0.1624,
      "step": 1216
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015367537347495822,
      "loss": 0.089,
      "step": 1217
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015360515670376375,
      "loss": 0.1451,
      "step": 1218
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00015353490282672488,
      "loss": 0.0915,
      "step": 1219
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001534646118924718,
      "loss": 0.2215,
      "step": 1220
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001533942839496604,
      "loss": 0.171,
      "step": 1221
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015332391904697212,
      "loss": 0.2381,
      "step": 1222
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000153253517233114,
      "loss": 0.268,
      "step": 1223
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015318307855681867,
      "loss": 0.2391,
      "step": 1224
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015311260306684422,
      "loss": 0.2322,
      "step": 1225
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015304209081197425,
      "loss": 0.2244,
      "step": 1226
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015297154184101786,
      "loss": 0.1248,
      "step": 1227
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015290095620280945,
      "loss": 0.2095,
      "step": 1228
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0001528303339462089,
      "loss": 0.1651,
      "step": 1229
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0001527596751201014,
      "loss": 0.0803,
      "step": 1230
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015268897977339748,
      "loss": 0.212,
      "step": 1231
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015261824795503287,
      "loss": 0.1429,
      "step": 1232
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015254747971396866,
      "loss": 0.144,
      "step": 1233
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015247667509919104,
      "loss": 0.3074,
      "step": 1234
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015240583415971148,
      "loss": 0.1789,
      "step": 1235
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015233495694456654,
      "loss": 0.1965,
      "step": 1236
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015226404350281788,
      "loss": 0.1805,
      "step": 1237
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015219309388355226,
      "loss": 0.1658,
      "step": 1238
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0001521221081358815,
      "loss": 0.2434,
      "step": 1239
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00015205108630894239,
      "loss": 0.2157,
      "step": 1240
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015198002845189665,
      "loss": 0.1305,
      "step": 1241
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015190893461393108,
      "loss": 0.2222,
      "step": 1242
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015183780484425726,
      "loss": 0.2424,
      "step": 1243
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015176663919211177,
      "loss": 0.0928,
      "step": 1244
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015169543770675577,
      "loss": 0.1543,
      "step": 1245
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0001516242004374756,
      "loss": 0.0936,
      "step": 1246
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015155292743358208,
      "loss": 0.1779,
      "step": 1247
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0001514816187444108,
      "loss": 0.1012,
      "step": 1248
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015141027441932216,
      "loss": 0.2297,
      "step": 1249
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015133889450770122,
      "loss": 0.2623,
      "step": 1250
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015126747905895758,
      "loss": 0.2347,
      "step": 1251
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015119602812252547,
      "loss": 0.1429,
      "step": 1252
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015112454174786373,
      "loss": 0.3506,
      "step": 1253
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015105301998445566,
      "loss": 0.214,
      "step": 1254
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015098146288180915,
      "loss": 0.0679,
      "step": 1255
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015090987048945652,
      "loss": 0.1214,
      "step": 1256
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015083824285695442,
      "loss": 0.2185,
      "step": 1257
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000150766580033884,
      "loss": 0.2326,
      "step": 1258
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00015069488206985077,
      "loss": 0.1628,
      "step": 1259
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00015062314901448452,
      "loss": 0.0787,
      "step": 1260
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00015055138091743934,
      "loss": 0.1122,
      "step": 1261
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00015047957782839358,
      "loss": 0.254,
      "step": 1262
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00015040773979704982,
      "loss": 0.1228,
      "step": 1263
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0001503358668731348,
      "loss": 0.1028,
      "step": 1264
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00015026395910639948,
      "loss": 0.1406,
      "step": 1265
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00015019201654661884,
      "loss": 0.1823,
      "step": 1266
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00015012003924359202,
      "loss": 0.212,
      "step": 1267
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00015004802724714224,
      "loss": 0.1563,
      "step": 1268
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014997598060711658,
      "loss": 0.0587,
      "step": 1269
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014990389937338624,
      "loss": 0.249,
      "step": 1270
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014983178359584634,
      "loss": 0.2052,
      "step": 1271
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014975963332441593,
      "loss": 0.1589,
      "step": 1272
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014968744860903782,
      "loss": 0.089,
      "step": 1273
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014961522949967886,
      "loss": 0.1644,
      "step": 1274
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014954297604632947,
      "loss": 0.0875,
      "step": 1275
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014947068829900405,
      "loss": 0.1373,
      "step": 1276
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014939836630774067,
      "loss": 0.1384,
      "step": 1277
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00014932601012260105,
      "loss": 0.1602,
      "step": 1278
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0001492536197936706,
      "loss": 0.162,
      "step": 1279
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014918119537105844,
      "loss": 0.1146,
      "step": 1280
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014910873690489725,
      "loss": 0.1141,
      "step": 1281
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014903624444534316,
      "loss": 0.1685,
      "step": 1282
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000148963718042576,
      "loss": 0.0956,
      "step": 1283
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014889115774679906,
      "loss": 0.305,
      "step": 1284
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000148818563608239,
      "loss": 0.1632,
      "step": 1285
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014874593567714594,
      "loss": 0.1049,
      "step": 1286
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014867327400379347,
      "loss": 0.1546,
      "step": 1287
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0001486005786384785,
      "loss": 0.1571,
      "step": 1288
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014852784963152117,
      "loss": 0.2059,
      "step": 1289
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014845508703326504,
      "loss": 0.1959,
      "step": 1290
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014838229089407683,
      "loss": 0.1803,
      "step": 1291
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0001483094612643465,
      "loss": 0.318,
      "step": 1292
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014823659819448724,
      "loss": 0.1568,
      "step": 1293
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014816370173493533,
      "loss": 0.1859,
      "step": 1294
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014809077193615023,
      "loss": 0.2267,
      "step": 1295
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00014801780884861432,
      "loss": 0.3156,
      "step": 1296
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0001479448125228332,
      "loss": 0.1866,
      "step": 1297
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014787178300933542,
      "loss": 0.2256,
      "step": 1298
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014779872035867252,
      "loss": 0.3029,
      "step": 1299
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014772562462141882,
      "loss": 0.1458,
      "step": 1300
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014765249584817184,
      "loss": 0.186,
      "step": 1301
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014757933408955168,
      "loss": 0.3387,
      "step": 1302
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014750613939620144,
      "loss": 0.1855,
      "step": 1303
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014743291181878697,
      "loss": 0.3866,
      "step": 1304
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0001473596514079969,
      "loss": 0.2217,
      "step": 1305
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014728635821454255,
      "loss": 0.2811,
      "step": 1306
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014721303228915793,
      "loss": 0.1642,
      "step": 1307
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0001471396736825998,
      "loss": 0.1376,
      "step": 1308
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0001470662824456474,
      "loss": 0.1926,
      "step": 1309
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014699285862910264,
      "loss": 0.2456,
      "step": 1310
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014691940228379004,
      "loss": 0.159,
      "step": 1311
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014684591346055648,
      "loss": 0.1393,
      "step": 1312
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014677239221027149,
      "loss": 0.2164,
      "step": 1313
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0001466988385838269,
      "loss": 0.1871,
      "step": 1314
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00014662525263213708,
      "loss": 0.2302,
      "step": 1315
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014655163440613865,
      "loss": 0.1361,
      "step": 1316
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014647798395679067,
      "loss": 0.2063,
      "step": 1317
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014640430133507445,
      "loss": 0.1093,
      "step": 1318
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014633058659199364,
      "loss": 0.1759,
      "step": 1319
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014625683977857398,
      "loss": 0.195,
      "step": 1320
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014618306094586357,
      "loss": 0.2586,
      "step": 1321
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0001461092501449326,
      "loss": 0.1396,
      "step": 1322
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0001460354074268734,
      "loss": 0.1524,
      "step": 1323
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014596153284280035,
      "loss": 0.2524,
      "step": 1324
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014588762644384993,
      "loss": 0.1327,
      "step": 1325
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0001458136882811807,
      "loss": 0.2365,
      "step": 1326
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014573971840597312,
      "loss": 0.1512,
      "step": 1327
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014566571686942955,
      "loss": 0.2322,
      "step": 1328
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014559168372277448,
      "loss": 0.1987,
      "step": 1329
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014551761901725402,
      "loss": 0.1151,
      "step": 1330
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014544352280413628,
      "loss": 0.1629,
      "step": 1331
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014536939513471123,
      "loss": 0.2446,
      "step": 1332
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014529523606029044,
      "loss": 0.3006,
      "step": 1333
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00014522104563220736,
      "loss": 0.276,
      "step": 1334
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014514682390181707,
      "loss": 0.2363,
      "step": 1335
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014507257092049637,
      "loss": 0.1,
      "step": 1336
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014499828673964364,
      "loss": 0.1606,
      "step": 1337
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014492397141067887,
      "loss": 0.1774,
      "step": 1338
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014484962498504368,
      "loss": 0.1974,
      "step": 1339
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014477524751420112,
      "loss": 0.0979,
      "step": 1340
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014470083904963573,
      "loss": 0.2117,
      "step": 1341
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0001446263996428536,
      "loss": 0.1755,
      "step": 1342
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014455192934538218,
      "loss": 0.1637,
      "step": 1343
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0001444774282087703,
      "loss": 0.1108,
      "step": 1344
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014440289628458807,
      "loss": 0.1364,
      "step": 1345
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0001443283336244271,
      "loss": 0.2803,
      "step": 1346
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014425374027990005,
      "loss": 0.073,
      "step": 1347
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014417911630264097,
      "loss": 0.1393,
      "step": 1348
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014410446174430508,
      "loss": 0.1491,
      "step": 1349
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014402977665656876,
      "loss": 0.22,
      "step": 1350
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0001439550610911295,
      "loss": 0.2035,
      "step": 1351
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014388031509970591,
      "loss": 0.3069,
      "step": 1352
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00014380553873403774,
      "loss": 0.2131,
      "step": 1353
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014373073204588556,
      "loss": 0.1682,
      "step": 1354
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0001436558950870311,
      "loss": 0.2059,
      "step": 1355
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014358102790927704,
      "loss": 0.092,
      "step": 1356
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014350613056444693,
      "loss": 0.248,
      "step": 1357
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014343120310438513,
      "loss": 0.1759,
      "step": 1358
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014335624558095702,
      "loss": 0.1482,
      "step": 1359
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014328125804604863,
      "loss": 0.0913,
      "step": 1360
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0001432062405515669,
      "loss": 0.14,
      "step": 1361
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014313119314943932,
      "loss": 0.1466,
      "step": 1362
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0001430561158916143,
      "loss": 0.219,
      "step": 1363
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0001429810088300608,
      "loss": 0.1564,
      "step": 1364
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014290587201676837,
      "loss": 0.2567,
      "step": 1365
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014283070550374726,
      "loss": 0.1253,
      "step": 1366
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014275550934302823,
      "loss": 0.097,
      "step": 1367
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014268028358666252,
      "loss": 0.1767,
      "step": 1368
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014260502828672192,
      "loss": 0.0934,
      "step": 1369
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0001425297434952987,
      "loss": 0.1582,
      "step": 1370
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014245442926450537,
      "loss": 0.2114,
      "step": 1371
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00014237908564647503,
      "loss": 0.1793,
      "step": 1372
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014230371269336104,
      "loss": 0.1359,
      "step": 1373
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014222831045733702,
      "loss": 0.2619,
      "step": 1374
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014215287899059688,
      "loss": 0.2032,
      "step": 1375
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001420774183453548,
      "loss": 0.1677,
      "step": 1376
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001420019285738452,
      "loss": 0.1659,
      "step": 1377
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001419264097283225,
      "loss": 0.189,
      "step": 1378
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001418508618610614,
      "loss": 0.1473,
      "step": 1379
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0001417752850243566,
      "loss": 0.1423,
      "step": 1380
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014169967927052293,
      "loss": 0.1249,
      "step": 1381
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014162404465189516,
      "loss": 0.2195,
      "step": 1382
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014154838122082807,
      "loss": 0.3112,
      "step": 1383
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014147268902969643,
      "loss": 0.1084,
      "step": 1384
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014139696813089485,
      "loss": 0.132,
      "step": 1385
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014132121857683783,
      "loss": 0.2931,
      "step": 1386
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014124544041995972,
      "loss": 0.2135,
      "step": 1387
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014116963371271464,
      "loss": 0.2767,
      "step": 1388
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014109379850757656,
      "loss": 0.1927,
      "step": 1389
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014101793485703904,
      "loss": 0.11,
      "step": 1390
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00014094204281361544,
      "loss": 0.2897,
      "step": 1391
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014086612242983872,
      "loss": 0.1441,
      "step": 1392
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014079017375826148,
      "loss": 0.1932,
      "step": 1393
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014071419685145585,
      "loss": 0.0787,
      "step": 1394
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014063819176201366,
      "loss": 0.2758,
      "step": 1395
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000140562158542546,
      "loss": 0.2632,
      "step": 1396
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014048609724568367,
      "loss": 0.1423,
      "step": 1397
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014041000792407675,
      "loss": 0.1784,
      "step": 1398
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001403338906303948,
      "loss": 0.1776,
      "step": 1399
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014025774541732666,
      "loss": 0.0947,
      "step": 1400
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014018157233758064,
      "loss": 0.1405,
      "step": 1401
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014010537144388416,
      "loss": 0.0845,
      "step": 1402
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00014002914278898407,
      "loss": 0.1475,
      "step": 1403
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001399528864256463,
      "loss": 0.1396,
      "step": 1404
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00013987660240665599,
      "loss": 0.168,
      "step": 1405
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00013980029078481746,
      "loss": 0.1548,
      "step": 1406
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00013972395161295414,
      "loss": 0.1024,
      "step": 1407
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00013964758494390854,
      "loss": 0.1519,
      "step": 1408
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00013957119083054205,
      "loss": 0.1565,
      "step": 1409
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0001394947693257353,
      "loss": 0.1061,
      "step": 1410
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013941832048238777,
      "loss": 0.0608,
      "step": 1411
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013934184435341777,
      "loss": 0.2847,
      "step": 1412
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001392653409917626,
      "loss": 0.2893,
      "step": 1413
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013918881045037846,
      "loss": 0.1615,
      "step": 1414
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013911225278224023,
      "loss": 0.1766,
      "step": 1415
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013903566804034166,
      "loss": 0.1391,
      "step": 1416
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013895905627769523,
      "loss": 0.2277,
      "step": 1417
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013888241754733208,
      "loss": 0.1513,
      "step": 1418
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013880575190230204,
      "loss": 0.1631,
      "step": 1419
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013872905939567363,
      "loss": 0.1735,
      "step": 1420
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013865234008053387,
      "loss": 0.1803,
      "step": 1421
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013857559400998838,
      "loss": 0.1897,
      "step": 1422
    },
    {
      "epoch": 0.75,
      "eval_loss": 0.16311943531036377,
      "eval_runtime": 683.209,
      "eval_samples_per_second": 9.817,
      "eval_steps_per_second": 2.455,
      "step": 1422
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001384988212371613,
      "loss": 0.1545,
      "step": 1423
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001384220218151953,
      "loss": 0.1964,
      "step": 1424
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013834519579725133,
      "loss": 0.1159,
      "step": 1425
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000138268343236509,
      "loss": 0.1972,
      "step": 1426
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013819146418616606,
      "loss": 0.2127,
      "step": 1427
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00013811455869943876,
      "loss": 0.1761,
      "step": 1428
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001380376268295616,
      "loss": 0.1634,
      "step": 1429
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013796066862978724,
      "loss": 0.1931,
      "step": 1430
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013788368415338677,
      "loss": 0.0715,
      "step": 1431
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013780667345364927,
      "loss": 0.1371,
      "step": 1432
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001377296365838821,
      "loss": 0.1818,
      "step": 1433
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013765257359741063,
      "loss": 0.323,
      "step": 1434
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013757548454757846,
      "loss": 0.1194,
      "step": 1435
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013749836948774707,
      "loss": 0.1639,
      "step": 1436
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001374212284712961,
      "loss": 0.109,
      "step": 1437
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013734406155162292,
      "loss": 0.2091,
      "step": 1438
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013726686878214313,
      "loss": 0.1529,
      "step": 1439
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013718965021629003,
      "loss": 0.1716,
      "step": 1440
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001371124059075148,
      "loss": 0.1304,
      "step": 1441
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013703513590928646,
      "loss": 0.2382,
      "step": 1442
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013695784027509185,
      "loss": 0.1077,
      "step": 1443
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013688051905843548,
      "loss": 0.1538,
      "step": 1444
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013680317231283965,
      "loss": 0.208,
      "step": 1445
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001367258000918443,
      "loss": 0.1104,
      "step": 1446
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00013664840244900688,
      "loss": 0.2242,
      "step": 1447
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0001365709794379027,
      "loss": 0.1211,
      "step": 1448
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001364935311121244,
      "loss": 0.0726,
      "step": 1449
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013641605752528224,
      "loss": 0.1732,
      "step": 1450
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001363385587310039,
      "loss": 0.1335,
      "step": 1451
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013626103478293469,
      "loss": 0.1469,
      "step": 1452
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013618348573473705,
      "loss": 0.2504,
      "step": 1453
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013610591164009103,
      "loss": 0.0891,
      "step": 1454
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001360283125526939,
      "loss": 0.1633,
      "step": 1455
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013595068852626028,
      "loss": 0.1743,
      "step": 1456
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013587303961452204,
      "loss": 0.2268,
      "step": 1457
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013579536587122828,
      "loss": 0.1837,
      "step": 1458
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013571766735014524,
      "loss": 0.1251,
      "step": 1459
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013563994410505635,
      "loss": 0.2054,
      "step": 1460
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013556219618976218,
      "loss": 0.1854,
      "step": 1461
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013548442365808036,
      "loss": 0.1198,
      "step": 1462
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013540662656384556,
      "loss": 0.2281,
      "step": 1463
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001353288049609094,
      "loss": 0.2934,
      "step": 1464
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001352509589031405,
      "loss": 0.2515,
      "step": 1465
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001351730884444245,
      "loss": 0.2408,
      "step": 1466
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00013509519363866379,
      "loss": 0.1965,
      "step": 1467
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013501727453977763,
      "loss": 0.1569,
      "step": 1468
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013493933120170223,
      "loss": 0.2497,
      "step": 1469
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001348613636783904,
      "loss": 0.1736,
      "step": 1470
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013478337202381185,
      "loss": 0.1352,
      "step": 1471
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013470535629195286,
      "loss": 0.1953,
      "step": 1472
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013462731653681644,
      "loss": 0.224,
      "step": 1473
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013454925281242226,
      "loss": 0.196,
      "step": 1474
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013447116517280652,
      "loss": 0.1986,
      "step": 1475
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013439305367202205,
      "loss": 0.1392,
      "step": 1476
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001343149183641381,
      "loss": 0.082,
      "step": 1477
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001342367593032404,
      "loss": 0.1167,
      "step": 1478
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013415857654343127,
      "loss": 0.1956,
      "step": 1479
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013408037013882927,
      "loss": 0.1579,
      "step": 1480
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013400214014356942,
      "loss": 0.1811,
      "step": 1481
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013392388661180303,
      "loss": 0.0997,
      "step": 1482
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013384560959769774,
      "loss": 0.0995,
      "step": 1483
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013376730915543736,
      "loss": 0.1569,
      "step": 1484
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000133688985339222,
      "loss": 0.141,
      "step": 1485
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00013361063820326797,
      "loss": 0.2677,
      "step": 1486
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013353226780180762,
      "loss": 0.1245,
      "step": 1487
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013345387418908946,
      "loss": 0.2866,
      "step": 1488
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001333754574193781,
      "loss": 0.1417,
      "step": 1489
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013329701754695412,
      "loss": 0.3077,
      "step": 1490
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013321855462611416,
      "loss": 0.2078,
      "step": 1491
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001331400687111707,
      "loss": 0.1619,
      "step": 1492
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013306155985645231,
      "loss": 0.1411,
      "step": 1493
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013298302811630324,
      "loss": 0.071,
      "step": 1494
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013290447354508372,
      "loss": 0.3154,
      "step": 1495
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001328258961971698,
      "loss": 0.0936,
      "step": 1496
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013274729612695315,
      "loss": 0.1365,
      "step": 1497
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001326686733888413,
      "loss": 0.1128,
      "step": 1498
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013259002803725744,
      "loss": 0.1136,
      "step": 1499
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001325113601266404,
      "loss": 0.1207,
      "step": 1500
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001324326697114446,
      "loss": 0.1977,
      "step": 1501
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0001323539568461401,
      "loss": 0.115,
      "step": 1502
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013227522158521245,
      "loss": 0.1634,
      "step": 1503
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013219646398316272,
      "loss": 0.1757,
      "step": 1504
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00013211768409450742,
      "loss": 0.1698,
      "step": 1505
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013203888197377855,
      "loss": 0.148,
      "step": 1506
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013196005767552346,
      "loss": 0.1331,
      "step": 1507
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013188121125430483,
      "loss": 0.1436,
      "step": 1508
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001318023427647007,
      "loss": 0.1761,
      "step": 1509
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013172345226130432,
      "loss": 0.1878,
      "step": 1510
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013164453979872424,
      "loss": 0.248,
      "step": 1511
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013156560543158417,
      "loss": 0.1853,
      "step": 1512
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013148664921452306,
      "loss": 0.0739,
      "step": 1513
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001314076712021949,
      "loss": 0.0854,
      "step": 1514
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013132867144926874,
      "loss": 0.1526,
      "step": 1515
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013124965001042879,
      "loss": 0.3326,
      "step": 1516
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001311706069403742,
      "loss": 0.156,
      "step": 1517
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001310915422938191,
      "loss": 0.2221,
      "step": 1518
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013101245612549256,
      "loss": 0.1303,
      "step": 1519
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013093334849013856,
      "loss": 0.1734,
      "step": 1520
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013085421944251593,
      "loss": 0.1382,
      "step": 1521
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001307750690373983,
      "loss": 0.1389,
      "step": 1522
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013069589732957412,
      "loss": 0.2068,
      "step": 1523
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.00013061670437384656,
      "loss": 0.2016,
      "step": 1524
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00013053749022503354,
      "loss": 0.1887,
      "step": 1525
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0001304582549379676,
      "loss": 0.1931,
      "step": 1526
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0001303789985674959,
      "loss": 0.1007,
      "step": 1527
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00013029972116848025,
      "loss": 0.2514,
      "step": 1528
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00013022042279579697,
      "loss": 0.2173,
      "step": 1529
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.000130141103504337,
      "loss": 0.2476,
      "step": 1530
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00013006176334900556,
      "loss": 0.1618,
      "step": 1531
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00012998240238472252,
      "loss": 0.154,
      "step": 1532
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.000129903020666422,
      "loss": 0.2169,
      "step": 1533
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00012982361824905261,
      "loss": 0.1985,
      "step": 1534
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00012974419518757722,
      "loss": 0.271,
      "step": 1535
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00012966475153697296,
      "loss": 0.1763,
      "step": 1536
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0001295852873522313,
      "loss": 0.1763,
      "step": 1537
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00012950580268835784,
      "loss": 0.1879,
      "step": 1538
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00012942629760037238,
      "loss": 0.2579,
      "step": 1539
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00012934677214330886,
      "loss": 0.1994,
      "step": 1540
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0001292672263722154,
      "loss": 0.1505,
      "step": 1541
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00012918766034215407,
      "loss": 0.1558,
      "step": 1542
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00012910807410820095,
      "loss": 0.2174,
      "step": 1543
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012902846772544624,
      "loss": 0.1237,
      "step": 1544
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012894884124899398,
      "loss": 0.1455,
      "step": 1545
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001288691947339621,
      "loss": 0.1516,
      "step": 1546
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001287895282354825,
      "loss": 0.1453,
      "step": 1547
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012870984180870086,
      "loss": 0.2567,
      "step": 1548
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012863013550877658,
      "loss": 0.1135,
      "step": 1549
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012855040939088296,
      "loss": 0.1534,
      "step": 1550
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012847066351020695,
      "loss": 0.1167,
      "step": 1551
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012839089792194912,
      "loss": 0.1478,
      "step": 1552
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012831111268132378,
      "loss": 0.092,
      "step": 1553
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001282313078435588,
      "loss": 0.0851,
      "step": 1554
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012815148346389564,
      "loss": 0.1036,
      "step": 1555
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012807163959758917,
      "loss": 0.1571,
      "step": 1556
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012799177629990792,
      "loss": 0.2327,
      "step": 1557
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001279118936261338,
      "loss": 0.1845,
      "step": 1558
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001278319916315621,
      "loss": 0.1217,
      "step": 1559
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001277520703715015,
      "loss": 0.1934,
      "step": 1560
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012767212990127404,
      "loss": 0.1503,
      "step": 1561
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.00012759217027621505,
      "loss": 0.1011,
      "step": 1562
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012751219155167313,
      "loss": 0.207,
      "step": 1563
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012743219378301003,
      "loss": 0.1488,
      "step": 1564
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012735217702560076,
      "loss": 0.1159,
      "step": 1565
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012727214133483346,
      "loss": 0.1978,
      "step": 1566
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012719208676610934,
      "loss": 0.198,
      "step": 1567
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012711201337484276,
      "loss": 0.0646,
      "step": 1568
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.000127031921216461,
      "loss": 0.1694,
      "step": 1569
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012695181034640436,
      "loss": 0.0934,
      "step": 1570
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012687168082012616,
      "loss": 0.2773,
      "step": 1571
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001267915326930926,
      "loss": 0.0519,
      "step": 1572
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001267113660207827,
      "loss": 0.1498,
      "step": 1573
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012663118085868836,
      "loss": 0.2086,
      "step": 1574
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012655097726231435,
      "loss": 0.0917,
      "step": 1575
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001264707552871781,
      "loss": 0.1225,
      "step": 1576
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0001263905149888097,
      "loss": 0.1194,
      "step": 1577
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012631025642275212,
      "loss": 0.1074,
      "step": 1578
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012622997964456087,
      "loss": 0.1671,
      "step": 1579
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012614968470980403,
      "loss": 0.167,
      "step": 1580
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00012606937167406227,
      "loss": 0.301,
      "step": 1581
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012598904059292883,
      "loss": 0.1217,
      "step": 1582
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012590869152200943,
      "loss": 0.1397,
      "step": 1583
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0001258283245169222,
      "loss": 0.137,
      "step": 1584
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012574793963329777,
      "loss": 0.144,
      "step": 1585
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.000125667536926779,
      "loss": 0.1195,
      "step": 1586
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012558711645302125,
      "loss": 0.2136,
      "step": 1587
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012550667826769207,
      "loss": 0.1265,
      "step": 1588
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012542622242647133,
      "loss": 0.1397,
      "step": 1589
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012534574898505105,
      "loss": 0.204,
      "step": 1590
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0001252652579991355,
      "loss": 0.1662,
      "step": 1591
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0001251847495244411,
      "loss": 0.1531,
      "step": 1592
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012510422361669633,
      "loss": 0.1859,
      "step": 1593
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012502368033164176,
      "loss": 0.1279,
      "step": 1594
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012494311972502994,
      "loss": 0.1303,
      "step": 1595
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012486254185262554,
      "loss": 0.103,
      "step": 1596
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012478194677020504,
      "loss": 0.1024,
      "step": 1597
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0001247013345335569,
      "loss": 0.2223,
      "step": 1598
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012462070519848146,
      "loss": 0.1945,
      "step": 1599
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.00012454005882079082,
      "loss": 0.1192,
      "step": 1600
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012445939545630902,
      "loss": 0.113,
      "step": 1601
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012437871516087173,
      "loss": 0.3204,
      "step": 1602
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001242980179903264,
      "loss": 0.205,
      "step": 1603
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012421730400053214,
      "loss": 0.1563,
      "step": 1604
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001241365732473597,
      "loss": 0.1735,
      "step": 1605
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012405582578669145,
      "loss": 0.1196,
      "step": 1606
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001239750616744213,
      "loss": 0.1918,
      "step": 1607
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012389428096645474,
      "loss": 0.1386,
      "step": 1608
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012381348371870866,
      "loss": 0.1293,
      "step": 1609
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001237326699871115,
      "loss": 0.0878,
      "step": 1610
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012365183982760304,
      "loss": 0.126,
      "step": 1611
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012357099329613443,
      "loss": 0.153,
      "step": 1612
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012349013044866822,
      "loss": 0.2073,
      "step": 1613
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012340925134117817,
      "loss": 0.1982,
      "step": 1614
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012332835602964933,
      "loss": 0.067,
      "step": 1615
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012324744457007804,
      "loss": 0.1066,
      "step": 1616
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00012316651701847165,
      "loss": 0.1575,
      "step": 1617
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0001230855734308488,
      "loss": 0.0667,
      "step": 1618
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012300461386323915,
      "loss": 0.1103,
      "step": 1619
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012292363837168352,
      "loss": 0.1956,
      "step": 1620
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001228426470122336,
      "loss": 0.1939,
      "step": 1621
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012276163984095218,
      "loss": 0.1429,
      "step": 1622
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012268061691391298,
      "loss": 0.2205,
      "step": 1623
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012259957828720052,
      "loss": 0.0752,
      "step": 1624
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012251852401691038,
      "loss": 0.0957,
      "step": 1625
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012243745415914883,
      "loss": 0.238,
      "step": 1626
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001223563687700329,
      "loss": 0.1446,
      "step": 1627
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012227526790569053,
      "loss": 0.1751,
      "step": 1628
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001221941516222602,
      "loss": 0.1504,
      "step": 1629
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001221130199758912,
      "loss": 0.0831,
      "step": 1630
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012203187302274334,
      "loss": 0.1503,
      "step": 1631
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001219507108189871,
      "loss": 0.0945,
      "step": 1632
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001218695334208035,
      "loss": 0.1743,
      "step": 1633
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012178834088438404,
      "loss": 0.0649,
      "step": 1634
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0001217071332659308,
      "loss": 0.1213,
      "step": 1635
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012162591062165615,
      "loss": 0.0768,
      "step": 1636
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00012154467300778301,
      "loss": 0.1114,
      "step": 1637
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012146342048054453,
      "loss": 0.1131,
      "step": 1638
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012138215309618433,
      "loss": 0.0934,
      "step": 1639
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012130087091095617,
      "loss": 0.1513,
      "step": 1640
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012121957398112411,
      "loss": 0.1658,
      "step": 1641
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012113826236296244,
      "loss": 0.0833,
      "step": 1642
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012105693611275562,
      "loss": 0.1723,
      "step": 1643
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012097559528679821,
      "loss": 0.1107,
      "step": 1644
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012089423994139484,
      "loss": 0.0762,
      "step": 1645
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012081287013286025,
      "loss": 0.1374,
      "step": 1646
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012073148591751915,
      "loss": 0.1672,
      "step": 1647
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012065008735170624,
      "loss": 0.1964,
      "step": 1648
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012056867449176614,
      "loss": 0.2148,
      "step": 1649
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012048724739405338,
      "loss": 0.2328,
      "step": 1650
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012040580611493229,
      "loss": 0.0594,
      "step": 1651
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012032435071077712,
      "loss": 0.1178,
      "step": 1652
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0001202428812379718,
      "loss": 0.1606,
      "step": 1653
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0001201613977529101,
      "loss": 0.2155,
      "step": 1654
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00012007990031199534,
      "loss": 0.0694,
      "step": 1655
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.00011999838897164066,
      "loss": 0.1084,
      "step": 1656
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011991686378826869,
      "loss": 0.1289,
      "step": 1657
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001198353248183118,
      "loss": 0.138,
      "step": 1658
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011975377211821168,
      "loss": 0.1323,
      "step": 1659
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011967220574441973,
      "loss": 0.1091,
      "step": 1660
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011959062575339673,
      "loss": 0.2315,
      "step": 1661
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011950903220161285,
      "loss": 0.2365,
      "step": 1662
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0001194274251455477,
      "loss": 0.2572,
      "step": 1663
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011934580464169025,
      "loss": 0.1066,
      "step": 1664
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011926417074653872,
      "loss": 0.2786,
      "step": 1665
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011918252351660066,
      "loss": 0.2567,
      "step": 1666
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011910086300839282,
      "loss": 0.1806,
      "step": 1667
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011901918927844114,
      "loss": 0.137,
      "step": 1668
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011893750238328072,
      "loss": 0.1901,
      "step": 1669
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011885580237945572,
      "loss": 0.06,
      "step": 1670
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011877408932351953,
      "loss": 0.153,
      "step": 1671
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011869236327203434,
      "loss": 0.205,
      "step": 1672
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011861062428157156,
      "loss": 0.159,
      "step": 1673
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011852887240871145,
      "loss": 0.1489,
      "step": 1674
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00011844710771004315,
      "loss": 0.1525,
      "step": 1675
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011836533024216476,
      "loss": 0.1809,
      "step": 1676
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011828354006168318,
      "loss": 0.0825,
      "step": 1677
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011820173722521413,
      "loss": 0.2067,
      "step": 1678
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011811992178938203,
      "loss": 0.1473,
      "step": 1679
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001180380938108201,
      "loss": 0.1131,
      "step": 1680
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001179562533461702,
      "loss": 0.0838,
      "step": 1681
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011787440045208286,
      "loss": 0.1138,
      "step": 1682
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011779253518521716,
      "loss": 0.1128,
      "step": 1683
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001177106576022408,
      "loss": 0.1019,
      "step": 1684
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011762876775983,
      "loss": 0.1055,
      "step": 1685
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011754686571466944,
      "loss": 0.2101,
      "step": 1686
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011746495152345223,
      "loss": 0.2123,
      "step": 1687
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011738302524287997,
      "loss": 0.137,
      "step": 1688
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0001173010869296625,
      "loss": 0.1267,
      "step": 1689
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011721913664051813,
      "loss": 0.1773,
      "step": 1690
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011713717443217336,
      "loss": 0.1089,
      "step": 1691
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.000117055200361363,
      "loss": 0.1546,
      "step": 1692
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011697321448482998,
      "loss": 0.189,
      "step": 1693
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00011689121685932556,
      "loss": 0.1515,
      "step": 1694
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.000116809207541609,
      "loss": 0.0684,
      "step": 1695
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011672718658844765,
      "loss": 0.1203,
      "step": 1696
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011664515405661698,
      "loss": 0.0797,
      "step": 1697
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0001165631100029005,
      "loss": 0.2115,
      "step": 1698
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011648105448408962,
      "loss": 0.1844,
      "step": 1699
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011639898755698366,
      "loss": 0.0655,
      "step": 1700
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011631690927838994,
      "loss": 0.2228,
      "step": 1701
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0001162348197051236,
      "loss": 0.1263,
      "step": 1702
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011615271889400758,
      "loss": 0.1677,
      "step": 1703
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011607060690187257,
      "loss": 0.1644,
      "step": 1704
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011598848378555707,
      "loss": 0.1485,
      "step": 1705
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011590634960190721,
      "loss": 0.1843,
      "step": 1706
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011582420440777686,
      "loss": 0.1434,
      "step": 1707
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011574204826002743,
      "loss": 0.0558,
      "step": 1708
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.000115659881215528,
      "loss": 0.1713,
      "step": 1709
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011557770333115506,
      "loss": 0.1873,
      "step": 1710
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011549551466379276,
      "loss": 0.1318,
      "step": 1711
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011541331527033263,
      "loss": 0.1211,
      "step": 1712
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.00011533110520767358,
      "loss": 0.2328,
      "step": 1713
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.000115248884532722,
      "loss": 0.1375,
      "step": 1714
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001151666533023916,
      "loss": 0.1278,
      "step": 1715
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011508441157360336,
      "loss": 0.2127,
      "step": 1716
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011500215940328556,
      "loss": 0.2144,
      "step": 1717
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011491989684837366,
      "loss": 0.1608,
      "step": 1718
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011483762396581044,
      "loss": 0.293,
      "step": 1719
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011475534081254562,
      "loss": 0.1608,
      "step": 1720
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011467304744553618,
      "loss": 0.1171,
      "step": 1721
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011459074392174618,
      "loss": 0.1557,
      "step": 1722
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001145084302981466,
      "loss": 0.1111,
      "step": 1723
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011442610663171553,
      "loss": 0.1082,
      "step": 1724
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011434377297943793,
      "loss": 0.1116,
      "step": 1725
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011426142939830571,
      "loss": 0.1052,
      "step": 1726
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001141790759453176,
      "loss": 0.113,
      "step": 1727
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011409671267747927,
      "loss": 0.1554,
      "step": 1728
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001140143396518031,
      "loss": 0.0897,
      "step": 1729
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011393195692530823,
      "loss": 0.2029,
      "step": 1730
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00011384956455502049,
      "loss": 0.1545,
      "step": 1731
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0001137671625979725,
      "loss": 0.179,
      "step": 1732
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011368475111120343,
      "loss": 0.0812,
      "step": 1733
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011360233015175899,
      "loss": 0.166,
      "step": 1734
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011351989977669156,
      "loss": 0.1095,
      "step": 1735
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011343746004306002,
      "loss": 0.2102,
      "step": 1736
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011335501100792962,
      "loss": 0.0615,
      "step": 1737
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011327255272837221,
      "loss": 0.1674,
      "step": 1738
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011319008526146593,
      "loss": 0.1273,
      "step": 1739
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011310760866429529,
      "loss": 0.12,
      "step": 1740
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011302512299395112,
      "loss": 0.1327,
      "step": 1741
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011294262830753057,
      "loss": 0.1251,
      "step": 1742
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011286012466213705,
      "loss": 0.1302,
      "step": 1743
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011277761211488007,
      "loss": 0.1249,
      "step": 1744
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011269509072287538,
      "loss": 0.1082,
      "step": 1745
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011261256054324488,
      "loss": 0.1964,
      "step": 1746
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011253002163311641,
      "loss": 0.157,
      "step": 1747
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011244747404962401,
      "loss": 0.2006,
      "step": 1748
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011236491784990767,
      "loss": 0.1527,
      "step": 1749
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011228235309111332,
      "loss": 0.1412,
      "step": 1750
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.00011219977983039285,
      "loss": 0.0657,
      "step": 1751
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011211719812490397,
      "loss": 0.0959,
      "step": 1752
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011203460803181038,
      "loss": 0.0869,
      "step": 1753
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011195200960828139,
      "loss": 0.0696,
      "step": 1754
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0001118694029114922,
      "loss": 0.1908,
      "step": 1755
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011178678799862377,
      "loss": 0.2232,
      "step": 1756
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011170416492686263,
      "loss": 0.216,
      "step": 1757
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011162153375340101,
      "loss": 0.1204,
      "step": 1758
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011153889453543676,
      "loss": 0.1785,
      "step": 1759
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011145624733017338,
      "loss": 0.2871,
      "step": 1760
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011137359219481969,
      "loss": 0.2227,
      "step": 1761
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011129092918659019,
      "loss": 0.1612,
      "step": 1762
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011120825836270476,
      "loss": 0.1214,
      "step": 1763
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011112557978038866,
      "loss": 0.0823,
      "step": 1764
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0001110428934968726,
      "loss": 0.2006,
      "step": 1765
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011096019956939254,
      "loss": 0.0933,
      "step": 1766
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011087749805518977,
      "loss": 0.1333,
      "step": 1767
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011079478901151081,
      "loss": 0.1845,
      "step": 1768
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011071207249560742,
      "loss": 0.1276,
      "step": 1769
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.00011062934856473655,
      "loss": 0.0901,
      "step": 1770
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00011054661727616022,
      "loss": 0.08,
      "step": 1771
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00011046387868714556,
      "loss": 0.2028,
      "step": 1772
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00011038113285496481,
      "loss": 0.1855,
      "step": 1773
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00011029837983689517,
      "loss": 0.1542,
      "step": 1774
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00011021561969021879,
      "loss": 0.1194,
      "step": 1775
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00011013285247222283,
      "loss": 0.0821,
      "step": 1776
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00011005007824019933,
      "loss": 0.1163,
      "step": 1777
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001099672970514451,
      "loss": 0.1966,
      "step": 1778
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00010988450896326186,
      "loss": 0.1742,
      "step": 1779
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001098017140329561,
      "loss": 0.121,
      "step": 1780
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00010971891231783893,
      "loss": 0.1411,
      "step": 1781
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00010963610387522631,
      "loss": 0.1025,
      "step": 1782
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00010955328876243885,
      "loss": 0.1493,
      "step": 1783
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00010947046703680161,
      "loss": 0.1968,
      "step": 1784
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00010938763875564437,
      "loss": 0.1487,
      "step": 1785
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00010930480397630145,
      "loss": 0.0742,
      "step": 1786
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00010922196275611161,
      "loss": 0.1701,
      "step": 1787
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0001091391151524181,
      "loss": 0.1315,
      "step": 1788
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00010905626122256855,
      "loss": 0.072,
      "step": 1789
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010897340102391505,
      "loss": 0.1369,
      "step": 1790
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010889053461381391,
      "loss": 0.1499,
      "step": 1791
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010880766204962585,
      "loss": 0.3475,
      "step": 1792
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0001087247833887158,
      "loss": 0.0397,
      "step": 1793
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010864189868845297,
      "loss": 0.1144,
      "step": 1794
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010855900800621062,
      "loss": 0.1164,
      "step": 1795
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010847611139936622,
      "loss": 0.1561,
      "step": 1796
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0001083932089253014,
      "loss": 0.182,
      "step": 1797
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010831030064140184,
      "loss": 0.1303,
      "step": 1798
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010822738660505714,
      "loss": 0.2402,
      "step": 1799
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0001081444668736609,
      "loss": 0.2039,
      "step": 1800
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010806154150461083,
      "loss": 0.1847,
      "step": 1801
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010797861055530831,
      "loss": 0.1169,
      "step": 1802
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010789567408315875,
      "loss": 0.1602,
      "step": 1803
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010781273214557133,
      "loss": 0.149,
      "step": 1804
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010772978479995899,
      "loss": 0.179,
      "step": 1805
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0001076468321037384,
      "loss": 0.1093,
      "step": 1806
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.00010756387411433,
      "loss": 0.143,
      "step": 1807
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0001074809108891579,
      "loss": 0.2171,
      "step": 1808
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010739794248564972,
      "loss": 0.15,
      "step": 1809
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010731496896123676,
      "loss": 0.0763,
      "step": 1810
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001072319903733539,
      "loss": 0.1037,
      "step": 1811
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010714900677943938,
      "loss": 0.0848,
      "step": 1812
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010706601823693503,
      "loss": 0.126,
      "step": 1813
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010698302480328605,
      "loss": 0.0759,
      "step": 1814
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010690002653594114,
      "loss": 0.1586,
      "step": 1815
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010681702349235212,
      "loss": 0.1472,
      "step": 1816
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001067340157299743,
      "loss": 0.1111,
      "step": 1817
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010665100330626625,
      "loss": 0.1259,
      "step": 1818
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010656798627868963,
      "loss": 0.2843,
      "step": 1819
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001064849647047094,
      "loss": 0.1387,
      "step": 1820
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010640193864179373,
      "loss": 0.1067,
      "step": 1821
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010631890814741372,
      "loss": 0.1479,
      "step": 1822
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010623587327904363,
      "loss": 0.1789,
      "step": 1823
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010615283409416077,
      "loss": 0.0879,
      "step": 1824
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010606979065024547,
      "loss": 0.1307,
      "step": 1825
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010598674300478084,
      "loss": 0.0865,
      "step": 1826
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00010590369121525305,
      "loss": 0.1207,
      "step": 1827
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010582063533915114,
      "loss": 0.1487,
      "step": 1828
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010573757543396692,
      "loss": 0.0681,
      "step": 1829
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010565451155719496,
      "loss": 0.1921,
      "step": 1830
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010557144376633268,
      "loss": 0.2491,
      "step": 1831
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010548837211888015,
      "loss": 0.1251,
      "step": 1832
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0001054052966723401,
      "loss": 0.2788,
      "step": 1833
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010532221748421787,
      "loss": 0.0796,
      "step": 1834
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010523913461202148,
      "loss": 0.1187,
      "step": 1835
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010515604811326143,
      "loss": 0.1425,
      "step": 1836
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010507295804545074,
      "loss": 0.1588,
      "step": 1837
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010498986446610488,
      "loss": 0.1257,
      "step": 1838
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010490676743274181,
      "loss": 0.1565,
      "step": 1839
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010482366700288184,
      "loss": 0.09,
      "step": 1840
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010474056323404762,
      "loss": 0.202,
      "step": 1841
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010465745618376417,
      "loss": 0.0654,
      "step": 1842
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010457434590955868,
      "loss": 0.128,
      "step": 1843
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010449123246896067,
      "loss": 0.0515,
      "step": 1844
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010440811591950182,
      "loss": 0.0764,
      "step": 1845
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00010432499631871595,
      "loss": 0.0944,
      "step": 1846
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010424187372413896,
      "loss": 0.0883,
      "step": 1847
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0001041587481933089,
      "loss": 0.1375,
      "step": 1848
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010407561978376584,
      "loss": 0.1641,
      "step": 1849
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010399248855305176,
      "loss": 0.099,
      "step": 1850
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010390935455871068,
      "loss": 0.1836,
      "step": 1851
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010382621785828848,
      "loss": 0.2102,
      "step": 1852
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010374307850933293,
      "loss": 0.1791,
      "step": 1853
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010365993656939367,
      "loss": 0.1498,
      "step": 1854
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010357679209602207,
      "loss": 0.0655,
      "step": 1855
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010349364514677131,
      "loss": 0.1723,
      "step": 1856
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010341049577919624,
      "loss": 0.2157,
      "step": 1857
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010332734405085342,
      "loss": 0.0909,
      "step": 1858
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010324419001930104,
      "loss": 0.106,
      "step": 1859
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0001031610337420988,
      "loss": 0.0797,
      "step": 1860
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0001030778752768081,
      "loss": 0.1445,
      "step": 1861
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010299471468099175,
      "loss": 0.0834,
      "step": 1862
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010291155201221407,
      "loss": 0.0835,
      "step": 1863
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010282838732804081,
      "loss": 0.0766,
      "step": 1864
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.00010274522068603911,
      "loss": 0.1143,
      "step": 1865
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010266205214377748,
      "loss": 0.1251,
      "step": 1866
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010257888175882575,
      "loss": 0.0818,
      "step": 1867
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010249570958875496,
      "loss": 0.1258,
      "step": 1868
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001024125356911375,
      "loss": 0.2629,
      "step": 1869
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010232936012354682,
      "loss": 0.1288,
      "step": 1870
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010224618294355768,
      "loss": 0.1566,
      "step": 1871
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010216300420874578,
      "loss": 0.1149,
      "step": 1872
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010207982397668809,
      "loss": 0.1949,
      "step": 1873
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010199664230496246,
      "loss": 0.1378,
      "step": 1874
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010191345925114781,
      "loss": 0.1126,
      "step": 1875
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010183027487282401,
      "loss": 0.1277,
      "step": 1876
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010174708922757181,
      "loss": 0.1663,
      "step": 1877
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010166390237297287,
      "loss": 0.0978,
      "step": 1878
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010158071436660972,
      "loss": 0.1918,
      "step": 1879
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0001014975252660656,
      "loss": 0.1256,
      "step": 1880
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010141433512892457,
      "loss": 0.1501,
      "step": 1881
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010133114401277139,
      "loss": 0.1142,
      "step": 1882
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010124795197519152,
      "loss": 0.2173,
      "step": 1883
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00010116475907377105,
      "loss": 0.0916,
      "step": 1884
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00010108156536609663,
      "loss": 0.1124,
      "step": 1885
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001009983709097555,
      "loss": 0.1679,
      "step": 1886
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001009151757623354,
      "loss": 0.1314,
      "step": 1887
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00010083197998142459,
      "loss": 0.0787,
      "step": 1888
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00010074878362461172,
      "loss": 0.0675,
      "step": 1889
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001006655867494859,
      "loss": 0.0938,
      "step": 1890
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00010058238941363646,
      "loss": 0.1692,
      "step": 1891
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00010049919167465327,
      "loss": 0.2454,
      "step": 1892
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00010041599359012628,
      "loss": 0.1315,
      "step": 1893
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00010033279521764578,
      "loss": 0.1474,
      "step": 1894
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001002495966148022,
      "loss": 0.0707,
      "step": 1895
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00010016639783918623,
      "loss": 0.1139,
      "step": 1896
    },
    {
      "epoch": 1.0,
      "eval_loss": 0.12808886170387268,
      "eval_runtime": 682.6741,
      "eval_samples_per_second": 9.825,
      "eval_steps_per_second": 2.457,
      "step": 1896
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00010008319894838856,
      "loss": 0.1809,
      "step": 1897
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0001,
      "loss": 0.1222,
      "step": 1898
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.991680105161146e-05,
      "loss": 0.0572,
      "step": 1899
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.983360216081378e-05,
      "loss": 0.0971,
      "step": 1900
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.975040338519779e-05,
      "loss": 0.2005,
      "step": 1901
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.966720478235422e-05,
      "loss": 0.1351,
      "step": 1902
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.958400640987374e-05,
      "loss": 0.1722,
      "step": 1903
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.950080832534677e-05,
      "loss": 0.1051,
      "step": 1904
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.941761058636355e-05,
      "loss": 0.1471,
      "step": 1905
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.933441325051414e-05,
      "loss": 0.1202,
      "step": 1906
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.92512163753883e-05,
      "loss": 0.2466,
      "step": 1907
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.916802001857543e-05,
      "loss": 0.1149,
      "step": 1908
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.90848242376646e-05,
      "loss": 0.1157,
      "step": 1909
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.900162909024452e-05,
      "loss": 0.1439,
      "step": 1910
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.891843463390343e-05,
      "loss": 0.2316,
      "step": 1911
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.883524092622898e-05,
      "loss": 0.0501,
      "step": 1912
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.875204802480849e-05,
      "loss": 0.1713,
      "step": 1913
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.866885598722863e-05,
      "loss": 0.1733,
      "step": 1914
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.858566487107545e-05,
      "loss": 0.1128,
      "step": 1915
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.850247473393443e-05,
      "loss": 0.1768,
      "step": 1916
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.84192856333903e-05,
      "loss": 0.131,
      "step": 1917
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.833609762702714e-05,
      "loss": 0.1679,
      "step": 1918
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.825291077242819e-05,
      "loss": 0.1538,
      "step": 1919
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.816972512717603e-05,
      "loss": 0.0984,
      "step": 1920
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.808654074885221e-05,
      "loss": 0.1314,
      "step": 1921
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.800335769503756e-05,
      "loss": 0.2006,
      "step": 1922
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.792017602331192e-05,
      "loss": 0.0678,
      "step": 1923
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.783699579125423e-05,
      "loss": 0.1137,
      "step": 1924
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.775381705644236e-05,
      "loss": 0.1066,
      "step": 1925
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.767063987645318e-05,
      "loss": 0.0795,
      "step": 1926
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.758746430886253e-05,
      "loss": 0.0934,
      "step": 1927
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.75042904112451e-05,
      "loss": 0.0678,
      "step": 1928
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.742111824117429e-05,
      "loss": 0.2114,
      "step": 1929
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.733794785622253e-05,
      "loss": 0.0559,
      "step": 1930
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.725477931396093e-05,
      "loss": 0.1194,
      "step": 1931
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.717161267195921e-05,
      "loss": 0.0953,
      "step": 1932
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.708844798778594e-05,
      "loss": 0.1606,
      "step": 1933
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.700528531900827e-05,
      "loss": 0.0657,
      "step": 1934
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.69221247231919e-05,
      "loss": 0.1083,
      "step": 1935
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.683896625790124e-05,
      "loss": 0.0764,
      "step": 1936
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.675580998069901e-05,
      "loss": 0.069,
      "step": 1937
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.667265594914662e-05,
      "loss": 0.0873,
      "step": 1938
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.658950422080377e-05,
      "loss": 0.1144,
      "step": 1939
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.65063548532287e-05,
      "loss": 0.1196,
      "step": 1940
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.642320790397795e-05,
      "loss": 0.106,
      "step": 1941
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.634006343060635e-05,
      "loss": 0.1625,
      "step": 1942
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.625692149066708e-05,
      "loss": 0.1158,
      "step": 1943
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.617378214171155e-05,
      "loss": 0.0688,
      "step": 1944
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.609064544128937e-05,
      "loss": 0.081,
      "step": 1945
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.600751144694827e-05,
      "loss": 0.1708,
      "step": 1946
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.592438021623417e-05,
      "loss": 0.1951,
      "step": 1947
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.584125180669112e-05,
      "loss": 0.1479,
      "step": 1948
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.575812627586105e-05,
      "loss": 0.17,
      "step": 1949
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.567500368128407e-05,
      "loss": 0.1934,
      "step": 1950
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.55918840804982e-05,
      "loss": 0.0873,
      "step": 1951
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.550876753103933e-05,
      "loss": 0.1978,
      "step": 1952
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.542565409044136e-05,
      "loss": 0.1225,
      "step": 1953
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.534254381623587e-05,
      "loss": 0.0817,
      "step": 1954
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.525943676595242e-05,
      "loss": 0.1636,
      "step": 1955
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.517633299711819e-05,
      "loss": 0.225,
      "step": 1956
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.509323256725821e-05,
      "loss": 0.1613,
      "step": 1957
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.501013553389516e-05,
      "loss": 0.2205,
      "step": 1958
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.492704195454929e-05,
      "loss": 0.1358,
      "step": 1959
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.484395188673857e-05,
      "loss": 0.0678,
      "step": 1960
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.476086538797853e-05,
      "loss": 0.1426,
      "step": 1961
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.467778251578217e-05,
      "loss": 0.1294,
      "step": 1962
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.459470332765996e-05,
      "loss": 0.0683,
      "step": 1963
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.451162788111988e-05,
      "loss": 0.0728,
      "step": 1964
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.442855623366735e-05,
      "loss": 0.154,
      "step": 1965
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.434548844280505e-05,
      "loss": 0.1079,
      "step": 1966
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.42624245660331e-05,
      "loss": 0.1589,
      "step": 1967
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.417936466084887e-05,
      "loss": 0.0469,
      "step": 1968
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.409630878474695e-05,
      "loss": 0.0954,
      "step": 1969
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.401325699521921e-05,
      "loss": 0.0792,
      "step": 1970
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.393020934975458e-05,
      "loss": 0.0647,
      "step": 1971
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.384716590583924e-05,
      "loss": 0.0899,
      "step": 1972
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.376412672095638e-05,
      "loss": 0.0895,
      "step": 1973
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.36810918525863e-05,
      "loss": 0.1174,
      "step": 1974
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.359806135820628e-05,
      "loss": 0.0871,
      "step": 1975
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.351503529529058e-05,
      "loss": 0.0639,
      "step": 1976
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.343201372131038e-05,
      "loss": 0.0629,
      "step": 1977
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.334899669373379e-05,
      "loss": 0.1075,
      "step": 1978
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.326598427002574e-05,
      "loss": 0.1258,
      "step": 1979
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.318297650764791e-05,
      "loss": 0.0736,
      "step": 1980
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.30999734640589e-05,
      "loss": 0.0551,
      "step": 1981
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.301697519671395e-05,
      "loss": 0.1548,
      "step": 1982
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.2933981763065e-05,
      "loss": 0.0641,
      "step": 1983
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.285099322056064e-05,
      "loss": 0.1345,
      "step": 1984
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.276800962664614e-05,
      "loss": 0.1183,
      "step": 1985
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.268503103876323e-05,
      "loss": 0.1012,
      "step": 1986
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.260205751435032e-05,
      "loss": 0.1297,
      "step": 1987
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.251908911084213e-05,
      "loss": 0.114,
      "step": 1988
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.243612588567e-05,
      "loss": 0.087,
      "step": 1989
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.235316789626162e-05,
      "loss": 0.118,
      "step": 1990
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.227021520004104e-05,
      "loss": 0.074,
      "step": 1991
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.218726785442869e-05,
      "loss": 0.1377,
      "step": 1992
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.210432591684126e-05,
      "loss": 0.2353,
      "step": 1993
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.202138944469168e-05,
      "loss": 0.0752,
      "step": 1994
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.193845849538921e-05,
      "loss": 0.0863,
      "step": 1995
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.185553312633912e-05,
      "loss": 0.2242,
      "step": 1996
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.177261339494291e-05,
      "loss": 0.0599,
      "step": 1997
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.16896993585982e-05,
      "loss": 0.1688,
      "step": 1998
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.16067910746986e-05,
      "loss": 0.07,
      "step": 1999
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.15238886006338e-05,
      "loss": 0.0465,
      "step": 2000
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.144099199378942e-05,
      "loss": 0.0719,
      "step": 2001
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.135810131154707e-05,
      "loss": 0.0878,
      "step": 2002
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.127521661128419e-05,
      "loss": 0.0258,
      "step": 2003
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.119233795037419e-05,
      "loss": 0.1069,
      "step": 2004
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.110946538618612e-05,
      "loss": 0.0965,
      "step": 2005
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.102659897608499e-05,
      "loss": 0.0761,
      "step": 2006
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.094373877743147e-05,
      "loss": 0.0835,
      "step": 2007
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.086088484758192e-05,
      "loss": 0.1381,
      "step": 2008
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.07780372438884e-05,
      "loss": 0.0933,
      "step": 2009
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.069519602369856e-05,
      "loss": 0.0856,
      "step": 2010
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.061236124435562e-05,
      "loss": 0.0816,
      "step": 2011
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.052953296319842e-05,
      "loss": 0.1544,
      "step": 2012
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.044671123756119e-05,
      "loss": 0.1296,
      "step": 2013
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.03638961247737e-05,
      "loss": 0.0841,
      "step": 2014
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.028108768216109e-05,
      "loss": 0.1728,
      "step": 2015
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.019828596704394e-05,
      "loss": 0.135,
      "step": 2016
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.011549103673817e-05,
      "loss": 0.0896,
      "step": 2017
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.003270294855491e-05,
      "loss": 0.1795,
      "step": 2018
    },
    {
      "epoch": 1.05,
      "learning_rate": 8.99499217598007e-05,
      "loss": 0.0875,
      "step": 2019
    },
    {
      "epoch": 1.05,
      "learning_rate": 8.986714752777716e-05,
      "loss": 0.1272,
      "step": 2020
    },
    {
      "epoch": 1.05,
      "learning_rate": 8.978438030978125e-05,
      "loss": 0.0817,
      "step": 2021
    },
    {
      "epoch": 1.05,
      "learning_rate": 8.970162016310487e-05,
      "loss": 0.0994,
      "step": 2022
    },
    {
      "epoch": 1.05,
      "learning_rate": 8.961886714503522e-05,
      "loss": 0.1456,
      "step": 2023
    },
    {
      "epoch": 1.05,
      "learning_rate": 8.953612131285446e-05,
      "loss": 0.0825,
      "step": 2024
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.94533827238398e-05,
      "loss": 0.0535,
      "step": 2025
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.937065143526347e-05,
      "loss": 0.1308,
      "step": 2026
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.928792750439257e-05,
      "loss": 0.1291,
      "step": 2027
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.920521098848918e-05,
      "loss": 0.1369,
      "step": 2028
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.912250194481027e-05,
      "loss": 0.1073,
      "step": 2029
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.903980043060748e-05,
      "loss": 0.2194,
      "step": 2030
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.895710650312743e-05,
      "loss": 0.0709,
      "step": 2031
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.887442021961135e-05,
      "loss": 0.0484,
      "step": 2032
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.879174163729525e-05,
      "loss": 0.05,
      "step": 2033
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.870907081340982e-05,
      "loss": 0.2039,
      "step": 2034
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.862640780518032e-05,
      "loss": 0.1409,
      "step": 2035
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.854375266982666e-05,
      "loss": 0.1073,
      "step": 2036
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.846110546456324e-05,
      "loss": 0.1062,
      "step": 2037
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.837846624659904e-05,
      "loss": 0.1571,
      "step": 2038
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.829583507313744e-05,
      "loss": 0.0683,
      "step": 2039
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.821321200137627e-05,
      "loss": 0.1132,
      "step": 2040
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.813059708850781e-05,
      "loss": 0.1064,
      "step": 2041
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.804799039171863e-05,
      "loss": 0.0996,
      "step": 2042
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.796539196818964e-05,
      "loss": 0.1281,
      "step": 2043
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.788280187509602e-05,
      "loss": 0.0391,
      "step": 2044
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.780022016960716e-05,
      "loss": 0.1447,
      "step": 2045
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.77176469088867e-05,
      "loss": 0.0686,
      "step": 2046
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.763508215009235e-05,
      "loss": 0.17,
      "step": 2047
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.755252595037602e-05,
      "loss": 0.086,
      "step": 2048
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.746997836688361e-05,
      "loss": 0.1176,
      "step": 2049
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.738743945675515e-05,
      "loss": 0.1142,
      "step": 2050
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.730490927712463e-05,
      "loss": 0.0797,
      "step": 2051
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.722238788511992e-05,
      "loss": 0.1197,
      "step": 2052
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.713987533786294e-05,
      "loss": 0.1267,
      "step": 2053
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.705737169246944e-05,
      "loss": 0.1704,
      "step": 2054
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.697487700604893e-05,
      "loss": 0.1192,
      "step": 2055
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.689239133570476e-05,
      "loss": 0.1859,
      "step": 2056
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.680991473853411e-05,
      "loss": 0.0989,
      "step": 2057
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.672744727162781e-05,
      "loss": 0.0836,
      "step": 2058
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.664498899207039e-05,
      "loss": 0.0665,
      "step": 2059
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.656253995694e-05,
      "loss": 0.1063,
      "step": 2060
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.648010022330845e-05,
      "loss": 0.1096,
      "step": 2061
    },
    {
      "epoch": 1.07,
      "learning_rate": 8.639766984824102e-05,
      "loss": 0.0627,
      "step": 2062
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.631524888879663e-05,
      "loss": 0.074,
      "step": 2063
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.62328374020275e-05,
      "loss": 0.1085,
      "step": 2064
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.615043544497953e-05,
      "loss": 0.1771,
      "step": 2065
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.60680430746918e-05,
      "loss": 0.0701,
      "step": 2066
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.598566034819691e-05,
      "loss": 0.1031,
      "step": 2067
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.590328732252074e-05,
      "loss": 0.043,
      "step": 2068
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.582092405468238e-05,
      "loss": 0.0923,
      "step": 2069
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.57385706016943e-05,
      "loss": 0.0827,
      "step": 2070
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.565622702056209e-05,
      "loss": 0.085,
      "step": 2071
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.55738933682845e-05,
      "loss": 0.0917,
      "step": 2072
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.549156970185341e-05,
      "loss": 0.1286,
      "step": 2073
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.540925607825384e-05,
      "loss": 0.1409,
      "step": 2074
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.532695255446383e-05,
      "loss": 0.0961,
      "step": 2075
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.524465918745442e-05,
      "loss": 0.1812,
      "step": 2076
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.516237603418958e-05,
      "loss": 0.1013,
      "step": 2077
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.508010315162634e-05,
      "loss": 0.0319,
      "step": 2078
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.499784059671445e-05,
      "loss": 0.1332,
      "step": 2079
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.491558842639667e-05,
      "loss": 0.0453,
      "step": 2080
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.483334669760841e-05,
      "loss": 0.0709,
      "step": 2081
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.475111546727802e-05,
      "loss": 0.0543,
      "step": 2082
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.466889479232645e-05,
      "loss": 0.1224,
      "step": 2083
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.45866847296674e-05,
      "loss": 0.1151,
      "step": 2084
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.450448533620725e-05,
      "loss": 0.0717,
      "step": 2085
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.442229666884494e-05,
      "loss": 0.0928,
      "step": 2086
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.434011878447201e-05,
      "loss": 0.1918,
      "step": 2087
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.425795173997259e-05,
      "loss": 0.0999,
      "step": 2088
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.417579559222318e-05,
      "loss": 0.0512,
      "step": 2089
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.409365039809281e-05,
      "loss": 0.1907,
      "step": 2090
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.401151621444296e-05,
      "loss": 0.0639,
      "step": 2091
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.392939309812746e-05,
      "loss": 0.0944,
      "step": 2092
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.384728110599245e-05,
      "loss": 0.1062,
      "step": 2093
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.37651802948764e-05,
      "loss": 0.1074,
      "step": 2094
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.368309072161007e-05,
      "loss": 0.0724,
      "step": 2095
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.360101244301635e-05,
      "loss": 0.091,
      "step": 2096
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.351894551591044e-05,
      "loss": 0.0855,
      "step": 2097
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.343688999709952e-05,
      "loss": 0.1376,
      "step": 2098
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.335484594338303e-05,
      "loss": 0.0891,
      "step": 2099
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.327281341155237e-05,
      "loss": 0.1144,
      "step": 2100
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.319079245839103e-05,
      "loss": 0.0876,
      "step": 2101
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.310878314067445e-05,
      "loss": 0.1128,
      "step": 2102
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.302678551517e-05,
      "loss": 0.1043,
      "step": 2103
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.2944799638637e-05,
      "loss": 0.1437,
      "step": 2104
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.286282556782666e-05,
      "loss": 0.1102,
      "step": 2105
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.27808633594819e-05,
      "loss": 0.0892,
      "step": 2106
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.269891307033752e-05,
      "loss": 0.1559,
      "step": 2107
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.261697475712007e-05,
      "loss": 0.1484,
      "step": 2108
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.253504847654779e-05,
      "loss": 0.1016,
      "step": 2109
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.245313428533058e-05,
      "loss": 0.227,
      "step": 2110
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.237123224017001e-05,
      "loss": 0.0775,
      "step": 2111
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.22893423977592e-05,
      "loss": 0.1187,
      "step": 2112
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.220746481478288e-05,
      "loss": 0.1604,
      "step": 2113
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.212559954791718e-05,
      "loss": 0.2187,
      "step": 2114
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.204374665382982e-05,
      "loss": 0.1124,
      "step": 2115
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.196190618917994e-05,
      "loss": 0.0862,
      "step": 2116
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.188007821061799e-05,
      "loss": 0.0454,
      "step": 2117
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.179826277478589e-05,
      "loss": 0.0513,
      "step": 2118
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.171645993831683e-05,
      "loss": 0.0855,
      "step": 2119
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.163466975783525e-05,
      "loss": 0.0899,
      "step": 2120
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.155289228995684e-05,
      "loss": 0.0801,
      "step": 2121
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.147112759128859e-05,
      "loss": 0.1321,
      "step": 2122
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.138937571842846e-05,
      "loss": 0.1308,
      "step": 2123
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.130763672796568e-05,
      "loss": 0.2348,
      "step": 2124
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.122591067648051e-05,
      "loss": 0.1164,
      "step": 2125
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.11441976205443e-05,
      "loss": 0.0773,
      "step": 2126
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.106249761671932e-05,
      "loss": 0.2037,
      "step": 2127
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.098081072155887e-05,
      "loss": 0.0763,
      "step": 2128
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.089913699160718e-05,
      "loss": 0.0564,
      "step": 2129
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.081747648339938e-05,
      "loss": 0.0364,
      "step": 2130
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.07358292534613e-05,
      "loss": 0.1078,
      "step": 2131
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.065419535830979e-05,
      "loss": 0.0973,
      "step": 2132
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.057257485445233e-05,
      "loss": 0.0884,
      "step": 2133
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.049096779838719e-05,
      "loss": 0.1564,
      "step": 2134
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.04093742466033e-05,
      "loss": 0.113,
      "step": 2135
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.032779425558028e-05,
      "loss": 0.0758,
      "step": 2136
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.024622788178833e-05,
      "loss": 0.0981,
      "step": 2137
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.016467518168821e-05,
      "loss": 0.0909,
      "step": 2138
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.008313621173132e-05,
      "loss": 0.1614,
      "step": 2139
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.000161102835937e-05,
      "loss": 0.0602,
      "step": 2140
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.992009968800468e-05,
      "loss": 0.1871,
      "step": 2141
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.983860224708993e-05,
      "loss": 0.1055,
      "step": 2142
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.975711876202822e-05,
      "loss": 0.091,
      "step": 2143
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.96756492892229e-05,
      "loss": 0.1165,
      "step": 2144
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.959419388506772e-05,
      "loss": 0.058,
      "step": 2145
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.951275260594666e-05,
      "loss": 0.0769,
      "step": 2146
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.943132550823391e-05,
      "loss": 0.0617,
      "step": 2147
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.934991264829379e-05,
      "loss": 0.2042,
      "step": 2148
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.926851408248088e-05,
      "loss": 0.0478,
      "step": 2149
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.918712986713979e-05,
      "loss": 0.2107,
      "step": 2150
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.910576005860519e-05,
      "loss": 0.1499,
      "step": 2151
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.90244047132018e-05,
      "loss": 0.1233,
      "step": 2152
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.894306388724439e-05,
      "loss": 0.087,
      "step": 2153
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.886173763703757e-05,
      "loss": 0.1241,
      "step": 2154
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.87804260188759e-05,
      "loss": 0.0735,
      "step": 2155
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.869912908904387e-05,
      "loss": 0.098,
      "step": 2156
    },
    {
      "epoch": 1.12,
      "learning_rate": 7.861784690381572e-05,
      "loss": 0.076,
      "step": 2157
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.853657951945548e-05,
      "loss": 0.1379,
      "step": 2158
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.8455326992217e-05,
      "loss": 0.0777,
      "step": 2159
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.837408937834387e-05,
      "loss": 0.0793,
      "step": 2160
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.829286673406923e-05,
      "loss": 0.0617,
      "step": 2161
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.821165911561596e-05,
      "loss": 0.1272,
      "step": 2162
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.813046657919652e-05,
      "loss": 0.1318,
      "step": 2163
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.804928918101295e-05,
      "loss": 0.1157,
      "step": 2164
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.79681269772567e-05,
      "loss": 0.0846,
      "step": 2165
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.788698002410884e-05,
      "loss": 0.109,
      "step": 2166
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.780584837773982e-05,
      "loss": 0.0796,
      "step": 2167
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.77247320943095e-05,
      "loss": 0.0634,
      "step": 2168
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.76436312299671e-05,
      "loss": 0.1676,
      "step": 2169
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.756254584085121e-05,
      "loss": 0.0631,
      "step": 2170
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.748147598308963e-05,
      "loss": 0.1638,
      "step": 2171
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.740042171279951e-05,
      "loss": 0.1178,
      "step": 2172
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.731938308608707e-05,
      "loss": 0.092,
      "step": 2173
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.723836015904786e-05,
      "loss": 0.0814,
      "step": 2174
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.715735298776642e-05,
      "loss": 0.1385,
      "step": 2175
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.707636162831649e-05,
      "loss": 0.1094,
      "step": 2176
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.699538613676085e-05,
      "loss": 0.0726,
      "step": 2177
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.69144265691512e-05,
      "loss": 0.1075,
      "step": 2178
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.683348298152836e-05,
      "loss": 0.0747,
      "step": 2179
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.675255542992199e-05,
      "loss": 0.1239,
      "step": 2180
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.667164397035069e-05,
      "loss": 0.0755,
      "step": 2181
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.659074865882186e-05,
      "loss": 0.0796,
      "step": 2182
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.650986955133181e-05,
      "loss": 0.1207,
      "step": 2183
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.642900670386561e-05,
      "loss": 0.1857,
      "step": 2184
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.634816017239699e-05,
      "loss": 0.0544,
      "step": 2185
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.626733001288851e-05,
      "loss": 0.0468,
      "step": 2186
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.618651628129136e-05,
      "loss": 0.1957,
      "step": 2187
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.610571903354529e-05,
      "loss": 0.3091,
      "step": 2188
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.602493832557875e-05,
      "loss": 0.1213,
      "step": 2189
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.594417421330859e-05,
      "loss": 0.1419,
      "step": 2190
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.586342675264035e-05,
      "loss": 0.0645,
      "step": 2191
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.57826959994679e-05,
      "loss": 0.0609,
      "step": 2192
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.570198200967362e-05,
      "loss": 0.0728,
      "step": 2193
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.562128483912829e-05,
      "loss": 0.0655,
      "step": 2194
    },
    {
      "epoch": 1.14,
      "learning_rate": 7.554060454369098e-05,
      "loss": 0.1095,
      "step": 2195
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.545994117920916e-05,
      "loss": 0.0474,
      "step": 2196
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.537929480151858e-05,
      "loss": 0.1487,
      "step": 2197
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.529866546644315e-05,
      "loss": 0.0932,
      "step": 2198
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.5218053229795e-05,
      "loss": 0.1389,
      "step": 2199
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.513745814737449e-05,
      "loss": 0.0745,
      "step": 2200
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.505688027497007e-05,
      "loss": 0.1044,
      "step": 2201
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.497631966835828e-05,
      "loss": 0.1111,
      "step": 2202
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.489577638330367e-05,
      "loss": 0.076,
      "step": 2203
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.481525047555891e-05,
      "loss": 0.0598,
      "step": 2204
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.47347420008645e-05,
      "loss": 0.1294,
      "step": 2205
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.465425101494899e-05,
      "loss": 0.0824,
      "step": 2206
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.45737775735287e-05,
      "loss": 0.0947,
      "step": 2207
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.449332173230796e-05,
      "loss": 0.0473,
      "step": 2208
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.441288354697876e-05,
      "loss": 0.0753,
      "step": 2209
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.4332463073221e-05,
      "loss": 0.1135,
      "step": 2210
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.425206036670224e-05,
      "loss": 0.1358,
      "step": 2211
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.417167548307779e-05,
      "loss": 0.0677,
      "step": 2212
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.409130847799055e-05,
      "loss": 0.0554,
      "step": 2213
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.401095940707117e-05,
      "loss": 0.1009,
      "step": 2214
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.393062832593778e-05,
      "loss": 0.0755,
      "step": 2215
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.385031529019602e-05,
      "loss": 0.1156,
      "step": 2216
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.377002035543916e-05,
      "loss": 0.0643,
      "step": 2217
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.368974357724789e-05,
      "loss": 0.1794,
      "step": 2218
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.360948501119031e-05,
      "loss": 0.128,
      "step": 2219
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.352924471282194e-05,
      "loss": 0.0759,
      "step": 2220
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.344902273768566e-05,
      "loss": 0.0749,
      "step": 2221
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.336881914131162e-05,
      "loss": 0.1086,
      "step": 2222
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.328863397921734e-05,
      "loss": 0.165,
      "step": 2223
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.320846730690744e-05,
      "loss": 0.1033,
      "step": 2224
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.312831917987388e-05,
      "loss": 0.0631,
      "step": 2225
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.304818965359567e-05,
      "loss": 0.1254,
      "step": 2226
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.296807878353904e-05,
      "loss": 0.1134,
      "step": 2227
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.288798662515727e-05,
      "loss": 0.039,
      "step": 2228
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.280791323389066e-05,
      "loss": 0.0754,
      "step": 2229
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.272785866516654e-05,
      "loss": 0.121,
      "step": 2230
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.264782297439926e-05,
      "loss": 0.1325,
      "step": 2231
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.256780621699e-05,
      "loss": 0.0727,
      "step": 2232
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.248780844832692e-05,
      "loss": 0.0874,
      "step": 2233
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.240782972378496e-05,
      "loss": 0.0509,
      "step": 2234
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.2327870098726e-05,
      "loss": 0.2142,
      "step": 2235
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.224792962849852e-05,
      "loss": 0.2235,
      "step": 2236
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.216800836843792e-05,
      "loss": 0.0944,
      "step": 2237
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.208810637386622e-05,
      "loss": 0.1484,
      "step": 2238
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.200822370009208e-05,
      "loss": 0.0793,
      "step": 2239
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.192836040241088e-05,
      "loss": 0.1619,
      "step": 2240
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.184851653610442e-05,
      "loss": 0.1602,
      "step": 2241
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.176869215644122e-05,
      "loss": 0.0604,
      "step": 2242
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.168888731867624e-05,
      "loss": 0.1189,
      "step": 2243
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.160910207805088e-05,
      "loss": 0.0334,
      "step": 2244
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.152933648979307e-05,
      "loss": 0.0496,
      "step": 2245
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.144959060911703e-05,
      "loss": 0.1736,
      "step": 2246
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.13698644912234e-05,
      "loss": 0.0863,
      "step": 2247
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.129015819129919e-05,
      "loss": 0.1006,
      "step": 2248
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.121047176451751e-05,
      "loss": 0.2065,
      "step": 2249
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.113080526603792e-05,
      "loss": 0.0953,
      "step": 2250
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.105115875100606e-05,
      "loss": 0.0894,
      "step": 2251
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.097153227455379e-05,
      "loss": 0.0769,
      "step": 2252
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.089192589179904e-05,
      "loss": 0.1112,
      "step": 2253
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.081233965784596e-05,
      "loss": 0.0999,
      "step": 2254
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.073277362778461e-05,
      "loss": 0.0397,
      "step": 2255
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.065322785669113e-05,
      "loss": 0.0917,
      "step": 2256
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.057370239962768e-05,
      "loss": 0.0755,
      "step": 2257
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.049419731164221e-05,
      "loss": 0.096,
      "step": 2258
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.041471264776874e-05,
      "loss": 0.1227,
      "step": 2259
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.033524846302706e-05,
      "loss": 0.1354,
      "step": 2260
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.025580481242278e-05,
      "loss": 0.107,
      "step": 2261
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.017638175094738e-05,
      "loss": 0.0616,
      "step": 2262
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.009697933357799e-05,
      "loss": 0.1436,
      "step": 2263
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.001759761527749e-05,
      "loss": 0.1332,
      "step": 2264
    },
    {
      "epoch": 1.18,
      "learning_rate": 6.993823665099446e-05,
      "loss": 0.0706,
      "step": 2265
    },
    {
      "epoch": 1.18,
      "learning_rate": 6.985889649566305e-05,
      "loss": 0.1238,
      "step": 2266
    },
    {
      "epoch": 1.18,
      "learning_rate": 6.977957720420304e-05,
      "loss": 0.087,
      "step": 2267
    },
    {
      "epoch": 1.18,
      "learning_rate": 6.970027883151978e-05,
      "loss": 0.0886,
      "step": 2268
    },
    {
      "epoch": 1.18,
      "learning_rate": 6.962100143250412e-05,
      "loss": 0.0533,
      "step": 2269
    },
    {
      "epoch": 1.18,
      "learning_rate": 6.954174506203243e-05,
      "loss": 0.1632,
      "step": 2270
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.946250977496647e-05,
      "loss": 0.0909,
      "step": 2271
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.938329562615345e-05,
      "loss": 0.0801,
      "step": 2272
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.930410267042588e-05,
      "loss": 0.1517,
      "step": 2273
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.922493096260174e-05,
      "loss": 0.0598,
      "step": 2274
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.91457805574841e-05,
      "loss": 0.1103,
      "step": 2275
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.906665150986145e-05,
      "loss": 0.1064,
      "step": 2276
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.898754387450745e-05,
      "loss": 0.0581,
      "step": 2277
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.890845770618091e-05,
      "loss": 0.0353,
      "step": 2278
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.88293930596258e-05,
      "loss": 0.0774,
      "step": 2279
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.875034998957121e-05,
      "loss": 0.105,
      "step": 2280
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.867132855073125e-05,
      "loss": 0.1468,
      "step": 2281
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.859232879780515e-05,
      "loss": 0.1467,
      "step": 2282
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.851335078547695e-05,
      "loss": 0.1356,
      "step": 2283
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.843439456841583e-05,
      "loss": 0.1113,
      "step": 2284
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.835546020127577e-05,
      "loss": 0.1265,
      "step": 2285
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.827654773869569e-05,
      "loss": 0.2106,
      "step": 2286
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.819765723529933e-05,
      "loss": 0.1681,
      "step": 2287
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.811878874569517e-05,
      "loss": 0.118,
      "step": 2288
    },
    {
      "epoch": 1.19,
      "learning_rate": 6.803994232447652e-05,
      "loss": 0.0609,
      "step": 2289
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.796111802622147e-05,
      "loss": 0.0638,
      "step": 2290
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.788231590549261e-05,
      "loss": 0.0321,
      "step": 2291
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.780353601683734e-05,
      "loss": 0.1388,
      "step": 2292
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.772477841478759e-05,
      "loss": 0.0638,
      "step": 2293
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.764604315385994e-05,
      "loss": 0.1644,
      "step": 2294
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.756733028855542e-05,
      "loss": 0.0567,
      "step": 2295
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.748863987335962e-05,
      "loss": 0.019,
      "step": 2296
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.740997196274257e-05,
      "loss": 0.0474,
      "step": 2297
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.73313266111587e-05,
      "loss": 0.0956,
      "step": 2298
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.725270387304687e-05,
      "loss": 0.0844,
      "step": 2299
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.717410380283023e-05,
      "loss": 0.1083,
      "step": 2300
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.709552645491628e-05,
      "loss": 0.048,
      "step": 2301
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.701697188369678e-05,
      "loss": 0.0696,
      "step": 2302
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.693844014354771e-05,
      "loss": 0.1292,
      "step": 2303
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.68599312888293e-05,
      "loss": 0.0697,
      "step": 2304
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.678144537388587e-05,
      "loss": 0.0868,
      "step": 2305
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.670298245304586e-05,
      "loss": 0.1185,
      "step": 2306
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.662454258062193e-05,
      "loss": 0.1026,
      "step": 2307
    },
    {
      "epoch": 1.2,
      "learning_rate": 6.654612581091058e-05,
      "loss": 0.1302,
      "step": 2308
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.646773219819242e-05,
      "loss": 0.0964,
      "step": 2309
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.638936179673205e-05,
      "loss": 0.0765,
      "step": 2310
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.6311014660778e-05,
      "loss": 0.1258,
      "step": 2311
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.623269084456265e-05,
      "loss": 0.0331,
      "step": 2312
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.615439040230227e-05,
      "loss": 0.1042,
      "step": 2313
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.607611338819697e-05,
      "loss": 0.1827,
      "step": 2314
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.599785985643058e-05,
      "loss": 0.0969,
      "step": 2315
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.591962986117076e-05,
      "loss": 0.1072,
      "step": 2316
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.584142345656877e-05,
      "loss": 0.0642,
      "step": 2317
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.576324069675963e-05,
      "loss": 0.0945,
      "step": 2318
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.568508163586196e-05,
      "loss": 0.055,
      "step": 2319
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.560694632797797e-05,
      "loss": 0.0669,
      "step": 2320
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.552883482719349e-05,
      "loss": 0.1266,
      "step": 2321
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.545074718757774e-05,
      "loss": 0.0607,
      "step": 2322
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.537268346318355e-05,
      "loss": 0.0901,
      "step": 2323
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.529464370804718e-05,
      "loss": 0.1093,
      "step": 2324
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.52166279761882e-05,
      "loss": 0.0809,
      "step": 2325
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.513863632160962e-05,
      "loss": 0.0882,
      "step": 2326
    },
    {
      "epoch": 1.21,
      "learning_rate": 6.506066879829779e-05,
      "loss": 0.0931,
      "step": 2327
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.498272546022238e-05,
      "loss": 0.1703,
      "step": 2328
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.490480636133625e-05,
      "loss": 0.1728,
      "step": 2329
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.48269115555755e-05,
      "loss": 0.035,
      "step": 2330
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.474904109685949e-05,
      "loss": 0.1754,
      "step": 2331
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.467119503909061e-05,
      "loss": 0.0657,
      "step": 2332
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.459337343615449e-05,
      "loss": 0.1171,
      "step": 2333
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.451557634191965e-05,
      "loss": 0.1274,
      "step": 2334
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.443780381023785e-05,
      "loss": 0.0411,
      "step": 2335
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.436005589494368e-05,
      "loss": 0.1159,
      "step": 2336
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.42823326498548e-05,
      "loss": 0.0535,
      "step": 2337
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.420463412877176e-05,
      "loss": 0.0813,
      "step": 2338
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.412696038547796e-05,
      "loss": 0.0857,
      "step": 2339
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.40493114737397e-05,
      "loss": 0.0376,
      "step": 2340
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.397168744730613e-05,
      "loss": 0.0262,
      "step": 2341
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.389408835990901e-05,
      "loss": 0.068,
      "step": 2342
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.381651426526298e-05,
      "loss": 0.0757,
      "step": 2343
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.373896521706535e-05,
      "loss": 0.0706,
      "step": 2344
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.36614412689961e-05,
      "loss": 0.0439,
      "step": 2345
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.358394247471778e-05,
      "loss": 0.1295,
      "step": 2346
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.35064688878756e-05,
      "loss": 0.0699,
      "step": 2347
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.34290205620973e-05,
      "loss": 0.0773,
      "step": 2348
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.335159755099314e-05,
      "loss": 0.076,
      "step": 2349
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.327419990815575e-05,
      "loss": 0.0824,
      "step": 2350
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.319682768716036e-05,
      "loss": 0.0734,
      "step": 2351
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.311948094156452e-05,
      "loss": 0.0296,
      "step": 2352
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.304215972490817e-05,
      "loss": 0.0458,
      "step": 2353
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.296486409071353e-05,
      "loss": 0.0699,
      "step": 2354
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.288759409248523e-05,
      "loss": 0.0904,
      "step": 2355
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.281034978370999e-05,
      "loss": 0.0753,
      "step": 2356
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.273313121785686e-05,
      "loss": 0.1511,
      "step": 2357
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.265593844837709e-05,
      "loss": 0.0488,
      "step": 2358
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.257877152870397e-05,
      "loss": 0.078,
      "step": 2359
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.250163051225294e-05,
      "loss": 0.0854,
      "step": 2360
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.242451545242156e-05,
      "loss": 0.1152,
      "step": 2361
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.234742640258938e-05,
      "loss": 0.058,
      "step": 2362
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.227036341611794e-05,
      "loss": 0.109,
      "step": 2363
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.219332654635074e-05,
      "loss": 0.0502,
      "step": 2364
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.211631584661325e-05,
      "loss": 0.0809,
      "step": 2365
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.203933137021278e-05,
      "loss": 0.1165,
      "step": 2366
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.196237317043845e-05,
      "loss": 0.0916,
      "step": 2367
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.188544130056125e-05,
      "loss": 0.1445,
      "step": 2368
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.180853581383396e-05,
      "loss": 0.0545,
      "step": 2369
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.173165676349103e-05,
      "loss": 0.1273,
      "step": 2370
    },
    {
      "epoch": 1.24,
      "eval_loss": 0.10482697188854218,
      "eval_runtime": 683.6084,
      "eval_samples_per_second": 9.811,
      "eval_steps_per_second": 2.453,
      "step": 2370
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.165480420274867e-05,
      "loss": 0.0909,
      "step": 2371
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.157797818480475e-05,
      "loss": 0.0945,
      "step": 2372
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.15011787628387e-05,
      "loss": 0.2044,
      "step": 2373
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.142440599001162e-05,
      "loss": 0.1513,
      "step": 2374
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.134765991946617e-05,
      "loss": 0.0956,
      "step": 2375
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.127094060432642e-05,
      "loss": 0.0323,
      "step": 2376
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.119424809769797e-05,
      "loss": 0.1372,
      "step": 2377
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.111758245266794e-05,
      "loss": 0.1613,
      "step": 2378
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.10409437223048e-05,
      "loss": 0.0947,
      "step": 2379
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.096433195965835e-05,
      "loss": 0.0909,
      "step": 2380
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.088774721775977e-05,
      "loss": 0.1425,
      "step": 2381
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.081118954962155e-05,
      "loss": 0.1001,
      "step": 2382
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.0734659008237423e-05,
      "loss": 0.1194,
      "step": 2383
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.065815564658227e-05,
      "loss": 0.0627,
      "step": 2384
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.0581679517612265e-05,
      "loss": 0.1062,
      "step": 2385
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.05052306742647e-05,
      "loss": 0.1167,
      "step": 2386
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.0428809169457945e-05,
      "loss": 0.1285,
      "step": 2387
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.0352415056091494e-05,
      "loss": 0.0943,
      "step": 2388
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.027604838704587e-05,
      "loss": 0.1228,
      "step": 2389
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.0199709215182544e-05,
      "loss": 0.1469,
      "step": 2390
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.0123397593344024e-05,
      "loss": 0.1138,
      "step": 2391
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.0047113574353755e-05,
      "loss": 0.0444,
      "step": 2392
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.9970857211015966e-05,
      "loss": 0.1636,
      "step": 2393
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.9894628556115854e-05,
      "loss": 0.1626,
      "step": 2394
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.981842766241938e-05,
      "loss": 0.127,
      "step": 2395
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.9742254582673354e-05,
      "loss": 0.0848,
      "step": 2396
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.966610936960524e-05,
      "loss": 0.1059,
      "step": 2397
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.9589992075923265e-05,
      "loss": 0.075,
      "step": 2398
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.951390275431634e-05,
      "loss": 0.0685,
      "step": 2399
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.9437841457454035e-05,
      "loss": 0.3008,
      "step": 2400
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.936180823798638e-05,
      "loss": 0.0862,
      "step": 2401
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.928580314854415e-05,
      "loss": 0.0646,
      "step": 2402
    },
    {
      "epoch": 1.25,
      "learning_rate": 5.920982624173856e-05,
      "loss": 0.1164,
      "step": 2403
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.913387757016131e-05,
      "loss": 0.0646,
      "step": 2404
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.905795718638457e-05,
      "loss": 0.1576,
      "step": 2405
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.898206514296097e-05,
      "loss": 0.0632,
      "step": 2406
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.890620149242345e-05,
      "loss": 0.0269,
      "step": 2407
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.883036628728538e-05,
      "loss": 0.0929,
      "step": 2408
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.8754559580040305e-05,
      "loss": 0.1621,
      "step": 2409
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.867878142316221e-05,
      "loss": 0.0806,
      "step": 2410
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.860303186910518e-05,
      "loss": 0.0844,
      "step": 2411
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.8527310970303575e-05,
      "loss": 0.108,
      "step": 2412
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.845161877917194e-05,
      "loss": 0.0541,
      "step": 2413
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.837595534810485e-05,
      "loss": 0.1124,
      "step": 2414
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.830032072947711e-05,
      "loss": 0.0885,
      "step": 2415
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.822471497564341e-05,
      "loss": 0.0719,
      "step": 2416
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.814913813893865e-05,
      "loss": 0.1202,
      "step": 2417
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.8073590271677536e-05,
      "loss": 0.1148,
      "step": 2418
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.799807142615483e-05,
      "loss": 0.1105,
      "step": 2419
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.792258165464519e-05,
      "loss": 0.0861,
      "step": 2420
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.7847121009403104e-05,
      "loss": 0.0772,
      "step": 2421
    },
    {
      "epoch": 1.26,
      "learning_rate": 5.777168954266301e-05,
      "loss": 0.0381,
      "step": 2422
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.7696287306638974e-05,
      "loss": 0.1223,
      "step": 2423
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.762091435352495e-05,
      "loss": 0.0875,
      "step": 2424
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.7545570735494644e-05,
      "loss": 0.142,
      "step": 2425
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.7470256504701347e-05,
      "loss": 0.0619,
      "step": 2426
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.739497171327808e-05,
      "loss": 0.0831,
      "step": 2427
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.7319716413337485e-05,
      "loss": 0.0541,
      "step": 2428
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.7244490656971815e-05,
      "loss": 0.0891,
      "step": 2429
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.716929449625277e-05,
      "loss": 0.0676,
      "step": 2430
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.709412798323164e-05,
      "loss": 0.0927,
      "step": 2431
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.701899116993923e-05,
      "loss": 0.1105,
      "step": 2432
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.694388410838569e-05,
      "loss": 0.1214,
      "step": 2433
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.6868806850560694e-05,
      "loss": 0.0973,
      "step": 2434
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.679375944843314e-05,
      "loss": 0.0546,
      "step": 2435
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.6718741953951404e-05,
      "loss": 0.0908,
      "step": 2436
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.664375441904302e-05,
      "loss": 0.1451,
      "step": 2437
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.656879689561487e-05,
      "loss": 0.0718,
      "step": 2438
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.64938694355531e-05,
      "loss": 0.1733,
      "step": 2439
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.6418972090722954e-05,
      "loss": 0.0944,
      "step": 2440
    },
    {
      "epoch": 1.27,
      "learning_rate": 5.634410491296888e-05,
      "loss": 0.0976,
      "step": 2441
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.626926795411447e-05,
      "loss": 0.046,
      "step": 2442
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.619446126596233e-05,
      "loss": 0.1905,
      "step": 2443
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.61196849002941e-05,
      "loss": 0.0697,
      "step": 2444
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.6044938908870527e-05,
      "loss": 0.0964,
      "step": 2445
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.597022334343125e-05,
      "loss": 0.1153,
      "step": 2446
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.589553825569491e-05,
      "loss": 0.0879,
      "step": 2447
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.582088369735903e-05,
      "loss": 0.0556,
      "step": 2448
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.574625972009998e-05,
      "loss": 0.1727,
      "step": 2449
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.567166637557293e-05,
      "loss": 0.1674,
      "step": 2450
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.559710371541196e-05,
      "loss": 0.093,
      "step": 2451
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.552257179122975e-05,
      "loss": 0.1594,
      "step": 2452
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.544807065461783e-05,
      "loss": 0.0793,
      "step": 2453
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.5373600357146395e-05,
      "loss": 0.1759,
      "step": 2454
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.5299160950364256e-05,
      "loss": 0.1561,
      "step": 2455
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.5224752485798924e-05,
      "loss": 0.2577,
      "step": 2456
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.515037501495634e-05,
      "loss": 0.1261,
      "step": 2457
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.507602858932113e-05,
      "loss": 0.0247,
      "step": 2458
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.50017132603564e-05,
      "loss": 0.0482,
      "step": 2459
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.4927429079503656e-05,
      "loss": 0.0624,
      "step": 2460
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.4853176098182936e-05,
      "loss": 0.0905,
      "step": 2461
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.4778954367792635e-05,
      "loss": 0.064,
      "step": 2462
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.470476393970958e-05,
      "loss": 0.0763,
      "step": 2463
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.4630604865288794e-05,
      "loss": 0.1469,
      "step": 2464
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.45564771958637e-05,
      "loss": 0.0748,
      "step": 2465
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.4482380982745986e-05,
      "loss": 0.1719,
      "step": 2466
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.440831627722558e-05,
      "loss": 0.115,
      "step": 2467
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.4334283130570453e-05,
      "loss": 0.0791,
      "step": 2468
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.426028159402692e-05,
      "loss": 0.1614,
      "step": 2469
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.4186311718819335e-05,
      "loss": 0.065,
      "step": 2470
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.411237355615009e-05,
      "loss": 0.0781,
      "step": 2471
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.403846715719969e-05,
      "loss": 0.088,
      "step": 2472
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.396459257312663e-05,
      "loss": 0.1521,
      "step": 2473
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.38907498550674e-05,
      "loss": 0.0367,
      "step": 2474
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.381693905413643e-05,
      "loss": 0.0486,
      "step": 2475
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.374316022142605e-05,
      "loss": 0.0711,
      "step": 2476
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.366941340800644e-05,
      "loss": 0.0555,
      "step": 2477
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.3595698664925573e-05,
      "loss": 0.0394,
      "step": 2478
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.3522016043209364e-05,
      "loss": 0.0749,
      "step": 2479
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.344836559386137e-05,
      "loss": 0.0764,
      "step": 2480
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.3374747367862944e-05,
      "loss": 0.1086,
      "step": 2481
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.330116141617308e-05,
      "loss": 0.0813,
      "step": 2482
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.322760778972853e-05,
      "loss": 0.1497,
      "step": 2483
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.315408653944356e-05,
      "loss": 0.0943,
      "step": 2484
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.308059771621001e-05,
      "loss": 0.1049,
      "step": 2485
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.3007141370897386e-05,
      "loss": 0.121,
      "step": 2486
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.2933717554352636e-05,
      "loss": 0.1011,
      "step": 2487
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.286032631740023e-05,
      "loss": 0.1002,
      "step": 2488
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.2786967710842075e-05,
      "loss": 0.0695,
      "step": 2489
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.27136417854575e-05,
      "loss": 0.0763,
      "step": 2490
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.264034859200313e-05,
      "loss": 0.1524,
      "step": 2491
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.256708818121303e-05,
      "loss": 0.0782,
      "step": 2492
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.249386060379859e-05,
      "loss": 0.1866,
      "step": 2493
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.2420665910448354e-05,
      "loss": 0.0765,
      "step": 2494
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.234750415182819e-05,
      "loss": 0.0458,
      "step": 2495
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.227437537858117e-05,
      "loss": 0.0347,
      "step": 2496
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.220127964132755e-05,
      "loss": 0.0488,
      "step": 2497
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.212821699066459e-05,
      "loss": 0.0812,
      "step": 2498
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.20551874771668e-05,
      "loss": 0.048,
      "step": 2499
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.198219115138569e-05,
      "loss": 0.1158,
      "step": 2500
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.1909228063849836e-05,
      "loss": 0.0245,
      "step": 2501
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.1836298265064675e-05,
      "loss": 0.048,
      "step": 2502
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.176340180551276e-05,
      "loss": 0.0849,
      "step": 2503
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.169053873565353e-05,
      "loss": 0.0572,
      "step": 2504
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.161770910592321e-05,
      "loss": 0.0726,
      "step": 2505
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.1544912966734994e-05,
      "loss": 0.0501,
      "step": 2506
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.147215036847884e-05,
      "loss": 0.1321,
      "step": 2507
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.1399421361521514e-05,
      "loss": 0.1114,
      "step": 2508
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.1326725996206514e-05,
      "loss": 0.0922,
      "step": 2509
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.125406432285408e-05,
      "loss": 0.0248,
      "step": 2510
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.1181436391761076e-05,
      "loss": 0.1126,
      "step": 2511
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.1108842253200987e-05,
      "loss": 0.0651,
      "step": 2512
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.103628195742401e-05,
      "loss": 0.1563,
      "step": 2513
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.096375555465686e-05,
      "loss": 0.0816,
      "step": 2514
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.089126309510278e-05,
      "loss": 0.0648,
      "step": 2515
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.081880462894154e-05,
      "loss": 0.0639,
      "step": 2516
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.0746380206329404e-05,
      "loss": 0.0654,
      "step": 2517
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.067398987739901e-05,
      "loss": 0.0434,
      "step": 2518
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.0601633692259364e-05,
      "loss": 0.1092,
      "step": 2519
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.0529311700995964e-05,
      "loss": 0.1263,
      "step": 2520
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.0457023953670544e-05,
      "loss": 0.0838,
      "step": 2521
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.0384770500321176e-05,
      "loss": 0.1008,
      "step": 2522
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.031255139096216e-05,
      "loss": 0.1314,
      "step": 2523
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.0240366675584115e-05,
      "loss": 0.0947,
      "step": 2524
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.016821640415367e-05,
      "loss": 0.0765,
      "step": 2525
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.009610062661381e-05,
      "loss": 0.1054,
      "step": 2526
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.002401939288348e-05,
      "loss": 0.0824,
      "step": 2527
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.9951972752857814e-05,
      "loss": 0.0416,
      "step": 2528
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.9879960756407985e-05,
      "loss": 0.0728,
      "step": 2529
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.980798345338117e-05,
      "loss": 0.056,
      "step": 2530
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.9736040893600564e-05,
      "loss": 0.1486,
      "step": 2531
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.966413312686522e-05,
      "loss": 0.0737,
      "step": 2532
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.9592260202950206e-05,
      "loss": 0.0503,
      "step": 2533
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.952042217160643e-05,
      "loss": 0.1406,
      "step": 2534
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.94486190825607e-05,
      "loss": 0.0658,
      "step": 2535
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.937685098551551e-05,
      "loss": 0.1016,
      "step": 2536
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.930511793014924e-05,
      "loss": 0.0585,
      "step": 2537
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.9233419966116036e-05,
      "loss": 0.127,
      "step": 2538
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.916175714304562e-05,
      "loss": 0.0981,
      "step": 2539
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.909012951054352e-05,
      "loss": 0.085,
      "step": 2540
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.9018537118190843e-05,
      "loss": 0.1226,
      "step": 2541
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.894698001554433e-05,
      "loss": 0.0733,
      "step": 2542
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.8875458252136306e-05,
      "loss": 0.096,
      "step": 2543
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.8803971877474555e-05,
      "loss": 0.0681,
      "step": 2544
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.873252094104248e-05,
      "loss": 0.1518,
      "step": 2545
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.866110549229881e-05,
      "loss": 0.1076,
      "step": 2546
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.8589725580677835e-05,
      "loss": 0.0821,
      "step": 2547
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.851838125558922e-05,
      "loss": 0.0615,
      "step": 2548
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.844707256641795e-05,
      "loss": 0.0256,
      "step": 2549
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.8375799562524395e-05,
      "loss": 0.1692,
      "step": 2550
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.8304562293244224e-05,
      "loss": 0.0635,
      "step": 2551
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.823336080788832e-05,
      "loss": 0.1632,
      "step": 2552
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.816219515574276e-05,
      "loss": 0.0472,
      "step": 2553
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.809106538606896e-05,
      "loss": 0.0594,
      "step": 2554
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.801997154810337e-05,
      "loss": 0.1395,
      "step": 2555
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.794891369105765e-05,
      "loss": 0.0625,
      "step": 2556
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.787789186411851e-05,
      "loss": 0.0471,
      "step": 2557
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.780690611644776e-05,
      "loss": 0.0416,
      "step": 2558
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.7735956497182134e-05,
      "loss": 0.1937,
      "step": 2559
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.766504305543351e-05,
      "loss": 0.0424,
      "step": 2560
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.759416584028855e-05,
      "loss": 0.0912,
      "step": 2561
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.752332490080899e-05,
      "loss": 0.1145,
      "step": 2562
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.7452520286031366e-05,
      "loss": 0.1212,
      "step": 2563
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.738175204496713e-05,
      "loss": 0.0465,
      "step": 2564
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.731102022660257e-05,
      "loss": 0.0843,
      "step": 2565
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.724032487989861e-05,
      "loss": 0.0817,
      "step": 2566
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.716966605379111e-05,
      "loss": 0.0512,
      "step": 2567
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.7099043797190556e-05,
      "loss": 0.1037,
      "step": 2568
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.702845815898218e-05,
      "loss": 0.0659,
      "step": 2569
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.695790918802576e-05,
      "loss": 0.0379,
      "step": 2570
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.6887396933155805e-05,
      "loss": 0.0669,
      "step": 2571
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.681692144318138e-05,
      "loss": 0.0832,
      "step": 2572
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.6746482766886034e-05,
      "loss": 0.1109,
      "step": 2573
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.66760809530279e-05,
      "loss": 0.0735,
      "step": 2574
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.660571605033961e-05,
      "loss": 0.1531,
      "step": 2575
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.6535388107528185e-05,
      "loss": 0.0459,
      "step": 2576
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.6465097173275153e-05,
      "loss": 0.0516,
      "step": 2577
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.639484329623627e-05,
      "loss": 0.0522,
      "step": 2578
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.6324626525041814e-05,
      "loss": 0.1115,
      "step": 2579
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.6254446908296246e-05,
      "loss": 0.0814,
      "step": 2580
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.6184304494578366e-05,
      "loss": 0.0854,
      "step": 2581
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.6114199332441245e-05,
      "loss": 0.0642,
      "step": 2582
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.604413147041212e-05,
      "loss": 0.1364,
      "step": 2583
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.597410095699245e-05,
      "loss": 0.0837,
      "step": 2584
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.590410784065783e-05,
      "loss": 0.0652,
      "step": 2585
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.58341521698579e-05,
      "loss": 0.0739,
      "step": 2586
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.576423399301652e-05,
      "loss": 0.0385,
      "step": 2587
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.569435335853143e-05,
      "loss": 0.071,
      "step": 2588
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.56245103147745e-05,
      "loss": 0.15,
      "step": 2589
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.555470491009157e-05,
      "loss": 0.0826,
      "step": 2590
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.548493719280238e-05,
      "loss": 0.0489,
      "step": 2591
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.541520721120065e-05,
      "loss": 0.0681,
      "step": 2592
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.5345515013553865e-05,
      "loss": 0.1,
      "step": 2593
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.5275860648103494e-05,
      "loss": 0.0777,
      "step": 2594
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.520624416306468e-05,
      "loss": 0.0749,
      "step": 2595
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.513666560662645e-05,
      "loss": 0.1063,
      "step": 2596
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.506712502695154e-05,
      "loss": 0.0649,
      "step": 2597
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.499762247217639e-05,
      "loss": 0.0362,
      "step": 2598
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.49281579904112e-05,
      "loss": 0.1094,
      "step": 2599
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.485873162973964e-05,
      "loss": 0.0671,
      "step": 2600
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.4789343438219135e-05,
      "loss": 0.0431,
      "step": 2601
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.47199934638807e-05,
      "loss": 0.0944,
      "step": 2602
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.465068175472875e-05,
      "loss": 0.1281,
      "step": 2603
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.458140835874137e-05,
      "loss": 0.1974,
      "step": 2604
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.4512173323870064e-05,
      "loss": 0.0482,
      "step": 2605
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.444297669803981e-05,
      "loss": 0.0498,
      "step": 2606
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.437381852914888e-05,
      "loss": 0.0866,
      "step": 2607
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.430469886506908e-05,
      "loss": 0.0882,
      "step": 2608
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.42356177536455e-05,
      "loss": 0.1468,
      "step": 2609
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.416657524269652e-05,
      "loss": 0.0793,
      "step": 2610
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.409757138001387e-05,
      "loss": 0.1312,
      "step": 2611
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.402860621336241e-05,
      "loss": 0.1066,
      "step": 2612
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3959679790480326e-05,
      "loss": 0.1139,
      "step": 2613
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.389079215907896e-05,
      "loss": 0.0957,
      "step": 2614
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3821943366842745e-05,
      "loss": 0.057,
      "step": 2615
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3753133461429285e-05,
      "loss": 0.0989,
      "step": 2616
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.368436249046925e-05,
      "loss": 0.1212,
      "step": 2617
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3615630501566384e-05,
      "loss": 0.0223,
      "step": 2618
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.354693754229744e-05,
      "loss": 0.0527,
      "step": 2619
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3478283660212086e-05,
      "loss": 0.0586,
      "step": 2620
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3409668902833036e-05,
      "loss": 0.1301,
      "step": 2621
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3341093317655826e-05,
      "loss": 0.0857,
      "step": 2622
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3272556952148966e-05,
      "loss": 0.0731,
      "step": 2623
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3204059853753784e-05,
      "loss": 0.09,
      "step": 2624
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.3135602069884426e-05,
      "loss": 0.092,
      "step": 2625
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.306718364792786e-05,
      "loss": 0.0905,
      "step": 2626
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.299880463524367e-05,
      "loss": 0.1044,
      "step": 2627
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.2930465079164386e-05,
      "loss": 0.0591,
      "step": 2628
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.286216502699498e-05,
      "loss": 0.0477,
      "step": 2629
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.279390452601326e-05,
      "loss": 0.0314,
      "step": 2630
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.27256836234696e-05,
      "loss": 0.1064,
      "step": 2631
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.265750236658696e-05,
      "loss": 0.0725,
      "step": 2632
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.258936080256083e-05,
      "loss": 0.0607,
      "step": 2633
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.252125897855932e-05,
      "loss": 0.1147,
      "step": 2634
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.245319694172289e-05,
      "loss": 0.0899,
      "step": 2635
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.238517473916458e-05,
      "loss": 0.1707,
      "step": 2636
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.231719241796976e-05,
      "loss": 0.0952,
      "step": 2637
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.2249250025196274e-05,
      "loss": 0.0865,
      "step": 2638
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.218134760787429e-05,
      "loss": 0.0673,
      "step": 2639
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.211348521300631e-05,
      "loss": 0.0841,
      "step": 2640
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.204566288756717e-05,
      "loss": 0.0602,
      "step": 2641
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.197788067850388e-05,
      "loss": 0.0494,
      "step": 2642
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.191013863273574e-05,
      "loss": 0.1432,
      "step": 2643
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.18424367971543e-05,
      "loss": 0.0579,
      "step": 2644
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.177477521862313e-05,
      "loss": 0.0826,
      "step": 2645
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.1707153943978075e-05,
      "loss": 0.0705,
      "step": 2646
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.163957302002701e-05,
      "loss": 0.153,
      "step": 2647
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.157203249354996e-05,
      "loss": 0.0865,
      "step": 2648
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.1504532411298826e-05,
      "loss": 0.0759,
      "step": 2649
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.143707281999767e-05,
      "loss": 0.0283,
      "step": 2650
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.136965376634245e-05,
      "loss": 0.1535,
      "step": 2651
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.13022752970011e-05,
      "loss": 0.1397,
      "step": 2652
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.123493745861346e-05,
      "loss": 0.1029,
      "step": 2653
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.116764029779117e-05,
      "loss": 0.0354,
      "step": 2654
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.110038386111783e-05,
      "loss": 0.0426,
      "step": 2655
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.103316819514873e-05,
      "loss": 0.1106,
      "step": 2656
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.0965993346411025e-05,
      "loss": 0.0949,
      "step": 2657
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.08988593614036e-05,
      "loss": 0.0622,
      "step": 2658
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.083176628659703e-05,
      "loss": 0.0333,
      "step": 2659
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.0764714168433604e-05,
      "loss": 0.0666,
      "step": 2660
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.069770305332726e-05,
      "loss": 0.2162,
      "step": 2661
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.063073298766351e-05,
      "loss": 0.1215,
      "step": 2662
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.0563804017799446e-05,
      "loss": 0.0484,
      "step": 2663
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.0496916190063764e-05,
      "loss": 0.0682,
      "step": 2664
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.0430069550756665e-05,
      "loss": 0.1131,
      "step": 2665
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.036326414614985e-05,
      "loss": 0.1029,
      "step": 2666
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.029650002248645e-05,
      "loss": 0.0645,
      "step": 2667
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.022977722598108e-05,
      "loss": 0.1441,
      "step": 2668
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.016309580281962e-05,
      "loss": 0.0385,
      "step": 2669
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.009645579915946e-05,
      "loss": 0.0505,
      "step": 2670
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.0029857261129175e-05,
      "loss": 0.118,
      "step": 2671
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.996330023482878e-05,
      "loss": 0.107,
      "step": 2672
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.989678476632947e-05,
      "loss": 0.0595,
      "step": 2673
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.9830310901673685e-05,
      "loss": 0.0969,
      "step": 2674
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.9763878686875124e-05,
      "loss": 0.1301,
      "step": 2675
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.969748816791852e-05,
      "loss": 0.0419,
      "step": 2676
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.963113939075989e-05,
      "loss": 0.1376,
      "step": 2677
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.95648324013263e-05,
      "loss": 0.1047,
      "step": 2678
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.949856724551584e-05,
      "loss": 0.1274,
      "step": 2679
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.9432343969197714e-05,
      "loss": 0.1325,
      "step": 2680
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.936616261821212e-05,
      "loss": 0.0674,
      "step": 2681
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.930002323837025e-05,
      "loss": 0.0652,
      "step": 2682
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.923392587545417e-05,
      "loss": 0.1155,
      "step": 2683
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.9167870575216937e-05,
      "loss": 0.0485,
      "step": 2684
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.9101857383382455e-05,
      "loss": 0.0799,
      "step": 2685
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.9035886345645514e-05,
      "loss": 0.1322,
      "step": 2686
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.896995750767173e-05,
      "loss": 0.0749,
      "step": 2687
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.89040709150974e-05,
      "loss": 0.0961,
      "step": 2688
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.8838226613529736e-05,
      "loss": 0.2192,
      "step": 2689
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.8772424648546545e-05,
      "loss": 0.2841,
      "step": 2690
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.87066650656964e-05,
      "loss": 0.0514,
      "step": 2691
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.8640947910498524e-05,
      "loss": 0.115,
      "step": 2692
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.857527322844276e-05,
      "loss": 0.0974,
      "step": 2693
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.850964106498958e-05,
      "loss": 0.0615,
      "step": 2694
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.8444051465570005e-05,
      "loss": 0.0616,
      "step": 2695
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.8378504475585584e-05,
      "loss": 0.0826,
      "step": 2696
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.831300014040834e-05,
      "loss": 0.113,
      "step": 2697
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.824753850538082e-05,
      "loss": 0.133,
      "step": 2698
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.8182119615816026e-05,
      "loss": 0.0927,
      "step": 2699
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.8116743516997325e-05,
      "loss": 0.1341,
      "step": 2700
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.80514102541785e-05,
      "loss": 0.0677,
      "step": 2701
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.7986119872583693e-05,
      "loss": 0.0986,
      "step": 2702
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.7920872417407314e-05,
      "loss": 0.053,
      "step": 2703
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.785566793381404e-05,
      "loss": 0.1009,
      "step": 2704
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.779050646693889e-05,
      "loss": 0.1696,
      "step": 2705
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.7725388061887056e-05,
      "loss": 0.0769,
      "step": 2706
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.766031276373394e-05,
      "loss": 0.0524,
      "step": 2707
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.7595280617525075e-05,
      "loss": 0.0864,
      "step": 2708
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.753029166827618e-05,
      "loss": 0.1651,
      "step": 2709
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.7465345960972985e-05,
      "loss": 0.0771,
      "step": 2710
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.740044354057137e-05,
      "loss": 0.089,
      "step": 2711
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.733558445199724e-05,
      "loss": 0.097,
      "step": 2712
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.7270768740146424e-05,
      "loss": 0.0826,
      "step": 2713
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.720599644988482e-05,
      "loss": 0.052,
      "step": 2714
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.714126762604824e-05,
      "loss": 0.0568,
      "step": 2715
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.707658231344243e-05,
      "loss": 0.1081,
      "step": 2716
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.7011940556842936e-05,
      "loss": 0.2073,
      "step": 2717
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.694734240099524e-05,
      "loss": 0.0513,
      "step": 2718
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.688278789061461e-05,
      "loss": 0.1981,
      "step": 2719
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.6818277070386144e-05,
      "loss": 0.0793,
      "step": 2720
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.6753809984964583e-05,
      "loss": 0.1517,
      "step": 2721
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.6689386678974505e-05,
      "loss": 0.0668,
      "step": 2722
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.662500719701019e-05,
      "loss": 0.0642,
      "step": 2723
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.6560671583635467e-05,
      "loss": 0.1131,
      "step": 2724
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.6496379883383916e-05,
      "loss": 0.0616,
      "step": 2725
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.643213214075867e-05,
      "loss": 0.1541,
      "step": 2726
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.6367928400232444e-05,
      "loss": 0.1014,
      "step": 2727
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.630376870624749e-05,
      "loss": 0.0771,
      "step": 2728
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.623965310321561e-05,
      "loss": 0.1044,
      "step": 2729
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.617558163551802e-05,
      "loss": 0.063,
      "step": 2730
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.611155434750537e-05,
      "loss": 0.1937,
      "step": 2731
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.604757128349782e-05,
      "loss": 0.0552,
      "step": 2732
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.5983632487784854e-05,
      "loss": 0.1233,
      "step": 2733
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.5919738004625333e-05,
      "loss": 0.0923,
      "step": 2734
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.5855887878247454e-05,
      "loss": 0.0892,
      "step": 2735
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.5792082152848725e-05,
      "loss": 0.0641,
      "step": 2736
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.5728320872595856e-05,
      "loss": 0.0591,
      "step": 2737
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.5664604081624786e-05,
      "loss": 0.1321,
      "step": 2738
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.560093182404074e-05,
      "loss": 0.0584,
      "step": 2739
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.553730414391808e-05,
      "loss": 0.0936,
      "step": 2740
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.547372108530031e-05,
      "loss": 0.0814,
      "step": 2741
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.541018269220003e-05,
      "loss": 0.0821,
      "step": 2742
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.5346689008598985e-05,
      "loss": 0.0815,
      "step": 2743
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.5283240078447845e-05,
      "loss": 0.154,
      "step": 2744
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.5219835945666424e-05,
      "loss": 0.1385,
      "step": 2745
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.5156476654143497e-05,
      "loss": 0.1228,
      "step": 2746
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.509316224773674e-05,
      "loss": 0.0569,
      "step": 2747
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.502989277027283e-05,
      "loss": 0.0512,
      "step": 2748
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.496666826554732e-05,
      "loss": 0.0946,
      "step": 2749
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.490348877732465e-05,
      "loss": 0.0707,
      "step": 2750
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.484035434933803e-05,
      "loss": 0.0866,
      "step": 2751
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.4777265025289565e-05,
      "loss": 0.0573,
      "step": 2752
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.4714220848850085e-05,
      "loss": 0.0549,
      "step": 2753
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.465122186365923e-05,
      "loss": 0.0614,
      "step": 2754
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.458826811332525e-05,
      "loss": 0.1258,
      "step": 2755
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.452535964142518e-05,
      "loss": 0.0856,
      "step": 2756
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.4462496491504705e-05,
      "loss": 0.0778,
      "step": 2757
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.439967870707807e-05,
      "loss": 0.0442,
      "step": 2758
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.433690633162816e-05,
      "loss": 0.1246,
      "step": 2759
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.427417940860646e-05,
      "loss": 0.0803,
      "step": 2760
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.421149798143294e-05,
      "loss": 0.0775,
      "step": 2761
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.414886209349615e-05,
      "loss": 0.1072,
      "step": 2762
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.408627178815299e-05,
      "loss": 0.0621,
      "step": 2763
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.4023727108728955e-05,
      "loss": 0.1154,
      "step": 2764
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.3961228098517805e-05,
      "loss": 0.0635,
      "step": 2765
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.389877480078183e-05,
      "loss": 0.093,
      "step": 2766
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.383636725875159e-05,
      "loss": 0.0586,
      "step": 2767
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.377400551562601e-05,
      "loss": 0.0668,
      "step": 2768
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.37116896145723e-05,
      "loss": 0.1031,
      "step": 2769
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.364941959872596e-05,
      "loss": 0.1159,
      "step": 2770
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.358719551119071e-05,
      "loss": 0.0793,
      "step": 2771
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.352501739503841e-05,
      "loss": 0.0605,
      "step": 2772
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.346288529330919e-05,
      "loss": 0.0823,
      "step": 2773
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.3400799249011326e-05,
      "loss": 0.1498,
      "step": 2774
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.333875930512119e-05,
      "loss": 0.1696,
      "step": 2775
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.327676550458323e-05,
      "loss": 0.0578,
      "step": 2776
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.321481789030999e-05,
      "loss": 0.0865,
      "step": 2777
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.315291650518197e-05,
      "loss": 0.0394,
      "step": 2778
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.309106139204778e-05,
      "loss": 0.1449,
      "step": 2779
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.302925259372388e-05,
      "loss": 0.0804,
      "step": 2780
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.296749015299474e-05,
      "loss": 0.092,
      "step": 2781
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.2905774112612754e-05,
      "loss": 0.1522,
      "step": 2782
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.2844104515298155e-05,
      "loss": 0.0536,
      "step": 2783
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.2782481403739096e-05,
      "loss": 0.0556,
      "step": 2784
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.272090482059142e-05,
      "loss": 0.1185,
      "step": 2785
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.265937480847889e-05,
      "loss": 0.1201,
      "step": 2786
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.2597891409992987e-05,
      "loss": 0.1286,
      "step": 2787
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.253645466769296e-05,
      "loss": 0.1267,
      "step": 2788
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.2475064624105665e-05,
      "loss": 0.0904,
      "step": 2789
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.241372132172572e-05,
      "loss": 0.1054,
      "step": 2790
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.23524248030154e-05,
      "loss": 0.1386,
      "step": 2791
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.22911751104045e-05,
      "loss": 0.0539,
      "step": 2792
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.22299722862905e-05,
      "loss": 0.0587,
      "step": 2793
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.216881637303839e-05,
      "loss": 0.0909,
      "step": 2794
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.21077074129807e-05,
      "loss": 0.075,
      "step": 2795
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.20466454484175e-05,
      "loss": 0.075,
      "step": 2796
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.198563052161621e-05,
      "loss": 0.0857,
      "step": 2797
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.1924662674811836e-05,
      "loss": 0.105,
      "step": 2798
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.1863741950206646e-05,
      "loss": 0.1458,
      "step": 2799
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.180286838997043e-05,
      "loss": 0.0356,
      "step": 2800
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.174204203624024e-05,
      "loss": 0.0705,
      "step": 2801
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.1681262931120504e-05,
      "loss": 0.1267,
      "step": 2802
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.1620531116682906e-05,
      "loss": 0.1076,
      "step": 2803
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.155984663496645e-05,
      "loss": 0.0893,
      "step": 2804
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.14992095279773e-05,
      "loss": 0.0784,
      "step": 2805
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.143861983768883e-05,
      "loss": 0.1158,
      "step": 2806
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.1378077606041666e-05,
      "loss": 0.1221,
      "step": 2807
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.131758287494353e-05,
      "loss": 0.0755,
      "step": 2808
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.125713568626928e-05,
      "loss": 0.0317,
      "step": 2809
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.119673608186085e-05,
      "loss": 0.0481,
      "step": 2810
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.1136384103527306e-05,
      "loss": 0.101,
      "step": 2811
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.10760797930446e-05,
      "loss": 0.058,
      "step": 2812
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.101582319215583e-05,
      "loss": 0.0965,
      "step": 2813
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.095561434257098e-05,
      "loss": 0.1303,
      "step": 2814
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.0895453285967024e-05,
      "loss": 0.0917,
      "step": 2815
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.083534006398786e-05,
      "loss": 0.1468,
      "step": 2816
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.077527471824424e-05,
      "loss": 0.0539,
      "step": 2817
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.071525729031384e-05,
      "loss": 0.0553,
      "step": 2818
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.065528782174104e-05,
      "loss": 0.0999,
      "step": 2819
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.0595366354037155e-05,
      "loss": 0.0561,
      "step": 2820
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.053549292868025e-05,
      "loss": 0.0911,
      "step": 2821
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.047566758711503e-05,
      "loss": 0.0963,
      "step": 2822
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.041589037075304e-05,
      "loss": 0.065,
      "step": 2823
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.0356161320972455e-05,
      "loss": 0.0557,
      "step": 2824
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.029648047911815e-05,
      "loss": 0.0813,
      "step": 2825
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.0236847886501542e-05,
      "loss": 0.0785,
      "step": 2826
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.0177263584400738e-05,
      "loss": 0.1265,
      "step": 2827
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.0117727614060376e-05,
      "loss": 0.1455,
      "step": 2828
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.0058240016691664e-05,
      "loss": 0.0173,
      "step": 2829
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.9998800833472317e-05,
      "loss": 0.1819,
      "step": 2830
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.9939410105546506e-05,
      "loss": 0.0658,
      "step": 2831
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.9880067874024887e-05,
      "loss": 0.0709,
      "step": 2832
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.982077417998459e-05,
      "loss": 0.0898,
      "step": 2833
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.976152906446903e-05,
      "loss": 0.0849,
      "step": 2834
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.970233256848811e-05,
      "loss": 0.0701,
      "step": 2835
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.9643184733018037e-05,
      "loss": 0.0577,
      "step": 2836
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.958408559900132e-05,
      "loss": 0.0778,
      "step": 2837
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.952503520734682e-05,
      "loss": 0.0356,
      "step": 2838
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.9466033598929542e-05,
      "loss": 0.0631,
      "step": 2839
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.9407080814590847e-05,
      "loss": 0.0477,
      "step": 2840
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.934817689513818e-05,
      "loss": 0.0781,
      "step": 2841
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.9289321881345254e-05,
      "loss": 0.0826,
      "step": 2842
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.9230515813951908e-05,
      "loss": 0.0893,
      "step": 2843
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.9171758733664088e-05,
      "loss": 0.0468,
      "step": 2844
    },
    {
      "epoch": 1.49,
      "eval_loss": 0.08970315754413605,
      "eval_runtime": 683.3858,
      "eval_samples_per_second": 9.814,
      "eval_steps_per_second": 2.454,
      "step": 2844
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.9113050681153865e-05,
      "loss": 0.1697,
      "step": 2845
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.905439169705927e-05,
      "loss": 0.0894,
      "step": 2846
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.8995781821984512e-05,
      "loss": 0.1855,
      "step": 2847
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.8937221096499677e-05,
      "loss": 0.0404,
      "step": 2848
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.8878709561140903e-05,
      "loss": 0.0811,
      "step": 2849
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.882024725641027e-05,
      "loss": 0.0398,
      "step": 2850
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.8761834222775787e-05,
      "loss": 0.1281,
      "step": 2851
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.8703470500671348e-05,
      "loss": 0.0465,
      "step": 2852
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.8645156130496665e-05,
      "loss": 0.1386,
      "step": 2853
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.8586891152617347e-05,
      "loss": 0.0433,
      "step": 2854
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.852867560736484e-05,
      "loss": 0.1056,
      "step": 2855
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.847050953503626e-05,
      "loss": 0.0539,
      "step": 2856
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.8412392975894588e-05,
      "loss": 0.1059,
      "step": 2857
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.8354325970168484e-05,
      "loss": 0.0377,
      "step": 2858
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.829630855805232e-05,
      "loss": 0.0602,
      "step": 2859
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.823834077970615e-05,
      "loss": 0.0262,
      "step": 2860
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.8180422675255613e-05,
      "loss": 0.075,
      "step": 2861
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.8122554284792014e-05,
      "loss": 0.0255,
      "step": 2862
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.8064735648372244e-05,
      "loss": 0.0512,
      "step": 2863
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.8006966806018787e-05,
      "loss": 0.0591,
      "step": 2864
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.7949247797719548e-05,
      "loss": 0.0936,
      "step": 2865
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.789157866342803e-05,
      "loss": 0.0343,
      "step": 2866
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.783395944306323e-05,
      "loss": 0.0913,
      "step": 2867
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.7776390176509493e-05,
      "loss": 0.0885,
      "step": 2868
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.7718870903616667e-05,
      "loss": 0.0573,
      "step": 2869
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.7661401664199993e-05,
      "loss": 0.0641,
      "step": 2870
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.7603982498040038e-05,
      "loss": 0.1368,
      "step": 2871
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.7546613444882775e-05,
      "loss": 0.0715,
      "step": 2872
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.748929454443937e-05,
      "loss": 0.1643,
      "step": 2873
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.743202583638641e-05,
      "loss": 0.0954,
      "step": 2874
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.7374807360365617e-05,
      "loss": 0.1119,
      "step": 2875
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.7317639155984022e-05,
      "loss": 0.0529,
      "step": 2876
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.7260521262813844e-05,
      "loss": 0.0484,
      "step": 2877
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.7203453720392458e-05,
      "loss": 0.0253,
      "step": 2878
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.71464365682224e-05,
      "loss": 0.0592,
      "step": 2879
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.7089469845771353e-05,
      "loss": 0.1353,
      "step": 2880
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.7032553592472032e-05,
      "loss": 0.1435,
      "step": 2881
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6975687847722196e-05,
      "loss": 0.0852,
      "step": 2882
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.691887265088473e-05,
      "loss": 0.0773,
      "step": 2883
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6862108041287503e-05,
      "loss": 0.0886,
      "step": 2884
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6805394058223333e-05,
      "loss": 0.0636,
      "step": 2885
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.674873074095002e-05,
      "loss": 0.1101,
      "step": 2886
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6692118128690336e-05,
      "loss": 0.064,
      "step": 2887
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.663555626063181e-05,
      "loss": 0.1281,
      "step": 2888
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6579045175927043e-05,
      "loss": 0.1028,
      "step": 2889
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6522584913693294e-05,
      "loss": 0.0706,
      "step": 2890
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6466175513012758e-05,
      "loss": 0.1134,
      "step": 2891
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6409817012932413e-05,
      "loss": 0.0574,
      "step": 2892
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6353509452463986e-05,
      "loss": 0.1831,
      "step": 2893
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6297252870583955e-05,
      "loss": 0.0781,
      "step": 2894
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6241047306233456e-05,
      "loss": 0.1934,
      "step": 2895
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6184892798318374e-05,
      "loss": 0.055,
      "step": 2896
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6128789385709263e-05,
      "loss": 0.0341,
      "step": 2897
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6072737107241208e-05,
      "loss": 0.0756,
      "step": 2898
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6016736001714015e-05,
      "loss": 0.0657,
      "step": 2899
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.596078610789201e-05,
      "loss": 0.0505,
      "step": 2900
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.5904887464504114e-05,
      "loss": 0.1116,
      "step": 2901
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.584904011024367e-05,
      "loss": 0.0663,
      "step": 2902
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.579324408376862e-05,
      "loss": 0.0542,
      "step": 2903
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.573749942370135e-05,
      "loss": 0.0542,
      "step": 2904
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.5681806168628676e-05,
      "loss": 0.045,
      "step": 2905
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.5626164357101857e-05,
      "loss": 0.0251,
      "step": 2906
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.5570574027636484e-05,
      "loss": 0.0679,
      "step": 2907
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.5515035218712612e-05,
      "loss": 0.1321,
      "step": 2908
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.545954796877449e-05,
      "loss": 0.0759,
      "step": 2909
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.5404112316230812e-05,
      "loss": 0.1152,
      "step": 2910
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.5348728299454482e-05,
      "loss": 0.0173,
      "step": 2911
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.5293395956782705e-05,
      "loss": 0.0782,
      "step": 2912
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.5238115326516885e-05,
      "loss": 0.1259,
      "step": 2913
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.518288644692267e-05,
      "loss": 0.027,
      "step": 2914
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.5127709356229824e-05,
      "loss": 0.0872,
      "step": 2915
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.5072584092632266e-05,
      "loss": 0.0917,
      "step": 2916
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.5017510694288105e-05,
      "loss": 0.041,
      "step": 2917
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4962489199319504e-05,
      "loss": 0.0739,
      "step": 2918
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4907519645812715e-05,
      "loss": 0.0512,
      "step": 2919
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.485260207181801e-05,
      "loss": 0.0859,
      "step": 2920
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4797736515349746e-05,
      "loss": 0.1467,
      "step": 2921
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4742923014386156e-05,
      "loss": 0.1037,
      "step": 2922
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4688161606869564e-05,
      "loss": 0.0555,
      "step": 2923
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.463345233070612e-05,
      "loss": 0.0566,
      "step": 2924
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4578795223765995e-05,
      "loss": 0.1963,
      "step": 2925
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4524190323883177e-05,
      "loss": 0.0896,
      "step": 2926
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.446963766885555e-05,
      "loss": 0.0695,
      "step": 2927
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4415137296444846e-05,
      "loss": 0.0524,
      "step": 2928
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4360689244376532e-05,
      "loss": 0.0355,
      "step": 2929
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.430629355033994e-05,
      "loss": 0.0604,
      "step": 2930
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4251950251988166e-05,
      "loss": 0.0872,
      "step": 2931
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.419765938693792e-05,
      "loss": 0.0612,
      "step": 2932
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.4143420992769738e-05,
      "loss": 0.1134,
      "step": 2933
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.4089235107027795e-05,
      "loss": 0.0449,
      "step": 2934
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.403510176721997e-05,
      "loss": 0.0625,
      "step": 2935
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3981021010817628e-05,
      "loss": 0.0497,
      "step": 2936
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3926992875255873e-05,
      "loss": 0.0777,
      "step": 2937
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3873017397933327e-05,
      "loss": 0.1613,
      "step": 2938
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3819094616212224e-05,
      "loss": 0.0479,
      "step": 2939
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.37652245674182e-05,
      "loss": 0.0639,
      "step": 2940
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.371140728884047e-05,
      "loss": 0.205,
      "step": 2941
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3657642817731774e-05,
      "loss": 0.0565,
      "step": 2942
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.360393119130815e-05,
      "loss": 0.1026,
      "step": 2943
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.355027244674919e-05,
      "loss": 0.0603,
      "step": 2944
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3496666621197818e-05,
      "loss": 0.0356,
      "step": 2945
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3443113751760336e-05,
      "loss": 0.1466,
      "step": 2946
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.338961387550641e-05,
      "loss": 0.0743,
      "step": 2947
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3336167029469037e-05,
      "loss": 0.079,
      "step": 2948
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.328277325064444e-05,
      "loss": 0.1784,
      "step": 2949
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3229432575992127e-05,
      "loss": 0.0556,
      "step": 2950
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.3176145042434883e-05,
      "loss": 0.1039,
      "step": 2951
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.31229106868587e-05,
      "loss": 0.1159,
      "step": 2952
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.306972954611275e-05,
      "loss": 0.0949,
      "step": 2953
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.301660165700936e-05,
      "loss": 0.1444,
      "step": 2954
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2963527056324054e-05,
      "loss": 0.0547,
      "step": 2955
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2910505780795388e-05,
      "loss": 0.0721,
      "step": 2956
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2857537867125e-05,
      "loss": 0.1097,
      "step": 2957
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.280462335197767e-05,
      "loss": 0.0667,
      "step": 2958
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2751762271981182e-05,
      "loss": 0.0717,
      "step": 2959
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.26989546637263e-05,
      "loss": 0.0656,
      "step": 2960
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2646200563766827e-05,
      "loss": 0.0232,
      "step": 2961
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2593500008619515e-05,
      "loss": 0.1099,
      "step": 2962
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2540853034763988e-05,
      "loss": 0.068,
      "step": 2963
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2488259678642852e-05,
      "loss": 0.1042,
      "step": 2964
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2435719976661597e-05,
      "loss": 0.1224,
      "step": 2965
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.238323396518852e-05,
      "loss": 0.0653,
      "step": 2966
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.233080168055479e-05,
      "loss": 0.1307,
      "step": 2967
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2278423159054396e-05,
      "loss": 0.0429,
      "step": 2968
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.222609843694411e-05,
      "loss": 0.0815,
      "step": 2969
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2173827550443417e-05,
      "loss": 0.0849,
      "step": 2970
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.2121610535734582e-05,
      "loss": 0.1121,
      "step": 2971
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.2069447428962575e-05,
      "loss": 0.0335,
      "step": 2972
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.2017338266235067e-05,
      "loss": 0.0715,
      "step": 2973
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1965283083622323e-05,
      "loss": 0.1417,
      "step": 2974
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.191328191715729e-05,
      "loss": 0.0977,
      "step": 2975
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1861334802835564e-05,
      "loss": 0.0664,
      "step": 2976
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1809441776615226e-05,
      "loss": 0.0979,
      "step": 2977
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1757602874416993e-05,
      "loss": 0.0926,
      "step": 2978
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1705818132124113e-05,
      "loss": 0.0177,
      "step": 2979
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1654087585582294e-05,
      "loss": 0.0382,
      "step": 2980
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1602411270599798e-05,
      "loss": 0.0314,
      "step": 2981
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1550789222947322e-05,
      "loss": 0.0214,
      "step": 2982
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1499221478357957e-05,
      "loss": 0.0525,
      "step": 2983
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1447708072527216e-05,
      "loss": 0.0976,
      "step": 2984
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.139624904111306e-05,
      "loss": 0.0966,
      "step": 2985
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1344844419735755e-05,
      "loss": 0.0484,
      "step": 2986
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.129349424397793e-05,
      "loss": 0.2254,
      "step": 2987
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1242198549384495e-05,
      "loss": 0.0924,
      "step": 2988
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1190957371462728e-05,
      "loss": 0.1534,
      "step": 2989
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.1139770745682065e-05,
      "loss": 0.1396,
      "step": 2990
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.1088638707474207e-05,
      "loss": 0.0535,
      "step": 2991
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.103756129223311e-05,
      "loss": 0.0771,
      "step": 2992
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0986538535314916e-05,
      "loss": 0.0365,
      "step": 2993
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.093557047203789e-05,
      "loss": 0.1484,
      "step": 2994
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0884657137682474e-05,
      "loss": 0.1226,
      "step": 2995
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0833798567491246e-05,
      "loss": 0.0885,
      "step": 2996
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0782994796668796e-05,
      "loss": 0.1306,
      "step": 2997
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0732245860381872e-05,
      "loss": 0.0615,
      "step": 2998
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.068155179375918e-05,
      "loss": 0.1227,
      "step": 2999
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0630912631891497e-05,
      "loss": 0.0951,
      "step": 3000
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0580328409831616e-05,
      "loss": 0.1047,
      "step": 3001
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0529799162594244e-05,
      "loss": 0.1146,
      "step": 3002
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0479324925156095e-05,
      "loss": 0.0766,
      "step": 3003
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0428905732455726e-05,
      "loss": 0.0582,
      "step": 3004
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0378541619393653e-05,
      "loss": 0.1033,
      "step": 3005
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0328232620832254e-05,
      "loss": 0.1062,
      "step": 3006
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0277978771595774e-05,
      "loss": 0.1081,
      "step": 3007
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0227780106470207e-05,
      "loss": 0.0555,
      "step": 3008
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.0177636660203425e-05,
      "loss": 0.0874,
      "step": 3009
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.0127548467505085e-05,
      "loss": 0.0267,
      "step": 3010
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.0077515563046513e-05,
      "loss": 0.068,
      "step": 3011
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.0027537981460842e-05,
      "loss": 0.0626,
      "step": 3012
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9977615757342883e-05,
      "loss": 0.1047,
      "step": 3013
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9927748925249135e-05,
      "loss": 0.0787,
      "step": 3014
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9877937519697776e-05,
      "loss": 0.0837,
      "step": 3015
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9828181575168526e-05,
      "loss": 0.0621,
      "step": 3016
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9778481126102845e-05,
      "loss": 0.0704,
      "step": 3017
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9728836206903656e-05,
      "loss": 0.0387,
      "step": 3018
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.967924685193552e-05,
      "loss": 0.0474,
      "step": 3019
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9629713095524525e-05,
      "loss": 0.0954,
      "step": 3020
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9580234971958268e-05,
      "loss": 0.0764,
      "step": 3021
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.953081251548583e-05,
      "loss": 0.0901,
      "step": 3022
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9481445760317786e-05,
      "loss": 0.0573,
      "step": 3023
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9432134740626107e-05,
      "loss": 0.0484,
      "step": 3024
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9382879490544183e-05,
      "loss": 0.1171,
      "step": 3025
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9333680044166845e-05,
      "loss": 0.1496,
      "step": 3026
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.928453643555028e-05,
      "loss": 0.1219,
      "step": 3027
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9235448698712023e-05,
      "loss": 0.051,
      "step": 3028
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.918641686763092e-05,
      "loss": 0.0577,
      "step": 3029
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.9137440976247156e-05,
      "loss": 0.1363,
      "step": 3030
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.908852105846213e-05,
      "loss": 0.0343,
      "step": 3031
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.9039657148138567e-05,
      "loss": 0.065,
      "step": 3032
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8990849279100353e-05,
      "loss": 0.0415,
      "step": 3033
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8942097485132626e-05,
      "loss": 0.0819,
      "step": 3034
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8893401799981714e-05,
      "loss": 0.0411,
      "step": 3035
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8844762257355097e-05,
      "loss": 0.042,
      "step": 3036
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.87961788909214e-05,
      "loss": 0.1114,
      "step": 3037
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8747651734310314e-05,
      "loss": 0.1179,
      "step": 3038
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8699180821112673e-05,
      "loss": 0.0813,
      "step": 3039
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8650766184880365e-05,
      "loss": 0.0531,
      "step": 3040
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8602407859126347e-05,
      "loss": 0.1137,
      "step": 3041
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8554105877324522e-05,
      "loss": 0.1069,
      "step": 3042
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.850586027290987e-05,
      "loss": 0.1888,
      "step": 3043
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.845767107927835e-05,
      "loss": 0.1233,
      "step": 3044
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.840953832978678e-05,
      "loss": 0.0566,
      "step": 3045
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8361462057752986e-05,
      "loss": 0.0771,
      "step": 3046
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.8313442296455695e-05,
      "loss": 0.0891,
      "step": 3047
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.8265479079134506e-05,
      "loss": 0.0628,
      "step": 3048
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.8217572438989895e-05,
      "loss": 0.0993,
      "step": 3049
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.8169722409183097e-05,
      "loss": 0.1087,
      "step": 3050
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.8121929022836282e-05,
      "loss": 0.0607,
      "step": 3051
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.807419231303229e-05,
      "loss": 0.1067,
      "step": 3052
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.8026512312814824e-05,
      "loss": 0.0781,
      "step": 3053
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7978889055188288e-05,
      "loss": 0.051,
      "step": 3054
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7931322573117816e-05,
      "loss": 0.055,
      "step": 3055
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.788381289952923e-05,
      "loss": 0.1096,
      "step": 3056
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.783636006730909e-05,
      "loss": 0.0508,
      "step": 3057
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7788964109304496e-05,
      "loss": 0.0347,
      "step": 3058
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7741625058323295e-05,
      "loss": 0.0443,
      "step": 3059
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7694342947133847e-05,
      "loss": 0.0417,
      "step": 3060
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7647117808465164e-05,
      "loss": 0.0521,
      "step": 3061
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7599949675006798e-05,
      "loss": 0.1143,
      "step": 3062
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.755283857940884e-05,
      "loss": 0.0247,
      "step": 3063
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.7505784554281936e-05,
      "loss": 0.0515,
      "step": 3064
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.745878763219714e-05,
      "loss": 0.0607,
      "step": 3065
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.741184784568608e-05,
      "loss": 0.0566,
      "step": 3066
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.7364965227240737e-05,
      "loss": 0.0284,
      "step": 3067
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.7318139809313604e-05,
      "loss": 0.0709,
      "step": 3068
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.7271371624317532e-05,
      "loss": 0.0676,
      "step": 3069
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.7224660704625782e-05,
      "loss": 0.0606,
      "step": 3070
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.717800708257198e-05,
      "loss": 0.0464,
      "step": 3071
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.713141079045001e-05,
      "loss": 0.0393,
      "step": 3072
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.7084871860514185e-05,
      "loss": 0.1122,
      "step": 3073
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.703839032497906e-05,
      "loss": 0.0999,
      "step": 3074
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.699196621601943e-05,
      "loss": 0.0686,
      "step": 3075
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6945599565770375e-05,
      "loss": 0.0507,
      "step": 3076
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.689929040632723e-05,
      "loss": 0.0384,
      "step": 3077
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6853038769745467e-05,
      "loss": 0.1292,
      "step": 3078
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6806844688040825e-05,
      "loss": 0.0404,
      "step": 3079
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6760708193189088e-05,
      "loss": 0.0327,
      "step": 3080
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6714629317126286e-05,
      "loss": 0.1191,
      "step": 3081
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6668608091748495e-05,
      "loss": 0.0556,
      "step": 3082
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.662264454891196e-05,
      "loss": 0.0591,
      "step": 3083
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6576738720432883e-05,
      "loss": 0.0709,
      "step": 3084
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6530890638087626e-05,
      "loss": 0.0293,
      "step": 3085
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.6485100333612548e-05,
      "loss": 0.0429,
      "step": 3086
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.6439367838703944e-05,
      "loss": 0.0978,
      "step": 3087
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.63936931850182e-05,
      "loss": 0.0594,
      "step": 3088
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.6348076404171587e-05,
      "loss": 0.0892,
      "step": 3089
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.6302517527740356e-05,
      "loss": 0.0542,
      "step": 3090
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.6257016587260686e-05,
      "loss": 0.0822,
      "step": 3091
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.621157361422858e-05,
      "loss": 0.1076,
      "step": 3092
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.616618864010001e-05,
      "loss": 0.1133,
      "step": 3093
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.6120861696290712e-05,
      "loss": 0.1729,
      "step": 3094
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.607559281417631e-05,
      "loss": 0.1073,
      "step": 3095
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.603038202509224e-05,
      "loss": 0.0782,
      "step": 3096
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5985229360333708e-05,
      "loss": 0.0467,
      "step": 3097
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5940134851155697e-05,
      "loss": 0.0279,
      "step": 3098
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5895098528772888e-05,
      "loss": 0.1307,
      "step": 3099
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.585012042435977e-05,
      "loss": 0.061,
      "step": 3100
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5805200569050428e-05,
      "loss": 0.1165,
      "step": 3101
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5760338993938716e-05,
      "loss": 0.0756,
      "step": 3102
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.5715535730078112e-05,
      "loss": 0.1128,
      "step": 3103
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.567079080848173e-05,
      "loss": 0.0791,
      "step": 3104
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5626104260122287e-05,
      "loss": 0.0741,
      "step": 3105
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.558147611593216e-05,
      "loss": 0.1446,
      "step": 3106
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5536906406803185e-05,
      "loss": 0.0599,
      "step": 3107
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5492395163586847e-05,
      "loss": 0.0646,
      "step": 3108
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5447942417094097e-05,
      "loss": 0.067,
      "step": 3109
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.540354819809542e-05,
      "loss": 0.0796,
      "step": 3110
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.535921253732081e-05,
      "loss": 0.0422,
      "step": 3111
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.531493546545969e-05,
      "loss": 0.0503,
      "step": 3112
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5270717013160973e-05,
      "loss": 0.0701,
      "step": 3113
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.522655721103291e-05,
      "loss": 0.0653,
      "step": 3114
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5182456089643249e-05,
      "loss": 0.1385,
      "step": 3115
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5138413679519081e-05,
      "loss": 0.1103,
      "step": 3116
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5094430011146821e-05,
      "loss": 0.0791,
      "step": 3117
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.505050511497227e-05,
      "loss": 0.0787,
      "step": 3118
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.5006639021400538e-05,
      "loss": 0.0637,
      "step": 3119
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.4962831760796048e-05,
      "loss": 0.1394,
      "step": 3120
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.4919083363482422e-05,
      "loss": 0.0425,
      "step": 3121
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.4875393859742625e-05,
      "loss": 0.0764,
      "step": 3122
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4831763279818811e-05,
      "loss": 0.1106,
      "step": 3123
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4788191653912375e-05,
      "loss": 0.058,
      "step": 3124
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4744679012183882e-05,
      "loss": 0.0827,
      "step": 3125
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4701225384753037e-05,
      "loss": 0.0464,
      "step": 3126
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4657830801698769e-05,
      "loss": 0.0636,
      "step": 3127
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4614495293059061e-05,
      "loss": 0.0335,
      "step": 3128
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4571218888831062e-05,
      "loss": 0.092,
      "step": 3129
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4528001618970966e-05,
      "loss": 0.02,
      "step": 3130
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4484843513394076e-05,
      "loss": 0.0708,
      "step": 3131
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4441744601974704e-05,
      "loss": 0.0653,
      "step": 3132
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4398704914546234e-05,
      "loss": 0.0866,
      "step": 3133
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.435572448090099e-05,
      "loss": 0.0573,
      "step": 3134
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4312803330790302e-05,
      "loss": 0.1042,
      "step": 3135
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4269941493924487e-05,
      "loss": 0.067,
      "step": 3136
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.42271389999728e-05,
      "loss": 0.0378,
      "step": 3137
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4184395878563395e-05,
      "loss": 0.1077,
      "step": 3138
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4141712159283372e-05,
      "loss": 0.0842,
      "step": 3139
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4099087871678673e-05,
      "loss": 0.1559,
      "step": 3140
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.4056523045254088e-05,
      "loss": 0.0507,
      "step": 3141
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.4014017709473314e-05,
      "loss": 0.0401,
      "step": 3142
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3971571893758784e-05,
      "loss": 0.101,
      "step": 3143
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3929185627491792e-05,
      "loss": 0.0535,
      "step": 3144
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3886858940012382e-05,
      "loss": 0.0417,
      "step": 3145
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3844591860619383e-05,
      "loss": 0.1112,
      "step": 3146
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3802384418570358e-05,
      "loss": 0.1215,
      "step": 3147
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3760236643081536e-05,
      "loss": 0.1369,
      "step": 3148
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3718148563327914e-05,
      "loss": 0.1723,
      "step": 3149
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3676120208443166e-05,
      "loss": 0.0736,
      "step": 3150
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.363415160751953e-05,
      "loss": 0.0466,
      "step": 3151
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3592242789608001e-05,
      "loss": 0.0978,
      "step": 3152
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3550393783718118e-05,
      "loss": 0.08,
      "step": 3153
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3508604618818065e-05,
      "loss": 0.1544,
      "step": 3154
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.346687532383455e-05,
      "loss": 0.0351,
      "step": 3155
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3425205927652872e-05,
      "loss": 0.1209,
      "step": 3156
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3383596459116876e-05,
      "loss": 0.0793,
      "step": 3157
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3342046947028896e-05,
      "loss": 0.1064,
      "step": 3158
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3300557420149818e-05,
      "loss": 0.0385,
      "step": 3159
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.3259127907198921e-05,
      "loss": 0.0515,
      "step": 3160
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3217758436854011e-05,
      "loss": 0.0358,
      "step": 3161
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3176449037751293e-05,
      "loss": 0.0572,
      "step": 3162
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3135199738485416e-05,
      "loss": 0.0527,
      "step": 3163
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3094010567609417e-05,
      "loss": 0.0696,
      "step": 3164
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3052881553634722e-05,
      "loss": 0.0425,
      "step": 3165
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.3011812725031091e-05,
      "loss": 0.2475,
      "step": 3166
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2970804110226675e-05,
      "loss": 0.1476,
      "step": 3167
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2929855737607876e-05,
      "loss": 0.0925,
      "step": 3168
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2888967635519411e-05,
      "loss": 0.0458,
      "step": 3169
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2848139832264328e-05,
      "loss": 0.053,
      "step": 3170
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2807372356103887e-05,
      "loss": 0.0375,
      "step": 3171
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2766665235257613e-05,
      "loss": 0.0538,
      "step": 3172
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2726018497903236e-05,
      "loss": 0.0901,
      "step": 3173
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2685432172176726e-05,
      "loss": 0.0729,
      "step": 3174
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2644906286172187e-05,
      "loss": 0.0599,
      "step": 3175
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.260444086794188e-05,
      "loss": 0.065,
      "step": 3176
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2564035945496256e-05,
      "loss": 0.0646,
      "step": 3177
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2523691546803873e-05,
      "loss": 0.139,
      "step": 3178
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.2483407699791395e-05,
      "loss": 0.0775,
      "step": 3179
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2443184432343557e-05,
      "loss": 0.0424,
      "step": 3180
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2403021772303191e-05,
      "loss": 0.136,
      "step": 3181
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2362919747471124e-05,
      "loss": 0.1193,
      "step": 3182
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2322878385606262e-05,
      "loss": 0.1205,
      "step": 3183
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2282897714425524e-05,
      "loss": 0.0492,
      "step": 3184
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2242977761603747e-05,
      "loss": 0.156,
      "step": 3185
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2203118554773806e-05,
      "loss": 0.0596,
      "step": 3186
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2163320121526511e-05,
      "loss": 0.0967,
      "step": 3187
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2123582489410624e-05,
      "loss": 0.0493,
      "step": 3188
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.208390568593275e-05,
      "loss": 0.0932,
      "step": 3189
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2044289738557469e-05,
      "loss": 0.0417,
      "step": 3190
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.2004734674707186e-05,
      "loss": 0.0511,
      "step": 3191
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.196524052176221e-05,
      "loss": 0.0848,
      "step": 3192
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.192580730706061e-05,
      "loss": 0.1065,
      "step": 3193
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.1886435057898337e-05,
      "loss": 0.1052,
      "step": 3194
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.1847123801529148e-05,
      "loss": 0.1746,
      "step": 3195
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.1807873565164506e-05,
      "loss": 0.116,
      "step": 3196
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.1768684375973715e-05,
      "loss": 0.0892,
      "step": 3197
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.1729556261083784e-05,
      "loss": 0.0575,
      "step": 3198
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.169048924757945e-05,
      "loss": 0.0494,
      "step": 3199
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1651483362503168e-05,
      "loss": 0.1183,
      "step": 3200
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1612538632855086e-05,
      "loss": 0.0634,
      "step": 3201
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1573655085592982e-05,
      "loss": 0.0871,
      "step": 3202
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1534832747632285e-05,
      "loss": 0.0231,
      "step": 3203
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1496071645846096e-05,
      "loss": 0.0942,
      "step": 3204
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1457371807065098e-05,
      "loss": 0.1288,
      "step": 3205
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1418733258077596e-05,
      "loss": 0.1094,
      "step": 3206
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.138015602562943e-05,
      "loss": 0.209,
      "step": 3207
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1341640136424047e-05,
      "loss": 0.0783,
      "step": 3208
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1303185617122369e-05,
      "loss": 0.0696,
      "step": 3209
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1264792494342857e-05,
      "loss": 0.0999,
      "step": 3210
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1226460794661509e-05,
      "loss": 0.097,
      "step": 3211
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1188190544611777e-05,
      "loss": 0.0734,
      "step": 3212
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.114998177068457e-05,
      "loss": 0.0594,
      "step": 3213
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1111834499328266e-05,
      "loss": 0.1071,
      "step": 3214
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1073748756948688e-05,
      "loss": 0.056,
      "step": 3215
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.1035724569908979e-05,
      "loss": 0.0246,
      "step": 3216
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.0997761964529773e-05,
      "loss": 0.0515,
      "step": 3217
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0959860967089041e-05,
      "loss": 0.0511,
      "step": 3218
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0922021603822064e-05,
      "loss": 0.0543,
      "step": 3219
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0884243900921532e-05,
      "loss": 0.0477,
      "step": 3220
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0846527884537415e-05,
      "loss": 0.0603,
      "step": 3221
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0808873580777002e-05,
      "loss": 0.0575,
      "step": 3222
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0771281015704825e-05,
      "loss": 0.059,
      "step": 3223
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.073375021534272e-05,
      "loss": 0.0647,
      "step": 3224
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0696281205669756e-05,
      "loss": 0.1027,
      "step": 3225
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0658874012622244e-05,
      "loss": 0.0794,
      "step": 3226
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.062152866209366e-05,
      "loss": 0.0513,
      "step": 3227
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0584245179934715e-05,
      "loss": 0.0686,
      "step": 3228
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0547023591953309e-05,
      "loss": 0.0573,
      "step": 3229
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.050986392391442e-05,
      "loss": 0.0193,
      "step": 3230
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0472766201540252e-05,
      "loss": 0.0772,
      "step": 3231
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0435730450510083e-05,
      "loss": 0.0275,
      "step": 3232
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.039875669646031e-05,
      "loss": 0.0799,
      "step": 3233
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0361844964984435e-05,
      "loss": 0.0337,
      "step": 3234
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0324995281632954e-05,
      "loss": 0.058,
      "step": 3235
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.0288207671913497e-05,
      "loss": 0.0582,
      "step": 3236
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.025148216129066e-05,
      "loss": 0.0468,
      "step": 3237
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.0214818775186108e-05,
      "loss": 0.0553,
      "step": 3238
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.0178217538978474e-05,
      "loss": 0.0896,
      "step": 3239
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.014167847800338e-05,
      "loss": 0.1495,
      "step": 3240
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.0105201617553406e-05,
      "loss": 0.0283,
      "step": 3241
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.0068786982878087e-05,
      "loss": 0.061,
      "step": 3242
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.003243459918386e-05,
      "loss": 0.0643,
      "step": 3243
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.996144491634063e-06,
      "loss": 0.0342,
      "step": 3244
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.95991668534898e-06,
      "loss": 0.0479,
      "step": 3245
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.92375120540573e-06,
      "loss": 0.0555,
      "step": 3246
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.887648076838285e-06,
      "loss": 0.0787,
      "step": 3247
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.851607324637501e-06,
      "loss": 0.0745,
      "step": 3248
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.815628973751012e-06,
      "loss": 0.0485,
      "step": 3249
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.77971304908325e-06,
      "loss": 0.0517,
      "step": 3250
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.743859575495495e-06,
      "loss": 0.1209,
      "step": 3251
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.708068577805717e-06,
      "loss": 0.0835,
      "step": 3252
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.6723400807887e-06,
      "loss": 0.0498,
      "step": 3253
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.636674109175936e-06,
      "loss": 0.032,
      "step": 3254
    },
    {
      "epoch": 1.7,
      "learning_rate": 9.601070687655667e-06,
      "loss": 0.0963,
      "step": 3255
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.565529840872822e-06,
      "loss": 0.0429,
      "step": 3256
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.530051593428979e-06,
      "loss": 0.0911,
      "step": 3257
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.494635969882426e-06,
      "loss": 0.0304,
      "step": 3258
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.459282994748097e-06,
      "loss": 0.0264,
      "step": 3259
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.423992692497584e-06,
      "loss": 0.0576,
      "step": 3260
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.388765087559015e-06,
      "loss": 0.0876,
      "step": 3261
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.353600204317203e-06,
      "loss": 0.0665,
      "step": 3262
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.318498067113524e-06,
      "loss": 0.0964,
      "step": 3263
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.283458700245873e-06,
      "loss": 0.0389,
      "step": 3264
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.248482127968749e-06,
      "loss": 0.0555,
      "step": 3265
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.213568374493175e-06,
      "loss": 0.0879,
      "step": 3266
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.178717463986675e-06,
      "loss": 0.0345,
      "step": 3267
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.143929420573293e-06,
      "loss": 0.1024,
      "step": 3268
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.109204268333516e-06,
      "loss": 0.072,
      "step": 3269
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.074542031304367e-06,
      "loss": 0.1063,
      "step": 3270
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.039942733479224e-06,
      "loss": 0.0558,
      "step": 3271
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.00540639880798e-06,
      "loss": 0.0721,
      "step": 3272
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.970933051196905e-06,
      "loss": 0.0591,
      "step": 3273
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.936522714508678e-06,
      "loss": 0.1078,
      "step": 3274
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.902175412562375e-06,
      "loss": 0.128,
      "step": 3275
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.867891169133436e-06,
      "loss": 0.031,
      "step": 3276
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.83367000795362e-06,
      "loss": 0.1051,
      "step": 3277
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.799511952711025e-06,
      "loss": 0.0696,
      "step": 3278
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.765417027050105e-06,
      "loss": 0.0753,
      "step": 3279
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.73138525457159e-06,
      "loss": 0.1053,
      "step": 3280
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.697416658832492e-06,
      "loss": 0.1133,
      "step": 3281
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.663511263346114e-06,
      "loss": 0.0908,
      "step": 3282
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.629669091582005e-06,
      "loss": 0.0397,
      "step": 3283
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.595890166965892e-06,
      "loss": 0.0833,
      "step": 3284
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.562174512879828e-06,
      "loss": 0.0599,
      "step": 3285
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.528522152661966e-06,
      "loss": 0.1267,
      "step": 3286
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.494933109606706e-06,
      "loss": 0.0365,
      "step": 3287
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.461407406964605e-06,
      "loss": 0.0373,
      "step": 3288
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.427945067942389e-06,
      "loss": 0.0403,
      "step": 3289
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.394546115702928e-06,
      "loss": 0.162,
      "step": 3290
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.361210573365141e-06,
      "loss": 0.0469,
      "step": 3291
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.327938464004148e-06,
      "loss": 0.049,
      "step": 3292
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.294729810651137e-06,
      "loss": 0.0355,
      "step": 3293
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.261584636293329e-06,
      "loss": 0.0463,
      "step": 3294
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.22850296387403e-06,
      "loss": 0.0336,
      "step": 3295
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.19548481629262e-06,
      "loss": 0.0525,
      "step": 3296
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.162530216404474e-06,
      "loss": 0.0311,
      "step": 3297
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.129639187020954e-06,
      "loss": 0.0881,
      "step": 3298
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.096811750909484e-06,
      "loss": 0.0476,
      "step": 3299
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.064047930793406e-06,
      "loss": 0.1941,
      "step": 3300
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.03134774935207e-06,
      "loss": 0.0366,
      "step": 3301
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.998711229220778e-06,
      "loss": 0.0235,
      "step": 3302
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.966138392990696e-06,
      "loss": 0.0758,
      "step": 3303
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.933629263208986e-06,
      "loss": 0.1438,
      "step": 3304
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.901183862378692e-06,
      "loss": 0.1006,
      "step": 3305
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.868802212958703e-06,
      "loss": 0.1277,
      "step": 3306
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.836484337363836e-06,
      "loss": 0.1044,
      "step": 3307
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.804230257964739e-06,
      "loss": 0.0382,
      "step": 3308
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.772039997087888e-06,
      "loss": 0.1703,
      "step": 3309
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.739913577015612e-06,
      "loss": 0.0677,
      "step": 3310
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.707851019986012e-06,
      "loss": 0.0883,
      "step": 3311
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.675852348193013e-06,
      "loss": 0.0458,
      "step": 3312
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.643917583786298e-06,
      "loss": 0.0741,
      "step": 3313
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.612046748871327e-06,
      "loss": 0.134,
      "step": 3314
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.580239865509309e-06,
      "loss": 0.0562,
      "step": 3315
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.548496955717166e-06,
      "loss": 0.0144,
      "step": 3316
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.516818041467588e-06,
      "loss": 0.0487,
      "step": 3317
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.4852031446888685e-06,
      "loss": 0.0292,
      "step": 3318
    },
    {
      "epoch": 1.74,
      "eval_loss": 0.08197121322154999,
      "eval_runtime": 683.494,
      "eval_samples_per_second": 9.813,
      "eval_steps_per_second": 2.454,
      "step": 3318
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.453652287265112e-06,
      "loss": 0.1026,
      "step": 3319
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.42216549103597e-06,
      "loss": 0.088,
      "step": 3320
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.3907427777968444e-06,
      "loss": 0.0201,
      "step": 3321
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.359384169298744e-06,
      "loss": 0.0914,
      "step": 3322
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.3280896872483074e-06,
      "loss": 0.0427,
      "step": 3323
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.296859353307795e-06,
      "loss": 0.0812,
      "step": 3324
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.265693189095002e-06,
      "loss": 0.2095,
      "step": 3325
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.234591216183395e-06,
      "loss": 0.0749,
      "step": 3326
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.203553456101975e-06,
      "loss": 0.0628,
      "step": 3327
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.172579930335243e-06,
      "loss": 0.0596,
      "step": 3328
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.141670660323308e-06,
      "loss": 0.0282,
      "step": 3329
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.110825667461762e-06,
      "loss": 0.0906,
      "step": 3330
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.080044973101718e-06,
      "loss": 0.0823,
      "step": 3331
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.049328598549787e-06,
      "loss": 0.0699,
      "step": 3332
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.0186765650680166e-06,
      "loss": 0.0988,
      "step": 3333
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.988088893873967e-06,
      "loss": 0.0969,
      "step": 3334
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.957565606140626e-06,
      "loss": 0.0687,
      "step": 3335
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.927106722996424e-06,
      "loss": 0.116,
      "step": 3336
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.8967122655251845e-06,
      "loss": 0.1125,
      "step": 3337
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.866382254766157e-06,
      "loss": 0.0428,
      "step": 3338
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.836116711713991e-06,
      "loss": 0.0411,
      "step": 3339
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.805915657318673e-06,
      "loss": 0.0227,
      "step": 3340
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.775779112485569e-06,
      "loss": 0.0656,
      "step": 3341
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.745707098075426e-06,
      "loss": 0.0575,
      "step": 3342
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.71569963490426e-06,
      "loss": 0.1418,
      "step": 3343
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.685756743743476e-06,
      "loss": 0.084,
      "step": 3344
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.655878445319696e-06,
      "loss": 0.0968,
      "step": 3345
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.62606476031491e-06,
      "loss": 0.1741,
      "step": 3346
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.596315709366319e-06,
      "loss": 0.1094,
      "step": 3347
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.566631313066429e-06,
      "loss": 0.0391,
      "step": 3348
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.537011591962972e-06,
      "loss": 0.0441,
      "step": 3349
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.5074565665589115e-06,
      "loss": 0.0847,
      "step": 3350
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.477966257312451e-06,
      "loss": 0.1007,
      "step": 3351
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.448540684636972e-06,
      "loss": 0.1353,
      "step": 3352
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.419179868901038e-06,
      "loss": 0.0759,
      "step": 3353
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.3898838304284e-06,
      "loss": 0.0755,
      "step": 3354
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.360652589497973e-06,
      "loss": 0.1011,
      "step": 3355
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.331486166343814e-06,
      "loss": 0.0556,
      "step": 3356
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.302384581155119e-06,
      "loss": 0.1242,
      "step": 3357
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.273347854076195e-06,
      "loss": 0.0385,
      "step": 3358
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.244376005206476e-06,
      "loss": 0.0327,
      "step": 3359
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.215469054600431e-06,
      "loss": 0.0345,
      "step": 3360
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.186627022267677e-06,
      "loss": 0.017,
      "step": 3361
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.1578499281728316e-06,
      "loss": 0.0495,
      "step": 3362
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.129137792235607e-06,
      "loss": 0.0792,
      "step": 3363
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.100490634330725e-06,
      "loss": 0.049,
      "step": 3364
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.071908474287935e-06,
      "loss": 0.1064,
      "step": 3365
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.0433913318920346e-06,
      "loss": 0.0564,
      "step": 3366
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.0149392268827206e-06,
      "loss": 0.0459,
      "step": 3367
    },
    {
      "epoch": 1.76,
      "learning_rate": 5.986552178954763e-06,
      "loss": 0.0367,
      "step": 3368
    },
    {
      "epoch": 1.76,
      "learning_rate": 5.958230207757876e-06,
      "loss": 0.1501,
      "step": 3369
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.929973332896677e-06,
      "loss": 0.1465,
      "step": 3370
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.901781573930787e-06,
      "loss": 0.0689,
      "step": 3371
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.8736549503747185e-06,
      "loss": 0.073,
      "step": 3372
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.8455934816979305e-06,
      "loss": 0.063,
      "step": 3373
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.8175971873247195e-06,
      "loss": 0.0793,
      "step": 3374
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.789666086634338e-06,
      "loss": 0.0641,
      "step": 3375
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.7618001989608536e-06,
      "loss": 0.0464,
      "step": 3376
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.733999543593227e-06,
      "loss": 0.1741,
      "step": 3377
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.706264139775286e-06,
      "loss": 0.0308,
      "step": 3378
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.678594006705606e-06,
      "loss": 0.1263,
      "step": 3379
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.65098916353769e-06,
      "loss": 0.0535,
      "step": 3380
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.623449629379751e-06,
      "loss": 0.0717,
      "step": 3381
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.595975423294853e-06,
      "loss": 0.1053,
      "step": 3382
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.568566564300825e-06,
      "loss": 0.1411,
      "step": 3383
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.54122307137025e-06,
      "loss": 0.0494,
      "step": 3384
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.513944963430495e-06,
      "loss": 0.1074,
      "step": 3385
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.486732259363647e-06,
      "loss": 0.0435,
      "step": 3386
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.459584978006516e-06,
      "loss": 0.0432,
      "step": 3387
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.432503138150602e-06,
      "loss": 0.0766,
      "step": 3388
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.405486758542155e-06,
      "loss": 0.0472,
      "step": 3389
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.378535857882094e-06,
      "loss": 0.0853,
      "step": 3390
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.351650454826007e-06,
      "loss": 0.0385,
      "step": 3391
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.324830567984151e-06,
      "loss": 0.0336,
      "step": 3392
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.298076215921444e-06,
      "loss": 0.0974,
      "step": 3393
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.271387417157392e-06,
      "loss": 0.0526,
      "step": 3394
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.2447641901661845e-06,
      "loss": 0.0474,
      "step": 3395
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.218206553376581e-06,
      "loss": 0.0614,
      "step": 3396
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.1917145251719645e-06,
      "loss": 0.0542,
      "step": 3397
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.165288123890277e-06,
      "loss": 0.0807,
      "step": 3398
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.138927367824065e-06,
      "loss": 0.1249,
      "step": 3399
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.112632275220419e-06,
      "loss": 0.1191,
      "step": 3400
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.0864028642809595e-06,
      "loss": 0.0967,
      "step": 3401
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.060239153161872e-06,
      "loss": 0.03,
      "step": 3402
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.03414115997386e-06,
      "loss": 0.0828,
      "step": 3403
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.008108902782094e-06,
      "loss": 0.0941,
      "step": 3404
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.982142399606293e-06,
      "loss": 0.1169,
      "step": 3405
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.956241668420647e-06,
      "loss": 0.1473,
      "step": 3406
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.930406727153835e-06,
      "loss": 0.0408,
      "step": 3407
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.9046375936889215e-06,
      "loss": 0.1671,
      "step": 3408
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.8789342858635126e-06,
      "loss": 0.1752,
      "step": 3409
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.853296821469588e-06,
      "loss": 0.0924,
      "step": 3410
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.8277252182535874e-06,
      "loss": 0.0674,
      "step": 3411
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.802219493916327e-06,
      "loss": 0.0413,
      "step": 3412
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.776779666113051e-06,
      "loss": 0.1984,
      "step": 3413
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.751405752453375e-06,
      "loss": 0.0667,
      "step": 3414
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.7260977705012696e-06,
      "loss": 0.0551,
      "step": 3415
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.700855737775111e-06,
      "loss": 0.1013,
      "step": 3416
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.675679671747601e-06,
      "loss": 0.0895,
      "step": 3417
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.6505695898457655e-06,
      "loss": 0.0378,
      "step": 3418
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.625525509450967e-06,
      "loss": 0.037,
      "step": 3419
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.600547447898917e-06,
      "loss": 0.058,
      "step": 3420
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.575635422479574e-06,
      "loss": 0.0462,
      "step": 3421
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.550789450437187e-06,
      "loss": 0.1403,
      "step": 3422
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.526009548970312e-06,
      "loss": 0.1015,
      "step": 3423
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.501295735231781e-06,
      "loss": 0.0298,
      "step": 3424
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.4766480263286445e-06,
      "loss": 0.0637,
      "step": 3425
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.4520664393222e-06,
      "loss": 0.0798,
      "step": 3426
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.427550991228024e-06,
      "loss": 0.0464,
      "step": 3427
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.4031016990158435e-06,
      "loss": 0.0701,
      "step": 3428
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.378718579609598e-06,
      "loss": 0.0983,
      "step": 3429
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.354401649887485e-06,
      "loss": 0.0503,
      "step": 3430
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.330150926681809e-06,
      "loss": 0.0439,
      "step": 3431
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.305966426779118e-06,
      "loss": 0.0524,
      "step": 3432
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.281848166920055e-06,
      "loss": 0.0351,
      "step": 3433
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.257796163799455e-06,
      "loss": 0.1669,
      "step": 3434
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.2338104340662655e-06,
      "loss": 0.0781,
      "step": 3435
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.209890994323562e-06,
      "loss": 0.1524,
      "step": 3436
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.186037861128566e-06,
      "loss": 0.1042,
      "step": 3437
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.162251050992538e-06,
      "loss": 0.0461,
      "step": 3438
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.1385305803808834e-06,
      "loss": 0.062,
      "step": 3439
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.114876465713069e-06,
      "loss": 0.069,
      "step": 3440
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.091288723362641e-06,
      "loss": 0.0446,
      "step": 3441
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.06776736965716e-06,
      "loss": 0.0632,
      "step": 3442
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.044312420878294e-06,
      "loss": 0.0911,
      "step": 3443
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.020923893261686e-06,
      "loss": 0.0303,
      "step": 3444
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.997601802997064e-06,
      "loss": 0.1005,
      "step": 3445
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.9743461662281045e-06,
      "loss": 0.0387,
      "step": 3446
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.9511569990525185e-06,
      "loss": 0.1496,
      "step": 3447
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.928034317522011e-06,
      "loss": 0.1958,
      "step": 3448
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.9049781376422455e-06,
      "loss": 0.0738,
      "step": 3449
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.8819884753728665e-06,
      "loss": 0.0662,
      "step": 3450
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.859065346627466e-06,
      "loss": 0.0586,
      "step": 3451
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.836208767273597e-06,
      "loss": 0.1341,
      "step": 3452
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.813418753132725e-06,
      "loss": 0.0427,
      "step": 3453
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.790695319980264e-06,
      "loss": 0.122,
      "step": 3454
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.7680384835455105e-06,
      "loss": 0.0769,
      "step": 3455
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.7454482595116634e-06,
      "loss": 0.04,
      "step": 3456
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.7229246635158365e-06,
      "loss": 0.107,
      "step": 3457
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.7004677111490247e-06,
      "loss": 0.0648,
      "step": 3458
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.6780774179560716e-06,
      "loss": 0.105,
      "step": 3459
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.6557537994356795e-06,
      "loss": 0.0938,
      "step": 3460
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.6334968710404316e-06,
      "loss": 0.056,
      "step": 3461
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.6113066481766932e-06,
      "loss": 0.0999,
      "step": 3462
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.589183146204678e-06,
      "loss": 0.1504,
      "step": 3463
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.5671263804384345e-06,
      "loss": 0.0426,
      "step": 3464
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.545136366145796e-06,
      "loss": 0.1202,
      "step": 3465
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.5232131185484076e-06,
      "loss": 0.0761,
      "step": 3466
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.5013566528216744e-06,
      "loss": 0.1301,
      "step": 3467
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.4795669840948065e-06,
      "loss": 0.1045,
      "step": 3468
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.457844127450727e-06,
      "loss": 0.0876,
      "step": 3469
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.436188097926174e-06,
      "loss": 0.026,
      "step": 3470
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.414598910511557e-06,
      "loss": 0.0926,
      "step": 3471
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.3930765801510776e-06,
      "loss": 0.0294,
      "step": 3472
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.3716211217426407e-06,
      "loss": 0.0322,
      "step": 3473
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.350232550137833e-06,
      "loss": 0.0837,
      "step": 3474
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.328910880142e-06,
      "loss": 0.1519,
      "step": 3475
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.307656126514114e-06,
      "loss": 0.0782,
      "step": 3476
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.2864683039668607e-06,
      "loss": 0.1028,
      "step": 3477
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.2653474271665983e-06,
      "loss": 0.1071,
      "step": 3478
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.244293510733343e-06,
      "loss": 0.0781,
      "step": 3479
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.223306569240736e-06,
      "loss": 0.0375,
      "step": 3480
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.202386617216091e-06,
      "loss": 0.0544,
      "step": 3481
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.181533669140346e-06,
      "loss": 0.1185,
      "step": 3482
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.160747739448022e-06,
      "loss": 0.2692,
      "step": 3483
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.140028842527298e-06,
      "loss": 0.0377,
      "step": 3484
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.1193769927199248e-06,
      "loss": 0.0499,
      "step": 3485
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.0987922043212546e-06,
      "loss": 0.0894,
      "step": 3486
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.078274491580224e-06,
      "loss": 0.0479,
      "step": 3487
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.0578238686993057e-06,
      "loss": 0.0985,
      "step": 3488
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.037440349834575e-06,
      "loss": 0.0919,
      "step": 3489
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.017123949095624e-06,
      "loss": 0.0812,
      "step": 3490
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.996874680545603e-06,
      "loss": 0.0924,
      "step": 3491
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.9766925582011997e-06,
      "loss": 0.0833,
      "step": 3492
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.956577596032606e-06,
      "loss": 0.071,
      "step": 3493
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.9365298079635283e-06,
      "loss": 0.0538,
      "step": 3494
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.9165492078712e-06,
      "loss": 0.091,
      "step": 3495
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.896635809586301e-06,
      "loss": 0.1289,
      "step": 3496
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.876789626893017e-06,
      "loss": 0.0486,
      "step": 3497
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.857010673529015e-06,
      "loss": 0.1509,
      "step": 3498
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.8372989631854084e-06,
      "loss": 0.1302,
      "step": 3499
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.8176545095067842e-06,
      "loss": 0.0478,
      "step": 3500
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.798077326091175e-06,
      "loss": 0.1503,
      "step": 3501
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.7785674264900197e-06,
      "loss": 0.0877,
      "step": 3502
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.7591248242081925e-06,
      "loss": 0.1076,
      "step": 3503
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.739749532704017e-06,
      "loss": 0.0715,
      "step": 3504
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.7204415653891758e-06,
      "loss": 0.0998,
      "step": 3505
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.701200935628767e-06,
      "loss": 0.0303,
      "step": 3506
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.682027656741304e-06,
      "loss": 0.1178,
      "step": 3507
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.6629217419986475e-06,
      "loss": 0.0527,
      "step": 3508
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.643883204626041e-06,
      "loss": 0.0457,
      "step": 3509
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.624912057802087e-06,
      "loss": 0.0974,
      "step": 3510
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.6060083146587257e-06,
      "loss": 0.0189,
      "step": 3511
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.587171988281256e-06,
      "loss": 0.0331,
      "step": 3512
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.568403091708338e-06,
      "loss": 0.0567,
      "step": 3513
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.5497016379318894e-06,
      "loss": 0.0658,
      "step": 3514
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.531067639897189e-06,
      "loss": 0.0534,
      "step": 3515
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.5125011105028293e-06,
      "loss": 0.0529,
      "step": 3516
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.4940020626006643e-06,
      "loss": 0.0663,
      "step": 3517
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.4755705089958615e-06,
      "loss": 0.1601,
      "step": 3518
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.4572064624468818e-06,
      "loss": 0.1391,
      "step": 3519
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.4389099356654234e-06,
      "loss": 0.1348,
      "step": 3520
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.4206809413164665e-06,
      "loss": 0.0694,
      "step": 3521
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.4025194920182403e-06,
      "loss": 0.085,
      "step": 3522
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.3844256003422438e-06,
      "loss": 0.1084,
      "step": 3523
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.3663992788131474e-06,
      "loss": 0.0633,
      "step": 3524
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.348440539908903e-06,
      "loss": 0.0298,
      "step": 3525
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.3305493960606882e-06,
      "loss": 0.1022,
      "step": 3526
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.3127258596528534e-06,
      "loss": 0.1011,
      "step": 3527
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.294969943022973e-06,
      "loss": 0.1254,
      "step": 3528
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.2772816584618273e-06,
      "loss": 0.117,
      "step": 3529
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.259661018213333e-06,
      "loss": 0.0331,
      "step": 3530
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.2421080344746436e-06,
      "loss": 0.0522,
      "step": 3531
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.224622719396019e-06,
      "loss": 0.106,
      "step": 3532
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.2072050850809322e-06,
      "loss": 0.0874,
      "step": 3533
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.1898551435859838e-06,
      "loss": 0.2156,
      "step": 3534
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.1725729069209e-06,
      "loss": 0.0439,
      "step": 3535
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.1553583870486003e-06,
      "loss": 0.1438,
      "step": 3536
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.1382115958850422e-06,
      "loss": 0.0583,
      "step": 3537
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.121132545299376e-06,
      "loss": 0.084,
      "step": 3538
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.1041212471138217e-06,
      "loss": 0.0617,
      "step": 3539
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0871777131037163e-06,
      "loss": 0.1234,
      "step": 3540
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.070301954997489e-06,
      "loss": 0.0537,
      "step": 3541
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.05349398447664e-06,
      "loss": 0.1546,
      "step": 3542
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0367538131757845e-06,
      "loss": 0.1139,
      "step": 3543
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.020081452682554e-06,
      "loss": 0.1057,
      "step": 3544
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.003476914537672e-06,
      "loss": 0.0941,
      "step": 3545
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.986940210234922e-06,
      "loss": 0.1273,
      "step": 3546
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.9704713512211035e-06,
      "loss": 0.0912,
      "step": 3547
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.9540703488960754e-06,
      "loss": 0.0716,
      "step": 3548
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.9377372146127227e-06,
      "loss": 0.0908,
      "step": 3549
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.921471959676957e-06,
      "loss": 0.0853,
      "step": 3550
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.9052745953476836e-06,
      "loss": 0.0817,
      "step": 3551
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.8891451328368227e-06,
      "loss": 0.0514,
      "step": 3552
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.8730835833093097e-06,
      "loss": 0.0657,
      "step": 3553
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.8570899578830292e-06,
      "loss": 0.0462,
      "step": 3554
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.8411642676289143e-06,
      "loss": 0.0575,
      "step": 3555
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.8253065235708021e-06,
      "loss": 0.0809,
      "step": 3556
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.8095167366855348e-06,
      "loss": 0.0605,
      "step": 3557
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.7937949179029134e-06,
      "loss": 0.0891,
      "step": 3558
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.7781410781056772e-06,
      "loss": 0.0292,
      "step": 3559
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.7625552281295254e-06,
      "loss": 0.09,
      "step": 3560
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.7470373787630833e-06,
      "loss": 0.0983,
      "step": 3561
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.7315875407479032e-06,
      "loss": 0.0693,
      "step": 3562
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.7162057247784857e-06,
      "loss": 0.0508,
      "step": 3563
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.7008919415021918e-06,
      "loss": 0.1085,
      "step": 3564
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.685646201519353e-06,
      "loss": 0.0541,
      "step": 3565
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6704685153831612e-06,
      "loss": 0.0368,
      "step": 3566
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6553588935997012e-06,
      "loss": 0.0608,
      "step": 3567
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.640317346627973e-06,
      "loss": 0.093,
      "step": 3568
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6253438848798153e-06,
      "loss": 0.0756,
      "step": 3569
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.610438518719981e-06,
      "loss": 0.1153,
      "step": 3570
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.5956012584660506e-06,
      "loss": 0.0701,
      "step": 3571
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.5808321143884641e-06,
      "loss": 0.0467,
      "step": 3572
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.5661310967105324e-06,
      "loss": 0.0542,
      "step": 3573
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.5514982156083936e-06,
      "loss": 0.1305,
      "step": 3574
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.5369334812110224e-06,
      "loss": 0.0451,
      "step": 3575
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.5224369036002217e-06,
      "loss": 0.0347,
      "step": 3576
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.5080084928106308e-06,
      "loss": 0.0272,
      "step": 3577
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.493648258829694e-06,
      "loss": 0.0897,
      "step": 3578
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.479356211597649e-06,
      "loss": 0.0325,
      "step": 3579
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.4651323610075485e-06,
      "loss": 0.0984,
      "step": 3580
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.450976716905228e-06,
      "loss": 0.0861,
      "step": 3581
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.436889289089316e-06,
      "loss": 0.0656,
      "step": 3582
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.4228700873112454e-06,
      "loss": 0.0288,
      "step": 3583
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.4089191212751872e-06,
      "loss": 0.0522,
      "step": 3584
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.3950364006380946e-06,
      "loss": 0.0567,
      "step": 3585
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.381221935009669e-06,
      "loss": 0.2253,
      "step": 3586
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.3674757339523947e-06,
      "loss": 0.0737,
      "step": 3587
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.3537978069814716e-06,
      "loss": 0.03,
      "step": 3588
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.3401881635648372e-06,
      "loss": 0.0917,
      "step": 3589
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.3266468131232001e-06,
      "loss": 0.0991,
      "step": 3590
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.3131737650299625e-06,
      "loss": 0.1079,
      "step": 3591
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.299769028611264e-06,
      "loss": 0.1456,
      "step": 3592
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2864326131459382e-06,
      "loss": 0.0526,
      "step": 3593
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2731645278655445e-06,
      "loss": 0.0536,
      "step": 3594
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.259964781954348e-06,
      "loss": 0.0287,
      "step": 3595
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.2468333845492952e-06,
      "loss": 0.0455,
      "step": 3596
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2337703447400262e-06,
      "loss": 0.0817,
      "step": 3597
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2207756715688745e-06,
      "loss": 0.0332,
      "step": 3598
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.2078493740308227e-06,
      "loss": 0.045,
      "step": 3599
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1949914610735358e-06,
      "loss": 0.1771,
      "step": 3600
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1822019415973607e-06,
      "loss": 0.0623,
      "step": 3601
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1694808244552824e-06,
      "loss": 0.0936,
      "step": 3602
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1568281184529462e-06,
      "loss": 0.0721,
      "step": 3603
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.144243832348635e-06,
      "loss": 0.0367,
      "step": 3604
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1317279748532694e-06,
      "loss": 0.0725,
      "step": 3605
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1192805546304307e-06,
      "loss": 0.1185,
      "step": 3606
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.1069015802962713e-06,
      "loss": 0.0455,
      "step": 3607
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.0945910604196252e-06,
      "loss": 0.109,
      "step": 3608
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.0823490035218987e-06,
      "loss": 0.044,
      "step": 3609
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.0701754180771462e-06,
      "loss": 0.057,
      "step": 3610
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.0580703125119829e-06,
      "loss": 0.0358,
      "step": 3611
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.0460336952056615e-06,
      "loss": 0.0996,
      "step": 3612
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.0340655744899953e-06,
      "loss": 0.0368,
      "step": 3613
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.0221659586494014e-06,
      "loss": 0.0629,
      "step": 3614
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.0103348559208692e-06,
      "loss": 0.0363,
      "step": 3615
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.985722744939696e-07,
      "loss": 0.0597,
      "step": 3616
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.868782225108342e-07,
      "loss": 0.039,
      "step": 3617
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.752527080661655e-07,
      "loss": 0.0716,
      "step": 3618
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.636957392072377e-07,
      "loss": 0.0507,
      "step": 3619
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.522073239338292e-07,
      "loss": 0.0448,
      "step": 3620
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.407874701983233e-07,
      "loss": 0.0712,
      "step": 3621
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.294361859056188e-07,
      "loss": 0.0437,
      "step": 3622
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.181534789131307e-07,
      "loss": 0.0433,
      "step": 3623
    },
    {
      "epoch": 1.9,
      "learning_rate": 9.069393570308227e-07,
      "loss": 0.0388,
      "step": 3624
    },
    {
      "epoch": 1.9,
      "learning_rate": 8.957938280212075e-07,
      "loss": 0.0809,
      "step": 3625
    },
    {
      "epoch": 1.9,
      "learning_rate": 8.847168995992916e-07,
      "loss": 0.0566,
      "step": 3626
    },
    {
      "epoch": 1.9,
      "learning_rate": 8.737085794325751e-07,
      "loss": 0.0466,
      "step": 3627
    },
    {
      "epoch": 1.9,
      "learning_rate": 8.627688751411067e-07,
      "loss": 0.0559,
      "step": 3628
    },
    {
      "epoch": 1.9,
      "learning_rate": 8.518977942974071e-07,
      "loss": 0.1481,
      "step": 3629
    },
    {
      "epoch": 1.9,
      "learning_rate": 8.410953444265346e-07,
      "loss": 0.0641,
      "step": 3630
    },
    {
      "epoch": 1.9,
      "learning_rate": 8.303615330060077e-07,
      "loss": 0.0657,
      "step": 3631
    },
    {
      "epoch": 1.9,
      "learning_rate": 8.196963674658276e-07,
      "loss": 0.1832,
      "step": 3632
    },
    {
      "epoch": 1.9,
      "learning_rate": 8.09099855188522e-07,
      "loss": 0.0298,
      "step": 3633
    },
    {
      "epoch": 1.9,
      "learning_rate": 7.985720035090239e-07,
      "loss": 0.0576,
      "step": 3634
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.881128197148146e-07,
      "loss": 0.0858,
      "step": 3635
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.777223110458032e-07,
      "loss": 0.1094,
      "step": 3636
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.674004846943695e-07,
      "loss": 0.0716,
      "step": 3637
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.571473478053426e-07,
      "loss": 0.0515,
      "step": 3638
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.469629074760231e-07,
      "loss": 0.0913,
      "step": 3639
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.368471707561497e-07,
      "loss": 0.0553,
      "step": 3640
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.268001446478767e-07,
      "loss": 0.1207,
      "step": 3641
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.16821836105841e-07,
      "loss": 0.0783,
      "step": 3642
    },
    {
      "epoch": 1.91,
      "learning_rate": 7.069122520370952e-07,
      "loss": 0.1211,
      "step": 3643
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.970713993011191e-07,
      "loss": 0.0455,
      "step": 3644
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.872992847098081e-07,
      "loss": 0.048,
      "step": 3645
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.775959150275068e-07,
      "loss": 0.0664,
      "step": 3646
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.679612969709204e-07,
      "loss": 0.0418,
      "step": 3647
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.583954372092138e-07,
      "loss": 0.0542,
      "step": 3648
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.48898342363935e-07,
      "loss": 0.0778,
      "step": 3649
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.394700190090253e-07,
      "loss": 0.0605,
      "step": 3650
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.301104736708529e-07,
      "loss": 0.1345,
      "step": 3651
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.208197128281467e-07,
      "loss": 0.0982,
      "step": 3652
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.11597742912029e-07,
      "loss": 0.0226,
      "step": 3653
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.024445703060156e-07,
      "loss": 0.0733,
      "step": 3654
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.933602013459937e-07,
      "loss": 0.1563,
      "step": 3655
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.843446423202115e-07,
      "loss": 0.0744,
      "step": 3656
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.75397899469321e-07,
      "loss": 0.0252,
      "step": 3657
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.665199789862907e-07,
      "loss": 0.1094,
      "step": 3658
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.577108870164827e-07,
      "loss": 0.0342,
      "step": 3659
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.489706296576303e-07,
      "loss": 0.0889,
      "step": 3660
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.402992129597717e-07,
      "loss": 0.0475,
      "step": 3661
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.316966429253278e-07,
      "loss": 0.0824,
      "step": 3662
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.231629255090686e-07,
      "loss": 0.0674,
      "step": 3663
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.146980666180689e-07,
      "loss": 0.0622,
      "step": 3664
    },
    {
      "epoch": 1.92,
      "learning_rate": 5.063020721117639e-07,
      "loss": 0.0752,
      "step": 3665
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.979749478019157e-07,
      "loss": 0.0634,
      "step": 3666
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.897166994526358e-07,
      "loss": 0.0292,
      "step": 3667
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.815273327803182e-07,
      "loss": 0.0692,
      "step": 3668
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.7340685345369506e-07,
      "loss": 0.1244,
      "step": 3669
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.6535526709381437e-07,
      "loss": 0.0615,
      "step": 3670
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.573725792740624e-07,
      "loss": 0.0579,
      "step": 3671
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.4945879552007466e-07,
      "loss": 0.1333,
      "step": 3672
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.41613921309858e-07,
      "loss": 0.0666,
      "step": 3673
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.3383796207365766e-07,
      "loss": 0.1802,
      "step": 3674
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.261309231940569e-07,
      "loss": 0.0405,
      "step": 3675
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.1849281000592157e-07,
      "loss": 0.0591,
      "step": 3676
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.1092362779640013e-07,
      "loss": 0.0709,
      "step": 3677
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.034233818049349e-07,
      "loss": 0.0534,
      "step": 3678
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.959920772232284e-07,
      "loss": 0.0237,
      "step": 3679
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.8862971919531033e-07,
      "loss": 0.0487,
      "step": 3680
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.813363128174263e-07,
      "loss": 0.1228,
      "step": 3681
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.7411186313812687e-07,
      "loss": 0.0428,
      "step": 3682
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.669563751582339e-07,
      "loss": 0.1382,
      "step": 3683
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.598698538308076e-07,
      "loss": 0.0954,
      "step": 3684
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.52852304061202e-07,
      "loss": 0.1443,
      "step": 3685
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.459037307069979e-07,
      "loss": 0.0522,
      "step": 3686
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.39024138578059e-07,
      "loss": 0.0566,
      "step": 3687
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.3221353243646505e-07,
      "loss": 0.0515,
      "step": 3688
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.254719169965781e-07,
      "loss": 0.0348,
      "step": 3689
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.1879929692498757e-07,
      "loss": 0.1805,
      "step": 3690
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.121956768405099e-07,
      "loss": 0.0594,
      "step": 3691
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.05661061314233e-07,
      "loss": 0.0594,
      "step": 3692
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.991954548694609e-07,
      "loss": 0.0475,
      "step": 3693
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.927988619817135e-07,
      "loss": 0.0433,
      "step": 3694
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.8647128707876004e-07,
      "loss": 0.0957,
      "step": 3695
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.802127345405969e-07,
      "loss": 0.0657,
      "step": 3696
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.7402320869943656e-07,
      "loss": 0.0529,
      "step": 3697
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.6790271383970723e-07,
      "loss": 0.1004,
      "step": 3698
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.6185125419804226e-07,
      "loss": 0.2202,
      "step": 3699
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.558688339633131e-07,
      "loss": 0.0463,
      "step": 3700
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.4995545727660717e-07,
      "loss": 0.0394,
      "step": 3701
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.4411112823117255e-07,
      "loss": 0.0544,
      "step": 3702
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.3833585087251753e-07,
      "loss": 0.0423,
      "step": 3703
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.32629629198311e-07,
      "loss": 0.0259,
      "step": 3704
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.269924671584378e-07,
      "loss": 0.1347,
      "step": 3705
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.2142436865499882e-07,
      "loss": 0.0496,
      "step": 3706
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.1592533754224432e-07,
      "loss": 0.1763,
      "step": 3707
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.1049537762665162e-07,
      "loss": 0.069,
      "step": 3708
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.0513449266686968e-07,
      "loss": 0.1339,
      "step": 3709
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.9984268637374127e-07,
      "loss": 0.0642,
      "step": 3710
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.946199624102918e-07,
      "loss": 0.0475,
      "step": 3711
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8946632439172939e-07,
      "loss": 0.0503,
      "step": 3712
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8438177588542272e-07,
      "loss": 0.023,
      "step": 3713
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.7936632041094525e-07,
      "loss": 0.0497,
      "step": 3714
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.7441996144003104e-07,
      "loss": 0.073,
      "step": 3715
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.6954270239655235e-07,
      "loss": 0.0507,
      "step": 3716
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.647345466566197e-07,
      "loss": 0.1405,
      "step": 3717
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.5999549754843745e-07,
      "loss": 0.0635,
      "step": 3718
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.55325558352426e-07,
      "loss": 0.0433,
      "step": 3719
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.5072473230114404e-07,
      "loss": 0.0627,
      "step": 3720
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.4619302257931068e-07,
      "loss": 0.0178,
      "step": 3721
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.4173043232380557e-07,
      "loss": 0.0436,
      "step": 3722
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.3733696462366886e-07,
      "loss": 0.0903,
      "step": 3723
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.3301262252007895e-07,
      "loss": 0.0786,
      "step": 3724
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.2875740900638588e-07,
      "loss": 0.0832,
      "step": 3725
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.2457132702806684e-07,
      "loss": 0.0306,
      "step": 3726
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.204543794827595e-07,
      "loss": 0.0768,
      "step": 3727
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.1640656922023985e-07,
      "loss": 0.0827,
      "step": 3728
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1242789904243322e-07,
      "loss": 0.0552,
      "step": 3729
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.0851837170340329e-07,
      "loss": 0.1057,
      "step": 3730
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.0467798990934086e-07,
      "loss": 0.1174,
      "step": 3731
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.0090675631858614e-07,
      "loss": 0.1323,
      "step": 3732
    },
    {
      "epoch": 1.96,
      "learning_rate": 9.720467354161766e-08,
      "loss": 0.0432,
      "step": 3733
    },
    {
      "epoch": 1.96,
      "learning_rate": 9.357174414104108e-08,
      "loss": 0.0448,
      "step": 3734
    },
    {
      "epoch": 1.96,
      "learning_rate": 9.000797063158927e-08,
      "loss": 0.1076,
      "step": 3735
    },
    {
      "epoch": 1.96,
      "learning_rate": 8.651335548013339e-08,
      "loss": 0.0427,
      "step": 3736
    },
    {
      "epoch": 1.96,
      "learning_rate": 8.308790110567177e-08,
      "loss": 0.0918,
      "step": 3737
    },
    {
      "epoch": 1.96,
      "learning_rate": 7.973160987931883e-08,
      "loss": 0.0525,
      "step": 3738
    },
    {
      "epoch": 1.96,
      "learning_rate": 7.644448412432726e-08,
      "loss": 0.0795,
      "step": 3739
    },
    {
      "epoch": 1.96,
      "learning_rate": 7.322652611606584e-08,
      "loss": 0.1365,
      "step": 3740
    },
    {
      "epoch": 1.96,
      "learning_rate": 7.007773808201945e-08,
      "loss": 0.0354,
      "step": 3741
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.699812220182234e-08,
      "loss": 0.0262,
      "step": 3742
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.398768060718041e-08,
      "loss": 0.0497,
      "step": 3743
    },
    {
      "epoch": 1.96,
      "learning_rate": 6.104641538194899e-08,
      "loss": 0.0353,
      "step": 3744
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.8174328562099475e-08,
      "loss": 0.04,
      "step": 3745
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.537142213569713e-08,
      "loss": 0.1018,
      "step": 3746
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.2637698042934394e-08,
      "loss": 0.0754,
      "step": 3747
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.997315817613091e-08,
      "loss": 0.0942,
      "step": 3748
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.737780437967798e-08,
      "loss": 0.041,
      "step": 3749
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.4851638450094105e-08,
      "loss": 0.1508,
      "step": 3750
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.239466213602494e-08,
      "loss": 0.0338,
      "step": 3751
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.000687713819895e-08,
      "loss": 0.0808,
      "step": 3752
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.768828510946065e-08,
      "loss": 0.0302,
      "step": 3753
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.5438887654737355e-08,
      "loss": 0.0379,
      "step": 3754
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.3258686331105734e-08,
      "loss": 0.0646,
      "step": 3755
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.114768264769197e-08,
      "loss": 0.0873,
      "step": 3756
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.910587806574938e-08,
      "loss": 0.064,
      "step": 3757
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.71332739986363e-08,
      "loss": 0.0382,
      "step": 3758
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.5229871811804917e-08,
      "loss": 0.1315,
      "step": 3759
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.3395672822790204e-08,
      "loss": 0.0734,
      "step": 3760
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.1630678301254315e-08,
      "loss": 0.1075,
      "step": 3761
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.993488946891997e-08,
      "loss": 0.0385,
      "step": 3762
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.8308307499637077e-08,
      "loss": 0.0489,
      "step": 3763
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.6750933519327217e-08,
      "loss": 0.0688,
      "step": 3764
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.5262768606016943e-08,
      "loss": 0.0994,
      "step": 3765
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.3843813789826688e-08,
      "loss": 0.0321,
      "step": 3766
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.2494070052959661e-08,
      "loss": 0.1103,
      "step": 3767
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.1213538329735152e-08,
      "loss": 0.0534,
      "step": 3768
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.0002219506521915e-08,
      "loss": 0.0446,
      "step": 3769
    },
    {
      "epoch": 1.98,
      "learning_rate": 8.860114421826993e-09,
      "loss": 0.058,
      "step": 3770
    },
    {
      "epoch": 1.98,
      "learning_rate": 7.787223866195793e-09,
      "loss": 0.0448,
      "step": 3771
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.783548582312005e-09,
      "loss": 0.0854,
      "step": 3772
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.849089264919893e-09,
      "loss": 0.0517,
      "step": 3773
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.983846560857597e-09,
      "loss": 0.0891,
      "step": 3774
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.18782106905713e-09,
      "loss": 0.1169,
      "step": 3775
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.461013340522179e-09,
      "loss": 0.133,
      "step": 3776
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.8034238783614108e-09,
      "loss": 0.1037,
      "step": 3777
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.215053137755163e-09,
      "loss": 0.101,
      "step": 3778
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.6959015259887523e-09,
      "loss": 0.0822,
      "step": 3779
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.245969402419167e-09,
      "loss": 0.0619,
      "step": 3780
    },
    {
      "epoch": 1.98,
      "learning_rate": 8.65257078475068e-10,
      "loss": 0.0964,
      "step": 3781
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.53764817712299e-10,
      "loss": 0.0567,
      "step": 3782
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.114928357472735e-10,
      "loss": 0.0399,
      "step": 3783
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.384413002569751e-10,
      "loss": 0.1223,
      "step": 3784
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.461033105667255e-11,
      "loss": 0.1102,
      "step": 3785
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0,
      "loss": 0.0386,
      "step": 3786
    }
  ],
  "logging_steps": 1,
  "max_steps": 3786,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 1893,
  "total_flos": 2.1421268786743345e+19,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}