{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9992952783650458,
  "eval_steps": 500,
  "global_step": 709,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 4.5454545454545455e-06,
      "loss": 1.019,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.090909090909091e-06,
      "loss": 1.0036,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.3636363636363637e-05,
      "loss": 0.9864,
      "step": 3
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.8181818181818182e-05,
      "loss": 1.0294,
      "step": 4
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.272727272727273e-05,
      "loss": 1.0158,
      "step": 5
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.7272727272727273e-05,
      "loss": 1.005,
      "step": 6
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.181818181818182e-05,
      "loss": 0.999,
      "step": 7
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.6363636363636364e-05,
      "loss": 1.1149,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.0909090909090915e-05,
      "loss": 1.1047,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.545454545454546e-05,
      "loss": 1.1206,
      "step": 10
    },
    {
      "epoch": 0.02,
      "learning_rate": 5e-05,
      "loss": 1.1102,
      "step": 11
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.4545454545454546e-05,
      "loss": 1.1411,
      "step": 12
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.90909090909091e-05,
      "loss": 1.1526,
      "step": 13
    },
    {
      "epoch": 0.02,
      "learning_rate": 6.363636363636364e-05,
      "loss": 1.1397,
      "step": 14
    },
    {
      "epoch": 0.02,
      "learning_rate": 6.818181818181818e-05,
      "loss": 1.1834,
      "step": 15
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.272727272727273e-05,
      "loss": 1.2163,
      "step": 16
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.727272727272727e-05,
      "loss": 1.2801,
      "step": 17
    },
    {
      "epoch": 0.03,
      "learning_rate": 8.181818181818183e-05,
      "loss": 1.2758,
      "step": 18
    },
    {
      "epoch": 0.03,
      "learning_rate": 8.636363636363637e-05,
      "loss": 1.2541,
      "step": 19
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.090909090909092e-05,
      "loss": 1.2804,
      "step": 20
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.545454545454546e-05,
      "loss": 1.2869,
      "step": 21
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001,
      "loss": 1.261,
      "step": 22
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.999947721210493e-05,
      "loss": 1.2938,
      "step": 23
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.999790885935198e-05,
      "loss": 1.3089,
      "step": 24
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.999529497453782e-05,
      "loss": 1.3996,
      "step": 25
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.999163561232272e-05,
      "loss": 1.4033,
      "step": 26
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.998693084922947e-05,
      "loss": 1.4061,
      "step": 27
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.998118078364184e-05,
      "loss": 1.3929,
      "step": 28
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.997438553580241e-05,
      "loss": 1.4128,
      "step": 29
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.996654524781009e-05,
      "loss": 1.4063,
      "step": 30
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.995766008361719e-05,
      "loss": 1.4088,
      "step": 31
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.994773022902597e-05,
      "loss": 1.3899,
      "step": 32
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.993675589168473e-05,
      "loss": 1.3813,
      "step": 33
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.992473730108354e-05,
      "loss": 1.3964,
      "step": 34
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.99116747085493e-05,
      "loss": 1.3606,
      "step": 35
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.989756838724064e-05,
      "loss": 1.4069,
      "step": 36
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.988241863214211e-05,
      "loss": 1.4116,
      "step": 37
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.986622576005806e-05,
      "loss": 1.3307,
      "step": 38
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.984899010960601e-05,
      "loss": 1.3983,
      "step": 39
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.983071204120951e-05,
      "loss": 1.3725,
      "step": 40
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.981139193709068e-05,
      "loss": 1.4223,
      "step": 41
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.979103020126218e-05,
      "loss": 1.3844,
      "step": 42
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.976962725951878e-05,
      "loss": 1.4282,
      "step": 43
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.974718355942843e-05,
      "loss": 1.3989,
      "step": 44
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.972369957032293e-05,
      "loss": 1.393,
      "step": 45
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.969917578328808e-05,
      "loss": 1.4606,
      "step": 46
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.967361271115343e-05,
      "loss": 1.4384,
      "step": 47
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.964701088848158e-05,
      "loss": 1.478,
      "step": 48
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.961937087155697e-05,
      "loss": 1.4254,
      "step": 49
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.959069323837424e-05,
      "loss": 1.454,
      "step": 50
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.956097858862619e-05,
      "loss": 1.4448,
      "step": 51
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.953022754369114e-05,
      "loss": 1.4567,
      "step": 52
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.94984407466201e-05,
      "loss": 1.4563,
      "step": 53
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.946561886212315e-05,
      "loss": 1.438,
      "step": 54
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.943176257655567e-05,
      "loss": 1.4096,
      "step": 55
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.93968725979039e-05,
      "loss": 1.4472,
      "step": 56
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.936094965577017e-05,
      "loss": 1.4223,
      "step": 57
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.932399450135766e-05,
      "loss": 1.439,
      "step": 58
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.928600790745466e-05,
      "loss": 1.4606,
      "step": 59
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.924699066841845e-05,
      "loss": 1.5223,
      "step": 60
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.920694360015863e-05,
      "loss": 1.4791,
      "step": 61
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.916586754012008e-05,
      "loss": 1.4432,
      "step": 62
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.91237633472655e-05,
      "loss": 1.4596,
      "step": 63
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.908063190205738e-05,
      "loss": 1.4488,
      "step": 64
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.903647410643963e-05,
      "loss": 1.4542,
      "step": 65
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.899129088381866e-05,
      "loss": 1.443,
      "step": 66
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.894508317904419e-05,
      "loss": 1.5506,
      "step": 67
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.88978519583893e-05,
      "loss": 1.4963,
      "step": 68
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.884959820953048e-05,
      "loss": 1.4135,
      "step": 69
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.880032294152673e-05,
      "loss": 1.4813,
      "step": 70
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.875002718479858e-05,
      "loss": 1.4471,
      "step": 71
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.869871199110656e-05,
      "loss": 1.4165,
      "step": 72
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.864637843352915e-05,
      "loss": 1.4395,
      "step": 73
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.859302760644036e-05,
      "loss": 1.4375,
      "step": 74
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.853866062548687e-05,
      "loss": 1.4995,
      "step": 75
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.848327862756466e-05,
      "loss": 1.4465,
      "step": 76
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.842688277079523e-05,
      "loss": 1.4374,
      "step": 77
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.836947423450147e-05,
      "loss": 1.4476,
      "step": 78
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.831105421918287e-05,
      "loss": 1.437,
      "step": 79
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.825162394649048e-05,
      "loss": 1.41,
      "step": 80
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.819118465920143e-05,
      "loss": 1.469,
      "step": 81
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.812973762119281e-05,
      "loss": 1.4168,
      "step": 82
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.806728411741533e-05,
      "loss": 1.5309,
      "step": 83
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.800382545386641e-05,
      "loss": 1.4883,
      "step": 84
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.79393629575629e-05,
      "loss": 1.4679,
      "step": 85
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.787389797651334e-05,
      "loss": 1.5121,
      "step": 86
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.780743187968968e-05,
      "loss": 1.4483,
      "step": 87
    },
    {
      "epoch": 0.12,
      "learning_rate": 9.773996605699875e-05,
      "loss": 1.4756,
      "step": 88
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.767150191925321e-05,
      "loss": 1.4383,
      "step": 89
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.760204089814192e-05,
      "loss": 1.4457,
      "step": 90
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.753158444620013e-05,
      "loss": 1.4332,
      "step": 91
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.746013403677905e-05,
      "loss": 1.4486,
      "step": 92
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.738769116401505e-05,
      "loss": 1.4434,
      "step": 93
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.73142573427984e-05,
      "loss": 1.4616,
      "step": 94
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.723983410874163e-05,
      "loss": 1.4519,
      "step": 95
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.716442301814735e-05,
      "loss": 1.4913,
      "step": 96
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.708802564797578e-05,
      "loss": 1.4632,
      "step": 97
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.701064359581176e-05,
      "loss": 1.4475,
      "step": 98
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.693227847983126e-05,
      "loss": 1.4445,
      "step": 99
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.685293193876765e-05,
      "loss": 1.458,
      "step": 100
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.67726056318774e-05,
      "loss": 1.4125,
      "step": 101
    },
    {
      "epoch": 0.14,
      "learning_rate": 9.669130123890533e-05,
      "loss": 1.4348,
      "step": 102
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.660902046004953e-05,
      "loss": 1.4535,
      "step": 103
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.652576501592583e-05,
      "loss": 1.4669,
      "step": 104
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.644153664753173e-05,
      "loss": 1.4212,
      "step": 105
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.635633711621012e-05,
      "loss": 1.4407,
      "step": 106
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.627016820361235e-05,
      "loss": 1.4285,
      "step": 107
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.618303171166094e-05,
      "loss": 1.4433,
      "step": 108
    },
    {
      "epoch": 0.15,
      "learning_rate": 9.60949294625121e-05,
      "loss": 1.4629,
      "step": 109
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.600586329851735e-05,
      "loss": 1.4309,
      "step": 110
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.59158350821852e-05,
      "loss": 1.4424,
      "step": 111
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.582484669614211e-05,
      "loss": 1.5388,
      "step": 112
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.573290004309318e-05,
      "loss": 1.4657,
      "step": 113
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.563999704578226e-05,
      "loss": 1.4654,
      "step": 114
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.554613964695189e-05,
      "loss": 1.4174,
      "step": 115
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.545132980930251e-05,
      "loss": 1.4004,
      "step": 116
    },
    {
      "epoch": 0.16,
      "learning_rate": 9.535556951545157e-05,
      "loss": 1.4183,
      "step": 117
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.525886076789194e-05,
      "loss": 1.4468,
      "step": 118
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.516120558895014e-05,
      "loss": 1.429,
      "step": 119
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.506260602074398e-05,
      "loss": 1.3977,
      "step": 120
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.496306412513988e-05,
      "loss": 1.433,
      "step": 121
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.486258198370981e-05,
      "loss": 1.4687,
      "step": 122
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.47611616976876e-05,
      "loss": 1.4017,
      "step": 123
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.465880538792518e-05,
      "loss": 1.4283,
      "step": 124
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.455551519484816e-05,
      "loss": 1.4018,
      "step": 125
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.445129327841102e-05,
      "loss": 1.4522,
      "step": 126
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.434614181805202e-05,
      "loss": 1.4229,
      "step": 127
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.424006301264761e-05,
      "loss": 1.4334,
      "step": 128
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.413305908046636e-05,
      "loss": 1.4241,
      "step": 129
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.402513225912273e-05,
      "loss": 1.4142,
      "step": 130
    },
    {
      "epoch": 0.18,
      "learning_rate": 9.391628480553013e-05,
      "loss": 1.4467,
      "step": 131
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.38065189958538e-05,
      "loss": 1.4676,
      "step": 132
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.369583712546322e-05,
      "loss": 1.4513,
      "step": 133
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.358424150888405e-05,
      "loss": 1.4245,
      "step": 134
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.347173447974982e-05,
      "loss": 1.406,
      "step": 135
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.335831839075304e-05,
      "loss": 1.3876,
      "step": 136
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.324399561359602e-05,
      "loss": 1.352,
      "step": 137
    },
    {
      "epoch": 0.19,
      "learning_rate": 9.312876853894134e-05,
      "loss": 1.4428,
      "step": 138
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.30126395763618e-05,
      "loss": 1.4213,
      "step": 139
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.289561115429004e-05,
      "loss": 1.3948,
      "step": 140
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.277768571996772e-05,
      "loss": 1.4475,
      "step": 141
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.265886573939447e-05,
      "loss": 1.4342,
      "step": 142
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.253915369727617e-05,
      "loss": 1.3901,
      "step": 143
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.241855209697307e-05,
      "loss": 1.4319,
      "step": 144
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.229706346044747e-05,
      "loss": 1.4263,
      "step": 145
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.217469032821088e-05,
      "loss": 1.3893,
      "step": 146
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.205143525927103e-05,
      "loss": 1.4633,
      "step": 147
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.192730083107819e-05,
      "loss": 1.4429,
      "step": 148
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.180228963947144e-05,
      "loss": 1.3981,
      "step": 149
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.167640429862429e-05,
      "loss": 1.441,
      "step": 150
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.154964744099006e-05,
      "loss": 1.4101,
      "step": 151
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.142202171724674e-05,
      "loss": 1.4725,
      "step": 152
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.129352979624169e-05,
      "loss": 1.3903,
      "step": 153
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.116417436493574e-05,
      "loss": 1.4252,
      "step": 154
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.103395812834705e-05,
      "loss": 1.4432,
      "step": 155
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.09028838094945e-05,
      "loss": 1.4416,
      "step": 156
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.077095414934075e-05,
      "loss": 1.3401,
      "step": 157
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.063817190673503e-05,
      "loss": 1.4355,
      "step": 158
    },
    {
      "epoch": 0.22,
      "learning_rate": 9.050453985835527e-05,
      "loss": 1.4083,
      "step": 159
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.037006079865016e-05,
      "loss": 1.4175,
      "step": 160
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.023473753978069e-05,
      "loss": 1.4193,
      "step": 161
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.009857291156134e-05,
      "loss": 1.4148,
      "step": 162
    },
    {
      "epoch": 0.23,
      "learning_rate": 8.996156976140086e-05,
      "loss": 1.4227,
      "step": 163
    },
    {
      "epoch": 0.23,
      "learning_rate": 8.98237309542428e-05,
      "loss": 1.3991,
      "step": 164
    },
    {
      "epoch": 0.23,
      "learning_rate": 8.968505937250555e-05,
      "loss": 1.4063,
      "step": 165
    },
    {
      "epoch": 0.23,
      "learning_rate": 8.954555791602211e-05,
      "loss": 1.4072,
      "step": 166
    },
    {
      "epoch": 0.24,
      "learning_rate": 8.940522950197935e-05,
      "loss": 1.4471,
      "step": 167
    },
    {
      "epoch": 0.24,
      "learning_rate": 8.926407706485713e-05,
      "loss": 1.3875,
      "step": 168
    },
    {
      "epoch": 0.24,
      "learning_rate": 8.91221035563669e-05,
      "loss": 1.469,
      "step": 169
    },
    {
      "epoch": 0.24,
      "learning_rate": 8.897931194538989e-05,
      "loss": 1.4398,
      "step": 170
    },
    {
      "epoch": 0.24,
      "learning_rate": 8.883570521791514e-05,
      "loss": 1.436,
      "step": 171
    },
    {
      "epoch": 0.24,
      "learning_rate": 8.869128637697702e-05,
      "loss": 1.4175,
      "step": 172
    },
    {
      "epoch": 0.24,
      "learning_rate": 8.854605844259237e-05,
      "loss": 1.3957,
      "step": 173
    },
    {
      "epoch": 0.25,
      "learning_rate": 8.840002445169746e-05,
      "loss": 1.4846,
      "step": 174
    },
    {
      "epoch": 0.25,
      "learning_rate": 8.82531874580844e-05,
      "loss": 1.3793,
      "step": 175
    },
    {
      "epoch": 0.25,
      "learning_rate": 8.810555053233729e-05,
      "loss": 1.4326,
      "step": 176
    },
    {
      "epoch": 0.25,
      "learning_rate": 8.795711676176803e-05,
      "loss": 1.4546,
      "step": 177
    },
    {
      "epoch": 0.25,
      "learning_rate": 8.780788925035178e-05,
      "loss": 1.4048,
      "step": 178
    },
    {
      "epoch": 0.25,
      "learning_rate": 8.765787111866198e-05,
      "loss": 1.3985,
      "step": 179
    },
    {
      "epoch": 0.25,
      "learning_rate": 8.750706550380518e-05,
      "loss": 1.3538,
      "step": 180
    },
    {
      "epoch": 0.26,
      "learning_rate": 8.735547555935537e-05,
      "loss": 1.402,
      "step": 181
    },
    {
      "epoch": 0.26,
      "learning_rate": 8.720310445528807e-05,
      "loss": 1.3701,
      "step": 182
    },
    {
      "epoch": 0.26,
      "learning_rate": 8.704995537791405e-05,
      "loss": 1.3698,
      "step": 183
    },
    {
      "epoch": 0.26,
      "learning_rate": 8.689603152981263e-05,
      "loss": 1.3827,
      "step": 184
    },
    {
      "epoch": 0.26,
      "learning_rate": 8.674133612976481e-05,
      "loss": 1.463,
      "step": 185
    },
    {
      "epoch": 0.26,
      "learning_rate": 8.658587241268587e-05,
      "loss": 1.363,
      "step": 186
    },
    {
      "epoch": 0.26,
      "learning_rate": 8.642964362955781e-05,
      "loss": 1.4031,
      "step": 187
    },
    {
      "epoch": 0.26,
      "learning_rate": 8.627265304736131e-05,
      "loss": 1.4026,
      "step": 188
    },
    {
      "epoch": 0.27,
      "learning_rate": 8.611490394900739e-05,
      "loss": 1.45,
      "step": 189
    },
    {
      "epoch": 0.27,
      "learning_rate": 8.595639963326881e-05,
      "loss": 1.3965,
      "step": 190
    },
    {
      "epoch": 0.27,
      "learning_rate": 8.579714341471106e-05,
      "loss": 1.4117,
      "step": 191
    },
    {
      "epoch": 0.27,
      "learning_rate": 8.56371386236231e-05,
      "loss": 1.3525,
      "step": 192
    },
    {
      "epoch": 0.27,
      "learning_rate": 8.547638860594764e-05,
      "loss": 1.4094,
      "step": 193
    },
    {
      "epoch": 0.27,
      "learning_rate": 8.531489672321122e-05,
      "loss": 1.3721,
      "step": 194
    },
    {
      "epoch": 0.27,
      "learning_rate": 8.515266635245389e-05,
      "loss": 1.3894,
      "step": 195
    },
    {
      "epoch": 0.28,
      "learning_rate": 8.498970088615861e-05,
      "loss": 1.3716,
      "step": 196
    },
    {
      "epoch": 0.28,
      "learning_rate": 8.48260037321803e-05,
      "loss": 1.4026,
      "step": 197
    },
    {
      "epoch": 0.28,
      "learning_rate": 8.46615783136746e-05,
      "loss": 1.3507,
      "step": 198
    },
    {
      "epoch": 0.28,
      "learning_rate": 8.449642806902623e-05,
      "loss": 1.3839,
      "step": 199
    },
    {
      "epoch": 0.28,
      "learning_rate": 8.433055645177714e-05,
      "loss": 1.3451,
      "step": 200
    },
    {
      "epoch": 0.28,
      "learning_rate": 8.416396693055429e-05,
      "loss": 1.3518,
      "step": 201
    },
    {
      "epoch": 0.28,
      "learning_rate": 8.399666298899706e-05,
      "loss": 1.3893,
      "step": 202
    },
    {
      "epoch": 0.29,
      "learning_rate": 8.382864812568452e-05,
      "loss": 1.3508,
      "step": 203
    },
    {
      "epoch": 0.29,
      "learning_rate": 8.365992585406207e-05,
      "loss": 1.335,
      "step": 204
    },
    {
      "epoch": 0.29,
      "learning_rate": 8.34904997023682e-05,
      "loss": 1.3734,
      "step": 205
    },
    {
      "epoch": 0.29,
      "learning_rate": 8.332037321356057e-05,
      "loss": 1.3876,
      "step": 206
    },
    {
      "epoch": 0.29,
      "learning_rate": 8.31495499452419e-05,
      "loss": 1.398,
      "step": 207
    },
    {
      "epoch": 0.29,
      "learning_rate": 8.297803346958571e-05,
      "loss": 1.4288,
      "step": 208
    },
    {
      "epoch": 0.29,
      "learning_rate": 8.280582737326146e-05,
      "loss": 1.389,
      "step": 209
    },
    {
      "epoch": 0.3,
      "learning_rate": 8.263293525735967e-05,
      "loss": 1.3819,
      "step": 210
    },
    {
      "epoch": 0.3,
      "learning_rate": 8.245936073731653e-05,
      "loss": 1.3584,
      "step": 211
    },
    {
      "epoch": 0.3,
      "learning_rate": 8.228510744283837e-05,
      "loss": 1.3803,
      "step": 212
    },
    {
      "epoch": 0.3,
      "learning_rate": 8.211017901782574e-05,
      "loss": 1.4161,
      "step": 213
    },
    {
      "epoch": 0.3,
      "learning_rate": 8.193457912029713e-05,
      "loss": 1.3701,
      "step": 214
    },
    {
      "epoch": 0.3,
      "learning_rate": 8.175831142231258e-05,
      "loss": 1.4184,
      "step": 215
    },
    {
      "epoch": 0.3,
      "learning_rate": 8.158137960989685e-05,
      "loss": 1.4,
      "step": 216
    },
    {
      "epoch": 0.31,
      "learning_rate": 8.140378738296233e-05,
      "loss": 1.3785,
      "step": 217
    },
    {
      "epoch": 0.31,
      "learning_rate": 8.122553845523166e-05,
      "loss": 1.4217,
      "step": 218
    },
    {
      "epoch": 0.31,
      "learning_rate": 8.104663655416014e-05,
      "loss": 1.3014,
      "step": 219
    },
    {
      "epoch": 0.31,
      "learning_rate": 8.086708542085768e-05,
      "loss": 1.3709,
      "step": 220
    },
    {
      "epoch": 0.31,
      "learning_rate": 8.068688881001065e-05,
      "loss": 1.3432,
      "step": 221
    },
    {
      "epoch": 0.31,
      "learning_rate": 8.050605048980333e-05,
      "loss": 1.3728,
      "step": 222
    },
    {
      "epoch": 0.31,
      "learning_rate": 8.03245742418391e-05,
      "loss": 1.4366,
      "step": 223
    },
    {
      "epoch": 0.32,
      "learning_rate": 8.014246386106138e-05,
      "loss": 1.3944,
      "step": 224
    },
    {
      "epoch": 0.32,
      "learning_rate": 7.995972315567431e-05,
      "loss": 1.4022,
      "step": 225
    },
    {
      "epoch": 0.32,
      "learning_rate": 7.977635594706299e-05,
      "loss": 1.3549,
      "step": 226
    },
    {
      "epoch": 0.32,
      "learning_rate": 7.959236606971375e-05,
      "loss": 1.4002,
      "step": 227
    },
    {
      "epoch": 0.32,
      "learning_rate": 7.940775737113378e-05,
      "loss": 1.3788,
      "step": 228
    },
    {
      "epoch": 0.32,
      "learning_rate": 7.922253371177082e-05,
      "loss": 1.3339,
      "step": 229
    },
    {
      "epoch": 0.32,
      "learning_rate": 7.903669896493233e-05,
      "loss": 1.3297,
      "step": 230
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.885025701670457e-05,
      "loss": 1.2987,
      "step": 231
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.866321176587129e-05,
      "loss": 1.4247,
      "step": 232
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.84755671238322e-05,
      "loss": 1.3754,
      "step": 233
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.828732701452119e-05,
      "loss": 1.3807,
      "step": 234
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.809849537432432e-05,
      "loss": 1.3654,
      "step": 235
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.790907615199736e-05,
      "loss": 1.3311,
      "step": 236
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.771907330858341e-05,
      "loss": 1.3283,
      "step": 237
    },
    {
      "epoch": 0.34,
      "learning_rate": 7.752849081732993e-05,
      "loss": 1.3423,
      "step": 238
    },
    {
      "epoch": 0.34,
      "learning_rate": 7.733733266360568e-05,
      "loss": 1.37,
      "step": 239
    },
    {
      "epoch": 0.34,
      "learning_rate": 7.714560284481742e-05,
      "loss": 1.355,
      "step": 240
    },
    {
      "epoch": 0.34,
      "learning_rate": 7.695330537032628e-05,
      "loss": 1.397,
      "step": 241
    },
    {
      "epoch": 0.34,
      "learning_rate": 7.676044426136397e-05,
      "loss": 1.3589,
      "step": 242
    },
    {
      "epoch": 0.34,
      "learning_rate": 7.656702355094859e-05,
      "loss": 1.3804,
      "step": 243
    },
    {
      "epoch": 0.34,
      "learning_rate": 7.637304728380036e-05,
      "loss": 1.3695,
      "step": 244
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.61785195162571e-05,
      "loss": 1.3546,
      "step": 245
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.598344431618926e-05,
      "loss": 1.33,
      "step": 246
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.578782576291501e-05,
      "loss": 1.3519,
      "step": 247
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.559166794711476e-05,
      "loss": 1.3526,
      "step": 248
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.539497497074584e-05,
      "loss": 1.37,
      "step": 249
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.519775094695649e-05,
      "loss": 1.343,
      "step": 250
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.500000000000001e-05,
      "loss": 1.3729,
      "step": 251
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.480172626514845e-05,
      "loss": 1.344,
      "step": 252
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.460293388860615e-05,
      "loss": 1.3071,
      "step": 253
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.440362702742305e-05,
      "loss": 1.3789,
      "step": 254
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.420380984940773e-05,
      "loss": 1.3207,
      "step": 255
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.400348653304022e-05,
      "loss": 1.3365,
      "step": 256
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.380266126738476e-05,
      "loss": 1.3616,
      "step": 257
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.360133825200205e-05,
      "loss": 1.3094,
      "step": 258
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.339952169686151e-05,
      "loss": 1.3491,
      "step": 259
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.319721582225323e-05,
      "loss": 1.3591,
      "step": 260
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.29944248586997e-05,
      "loss": 1.336,
      "step": 261
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.279115304686735e-05,
      "loss": 1.3376,
      "step": 262
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.258740463747788e-05,
      "loss": 1.3373,
      "step": 263
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.238318389121939e-05,
      "loss": 1.3632,
      "step": 264
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.217849507865724e-05,
      "loss": 1.3784,
      "step": 265
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.197334248014477e-05,
      "loss": 1.3099,
      "step": 266
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.176773038573377e-05,
      "loss": 1.3208,
      "step": 267
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.156166309508482e-05,
      "loss": 1.3293,
      "step": 268
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.13551449173773e-05,
      "loss": 1.3885,
      "step": 269
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.114818017121939e-05,
      "loss": 1.354,
      "step": 270
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.094077318455762e-05,
      "loss": 1.3362,
      "step": 271
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.073292829458645e-05,
      "loss": 1.3203,
      "step": 272
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.052464984765764e-05,
      "loss": 1.3777,
      "step": 273
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.031594219918916e-05,
      "loss": 1.3297,
      "step": 274
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.010680971357434e-05,
      "loss": 1.3249,
      "step": 275
    },
    {
      "epoch": 0.39,
      "learning_rate": 6.989725676409044e-05,
      "loss": 1.3271,
      "step": 276
    },
    {
      "epoch": 0.39,
      "learning_rate": 6.96872877328073e-05,
      "loss": 1.3234,
      "step": 277
    },
    {
      "epoch": 0.39,
      "learning_rate": 6.94769070104956e-05,
      "loss": 1.3597,
      "step": 278
    },
    {
      "epoch": 0.39,
      "learning_rate": 6.926611899653516e-05,
      "loss": 1.3421,
      "step": 279
    },
    {
      "epoch": 0.39,
      "learning_rate": 6.905492809882286e-05,
      "loss": 1.3048,
      "step": 280
    },
    {
      "epoch": 0.4,
      "learning_rate": 6.88433387336805e-05,
      "loss": 1.3737,
      "step": 281
    },
    {
      "epoch": 0.4,
      "learning_rate": 6.863135532576241e-05,
      "loss": 1.3015,
      "step": 282
    },
    {
      "epoch": 0.4,
      "learning_rate": 6.841898230796302e-05,
      "loss": 1.325,
      "step": 283
    },
    {
      "epoch": 0.4,
      "learning_rate": 6.820622412132402e-05,
      "loss": 1.3183,
      "step": 284
    },
    {
      "epoch": 0.4,
      "learning_rate": 6.799308521494156e-05,
      "loss": 1.3447,
      "step": 285
    },
    {
      "epoch": 0.4,
      "learning_rate": 6.777957004587331e-05,
      "loss": 1.2967,
      "step": 286
    },
    {
      "epoch": 0.4,
      "learning_rate": 6.756568307904508e-05,
      "loss": 1.3522,
      "step": 287
    },
    {
      "epoch": 0.41,
      "learning_rate": 6.735142878715754e-05,
      "loss": 1.3333,
      "step": 288
    },
    {
      "epoch": 0.41,
      "learning_rate": 6.713681165059271e-05,
      "loss": 1.4079,
      "step": 289
    },
    {
      "epoch": 0.41,
      "learning_rate": 6.692183615732025e-05,
      "loss": 1.2946,
      "step": 290
    },
    {
      "epoch": 0.41,
      "learning_rate": 6.670650680280358e-05,
      "loss": 1.3045,
      "step": 291
    },
    {
      "epoch": 0.41,
      "learning_rate": 6.649082808990586e-05,
      "loss": 1.358,
      "step": 292
    },
    {
      "epoch": 0.41,
      "learning_rate": 6.627480452879593e-05,
      "loss": 1.2442,
      "step": 293
    },
    {
      "epoch": 0.41,
      "learning_rate": 6.605844063685392e-05,
      "loss": 1.3063,
      "step": 294
    },
    {
      "epoch": 0.42,
      "learning_rate": 6.584174093857675e-05,
      "loss": 1.3009,
      "step": 295
    },
    {
      "epoch": 0.42,
      "learning_rate": 6.562470996548361e-05,
      "loss": 1.2866,
      "step": 296
    },
    {
      "epoch": 0.42,
      "learning_rate": 6.54073522560211e-05,
      "loss": 1.3432,
      "step": 297
    },
    {
      "epoch": 0.42,
      "learning_rate": 6.518967235546841e-05,
      "loss": 1.363,
      "step": 298
    },
    {
      "epoch": 0.42,
      "learning_rate": 6.497167481584221e-05,
      "loss": 1.3525,
      "step": 299
    },
    {
      "epoch": 0.42,
      "learning_rate": 6.475336419580151e-05,
      "loss": 1.3155,
      "step": 300
    },
    {
      "epoch": 0.42,
      "learning_rate": 6.453474506055228e-05,
      "loss": 1.3143,
      "step": 301
    },
    {
      "epoch": 0.43,
      "learning_rate": 6.431582198175203e-05,
      "loss": 1.3338,
      "step": 302
    },
    {
      "epoch": 0.43,
      "learning_rate": 6.409659953741416e-05,
      "loss": 1.3271,
      "step": 303
    },
    {
      "epoch": 0.43,
      "learning_rate": 6.387708231181229e-05,
      "loss": 1.2591,
      "step": 304
    },
    {
      "epoch": 0.43,
      "learning_rate": 6.365727489538437e-05,
      "loss": 1.2911,
      "step": 305
    },
    {
      "epoch": 0.43,
      "learning_rate": 6.343718188463663e-05,
      "loss": 1.2664,
      "step": 306
    },
    {
      "epoch": 0.43,
      "learning_rate": 6.321680788204758e-05,
      "loss": 1.2654,
      "step": 307
    },
    {
      "epoch": 0.43,
      "learning_rate": 6.299615749597165e-05,
      "loss": 1.256,
      "step": 308
    },
    {
      "epoch": 0.44,
      "learning_rate": 6.277523534054284e-05,
      "loss": 1.2852,
      "step": 309
    },
    {
      "epoch": 0.44,
      "learning_rate": 6.255404603557833e-05,
      "loss": 1.2917,
      "step": 310
    },
    {
      "epoch": 0.44,
      "learning_rate": 6.233259420648175e-05,
      "loss": 1.2603,
      "step": 311
    },
    {
      "epoch": 0.44,
      "learning_rate": 6.211088448414653e-05,
      "loss": 1.324,
      "step": 312
    },
    {
      "epoch": 0.44,
      "learning_rate": 6.188892150485903e-05,
      "loss": 1.3535,
      "step": 313
    },
    {
      "epoch": 0.44,
      "learning_rate": 6.166670991020162e-05,
      "loss": 1.2928,
      "step": 314
    },
    {
      "epoch": 0.44,
      "learning_rate": 6.144425434695551e-05,
      "loss": 1.3231,
      "step": 315
    },
    {
      "epoch": 0.45,
      "learning_rate": 6.122155946700381e-05,
      "loss": 1.313,
      "step": 316
    },
    {
      "epoch": 0.45,
      "learning_rate": 6.099862992723397e-05,
      "loss": 1.3226,
      "step": 317
    },
    {
      "epoch": 0.45,
      "learning_rate": 6.077547038944058e-05,
      "loss": 1.3328,
      "step": 318
    },
    {
      "epoch": 0.45,
      "learning_rate": 6.0552085520227875e-05,
      "loss": 1.2803,
      "step": 319
    },
    {
      "epoch": 0.45,
      "learning_rate": 6.032847999091206e-05,
      "loss": 1.3747,
      "step": 320
    },
    {
      "epoch": 0.45,
      "learning_rate": 6.010465847742368e-05,
      "loss": 1.2863,
      "step": 321
    },
    {
      "epoch": 0.45,
      "learning_rate": 5.988062566020987e-05,
      "loss": 1.3153,
      "step": 322
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.9656386224136426e-05,
      "loss": 1.2917,
      "step": 323
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.943194485838985e-05,
      "loss": 1.2688,
      "step": 324
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.920730625637934e-05,
      "loss": 1.3114,
      "step": 325
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.8982475115638515e-05,
      "loss": 1.3089,
      "step": 326
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.875745613772736e-05,
      "loss": 1.3035,
      "step": 327
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.85322540281338e-05,
      "loss": 1.3158,
      "step": 328
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.830687349617529e-05,
      "loss": 1.2819,
      "step": 329
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.808131925490039e-05,
      "loss": 1.3261,
      "step": 330
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.7855596020990186e-05,
      "loss": 1.2925,
      "step": 331
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.7629708514659655e-05,
      "loss": 1.2977,
      "step": 332
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.740366145955893e-05,
      "loss": 1.2672,
      "step": 333
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.7177459582674595e-05,
      "loss": 1.2812,
      "step": 334
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.6951107614230783e-05,
      "loss": 1.343,
      "step": 335
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.672461028759024e-05,
      "loss": 1.2741,
      "step": 336
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.649797233915539e-05,
      "loss": 1.2426,
      "step": 337
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.627119850826931e-05,
      "loss": 1.2673,
      "step": 338
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.6044293537116496e-05,
      "loss": 1.2921,
      "step": 339
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.5817262170623865e-05,
      "loss": 1.3118,
      "step": 340
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.559010915636143e-05,
      "loss": 1.2766,
      "step": 341
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.5362839244443034e-05,
      "loss": 1.311,
      "step": 342
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.513545718742702e-05,
      "loss": 1.327,
      "step": 343
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.490796774021687e-05,
      "loss": 1.3586,
      "step": 344
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.468037565996177e-05,
      "loss": 1.3274,
      "step": 345
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.445268570595708e-05,
      "loss": 1.2875,
      "step": 346
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.42249026395449e-05,
      "loss": 1.2926,
      "step": 347
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.399703122401441e-05,
      "loss": 1.3126,
      "step": 348
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.376907622450229e-05,
      "loss": 1.3025,
      "step": 349
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.3541042407893164e-05,
      "loss": 1.3226,
      "step": 350
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.331293454271974e-05,
      "loss": 1.3268,
      "step": 351
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.308475739906329e-05,
      "loss": 1.328,
      "step": 352
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.285651574845374e-05,
      "loss": 1.3183,
      "step": 353
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.262821436376998e-05,
      "loss": 1.2694,
      "step": 354
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.239985801914e-05,
      "loss": 1.3136,
      "step": 355
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.217145148984114e-05,
      "loss": 1.2293,
      "step": 356
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.1942999552200136e-05,
      "loss": 1.2772,
      "step": 357
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.171450698349329e-05,
      "loss": 1.3034,
      "step": 358
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.148597856184656e-05,
      "loss": 1.3186,
      "step": 359
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.125741906613565e-05,
      "loss": 1.2917,
      "step": 360
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.102883327588608e-05,
      "loss": 1.2597,
      "step": 361
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.080022597117318e-05,
      "loss": 1.2747,
      "step": 362
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.057160193252225e-05,
      "loss": 1.3008,
      "step": 363
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.0342965940808486e-05,
      "loss": 1.3098,
      "step": 364
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.011432277715702e-05,
      "loss": 1.2622,
      "step": 365
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.9885677222842984e-05,
      "loss": 1.2856,
      "step": 366
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.965703405919154e-05,
      "loss": 1.2775,
      "step": 367
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.942839806747775e-05,
      "loss": 1.2891,
      "step": 368
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.919977402882682e-05,
      "loss": 1.3406,
      "step": 369
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.897116672411395e-05,
      "loss": 1.302,
      "step": 370
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.8742580933864356e-05,
      "loss": 1.3238,
      "step": 371
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.851402143815345e-05,
      "loss": 1.2748,
      "step": 372
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.828549301650673e-05,
      "loss": 1.2535,
      "step": 373
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.8057000447799876e-05,
      "loss": 1.26,
      "step": 374
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.782854851015886e-05,
      "loss": 1.2667,
      "step": 375
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.760014198086002e-05,
      "loss": 1.2761,
      "step": 376
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.737178563623004e-05,
      "loss": 1.2922,
      "step": 377
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.714348425154627e-05,
      "loss": 1.2724,
      "step": 378
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.691524260093672e-05,
      "loss": 1.2786,
      "step": 379
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.668706545728026e-05,
      "loss": 1.3129,
      "step": 380
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.6458957592106855e-05,
      "loss": 1.2888,
      "step": 381
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.623092377549772e-05,
      "loss": 1.2741,
      "step": 382
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.600296877598561e-05,
      "loss": 1.2683,
      "step": 383
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.577509736045511e-05,
      "loss": 1.2782,
      "step": 384
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.554731429404293e-05,
      "loss": 1.2592,
      "step": 385
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.5319624340038244e-05,
      "loss": 1.3419,
      "step": 386
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.509203225978314e-05,
      "loss": 1.2586,
      "step": 387
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.486454281257299e-05,
      "loss": 1.3275,
      "step": 388
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.4637160755557e-05,
      "loss": 1.2152,
      "step": 389
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.4409890843638584e-05,
      "loss": 1.3169,
      "step": 390
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.418273782937613e-05,
      "loss": 1.2855,
      "step": 391
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.395570646288352e-05,
      "loss": 1.2351,
      "step": 392
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.372880149173071e-05,
      "loss": 1.3066,
      "step": 393
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.35020276608446e-05,
      "loss": 1.2788,
      "step": 394
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.327538971240978e-05,
      "loss": 1.3019,
      "step": 395
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.304889238576922e-05,
      "loss": 1.2754,
      "step": 396
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.2822540417325396e-05,
      "loss": 1.2644,
      "step": 397
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.2596338540441086e-05,
      "loss": 1.2097,
      "step": 398
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.237029148534036e-05,
      "loss": 1.2849,
      "step": 399
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.2144403979009826e-05,
      "loss": 1.2622,
      "step": 400
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.1918680745099614e-05,
      "loss": 1.2837,
      "step": 401
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.169312650382471e-05,
      "loss": 1.2705,
      "step": 402
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.1467745971866216e-05,
      "loss": 1.274,
      "step": 403
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.124254386227264e-05,
      "loss": 1.2877,
      "step": 404
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.101752488436149e-05,
      "loss": 1.2715,
      "step": 405
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.0792693743620695e-05,
      "loss": 1.2333,
      "step": 406
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.056805514161015e-05,
      "loss": 1.3097,
      "step": 407
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.034361377586357e-05,
      "loss": 1.3334,
      "step": 408
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.0119374339790136e-05,
      "loss": 1.2569,
      "step": 409
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.989534152257632e-05,
      "loss": 1.2795,
      "step": 410
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.967152000908796e-05,
      "loss": 1.3314,
      "step": 411
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.944791447977214e-05,
      "loss": 1.3027,
      "step": 412
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.922452961055941e-05,
      "loss": 1.2614,
      "step": 413
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.900137007276605e-05,
      "loss": 1.2581,
      "step": 414
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.8778440532996204e-05,
      "loss": 1.2899,
      "step": 415
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.855574565304448e-05,
      "loss": 1.2968,
      "step": 416
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.8333290089798415e-05,
      "loss": 1.2594,
      "step": 417
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.811107849514098e-05,
      "loss": 1.2745,
      "step": 418
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.788911551585348e-05,
      "loss": 1.2817,
      "step": 419
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.7667405793518264e-05,
      "loss": 1.2884,
      "step": 420
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.744595396442169e-05,
      "loss": 1.2748,
      "step": 421
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.722476465945718e-05,
      "loss": 1.309,
      "step": 422
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.7003842504028366e-05,
      "loss": 1.3223,
      "step": 423
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.678319211795242e-05,
      "loss": 1.3418,
      "step": 424
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.656281811536337e-05,
      "loss": 1.274,
      "step": 425
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.634272510461564e-05,
      "loss": 1.2765,
      "step": 426
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.612291768818772e-05,
      "loss": 1.2968,
      "step": 427
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.590340046258586e-05,
      "loss": 1.2538,
      "step": 428
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5684178018247996e-05,
      "loss": 1.3016,
      "step": 429
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.546525493944773e-05,
      "loss": 1.287,
      "step": 430
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.52466358041985e-05,
      "loss": 1.2595,
      "step": 431
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.502832518415778e-05,
      "loss": 1.2606,
      "step": 432
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.48103276445316e-05,
      "loss": 1.2743,
      "step": 433
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.459264774397891e-05,
      "loss": 1.2596,
      "step": 434
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.437529003451639e-05,
      "loss": 1.2515,
      "step": 435
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.415825906142326e-05,
      "loss": 1.2812,
      "step": 436
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.394155936314609e-05,
      "loss": 1.3057,
      "step": 437
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.372519547120407e-05,
      "loss": 1.2646,
      "step": 438
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.350917191009416e-05,
      "loss": 1.2425,
      "step": 439
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.329349319719644e-05,
      "loss": 1.2911,
      "step": 440
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.307816384267975e-05,
      "loss": 1.2944,
      "step": 441
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.286318834940729e-05,
      "loss": 1.2845,
      "step": 442
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.264857121284246e-05,
      "loss": 1.2854,
      "step": 443
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.2434316920954935e-05,
      "loss": 1.2476,
      "step": 444
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.222042995412669e-05,
      "loss": 1.3087,
      "step": 445
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.200691478505843e-05,
      "loss": 1.2395,
      "step": 446
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.1793775878676e-05,
      "loss": 1.2305,
      "step": 447
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.1581017692036985e-05,
      "loss": 1.2651,
      "step": 448
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.136864467423758e-05,
      "loss": 1.2375,
      "step": 449
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.115666126631952e-05,
      "loss": 1.2829,
      "step": 450
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.094507190117715e-05,
      "loss": 1.3191,
      "step": 451
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.073388100346484e-05,
      "loss": 1.2975,
      "step": 452
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.0523092989504415e-05,
      "loss": 1.2755,
      "step": 453
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.0312712267192713e-05,
      "loss": 1.2873,
      "step": 454
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.010274323590956e-05,
      "loss": 1.2662,
      "step": 455
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.989319028642567e-05,
      "loss": 1.3002,
      "step": 456
    },
    {
      "epoch": 0.64,
      "learning_rate": 2.9684057800810845e-05,
      "loss": 1.2369,
      "step": 457
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.9475350152342378e-05,
      "loss": 1.2125,
      "step": 458
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.9267071705413552e-05,
      "loss": 1.296,
      "step": 459
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.9059226815442385e-05,
      "loss": 1.2878,
      "step": 460
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.8851819828780623e-05,
      "loss": 1.308,
      "step": 461
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.8644855082622695e-05,
      "loss": 1.2083,
      "step": 462
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.8438336904915185e-05,
      "loss": 1.2431,
      "step": 463
    },
    {
      "epoch": 0.65,
      "learning_rate": 2.823226961426625e-05,
      "loss": 1.2315,
      "step": 464
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.802665751985525e-05,
      "loss": 1.3051,
      "step": 465
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.7821504921342777e-05,
      "loss": 1.2527,
      "step": 466
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.7616816108780623e-05,
      "loss": 1.295,
      "step": 467
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.741259536252213e-05,
      "loss": 1.2856,
      "step": 468
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.7208846953132682e-05,
      "loss": 1.28,
      "step": 469
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.700557514130032e-05,
      "loss": 1.2658,
      "step": 470
    },
    {
      "epoch": 0.66,
      "learning_rate": 2.6802784177746777e-05,
      "loss": 1.2511,
      "step": 471
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.6600478303138503e-05,
      "loss": 1.3012,
      "step": 472
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.6398661747997955e-05,
      "loss": 1.2492,
      "step": 473
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.619733873261524e-05,
      "loss": 1.2535,
      "step": 474
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5996513466959794e-05,
      "loss": 1.261,
      "step": 475
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.579619015059229e-05,
      "loss": 1.2847,
      "step": 476
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5596372972576967e-05,
      "loss": 1.2955,
      "step": 477
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.5397066111393853e-05,
      "loss": 1.2465,
      "step": 478
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.5198273734851553e-05,
      "loss": 1.2523,
      "step": 479
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.500000000000001e-05,
      "loss": 1.3038,
      "step": 480
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.4802249053043526e-05,
      "loss": 1.3138,
      "step": 481
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.4605025029254164e-05,
      "loss": 1.2282,
      "step": 482
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.4408332052885246e-05,
      "loss": 1.2731,
      "step": 483
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.4212174237085007e-05,
      "loss": 1.3086,
      "step": 484
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.401655568381074e-05,
      "loss": 1.3227,
      "step": 485
    },
    {
      "epoch": 0.68,
      "learning_rate": 2.382148048374292e-05,
      "loss": 1.2396,
      "step": 486
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.3626952716199647e-05,
      "loss": 1.2487,
      "step": 487
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.3432976449051442e-05,
      "loss": 1.2149,
      "step": 488
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.3239555738636044e-05,
      "loss": 1.2986,
      "step": 489
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.3046694629673716e-05,
      "loss": 1.2814,
      "step": 490
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.28543971551826e-05,
      "loss": 1.2465,
      "step": 491
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.266266733639434e-05,
      "loss": 1.307,
      "step": 492
    },
    {
      "epoch": 0.69,
      "learning_rate": 2.247150918267008e-05,
      "loss": 1.2532,
      "step": 493
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.2280926691416603e-05,
      "loss": 1.261,
      "step": 494
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.209092384800265e-05,
      "loss": 1.2771,
      "step": 495
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.190150462567569e-05,
      "loss": 1.2589,
      "step": 496
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.1712672985478815e-05,
      "loss": 1.278,
      "step": 497
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.1524432876167812e-05,
      "loss": 1.2595,
      "step": 498
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.133678823412873e-05,
      "loss": 1.299,
      "step": 499
    },
    {
      "epoch": 0.7,
      "learning_rate": 2.1149742983295446e-05,
      "loss": 1.3015,
      "step": 500
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.0963301035067685e-05,
      "loss": 1.2777,
      "step": 501
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.0777466288229207e-05,
      "loss": 1.2536,
      "step": 502
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.0592242628866236e-05,
      "loss": 1.2498,
      "step": 503
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.040763393028627e-05,
      "loss": 1.2639,
      "step": 504
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.022364405293703e-05,
      "loss": 1.2522,
      "step": 505
    },
    {
      "epoch": 0.71,
      "learning_rate": 2.0040276844325718e-05,
      "loss": 1.263,
      "step": 506
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9857536138938627e-05,
      "loss": 1.2423,
      "step": 507
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9675425758160925e-05,
      "loss": 1.2513,
      "step": 508
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.949394951019669e-05,
      "loss": 1.2495,
      "step": 509
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9313111189989375e-05,
      "loss": 1.2178,
      "step": 510
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.913291457914234e-05,
      "loss": 1.2509,
      "step": 511
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8953363445839877e-05,
      "loss": 1.2768,
      "step": 512
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8774461544768347e-05,
      "loss": 1.2517,
      "step": 513
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8596212617037694e-05,
      "loss": 1.2702,
      "step": 514
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8418620390103163e-05,
      "loss": 1.2566,
      "step": 515
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8241688577687426e-05,
      "loss": 1.2464,
      "step": 516
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.806542087970289e-05,
      "loss": 1.2356,
      "step": 517
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.788982098217427e-05,
      "loss": 1.2719,
      "step": 518
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7714892557161624e-05,
      "loss": 1.2667,
      "step": 519
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.754063926268349e-05,
      "loss": 1.2486,
      "step": 520
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7367064742640348e-05,
      "loss": 1.2685,
      "step": 521
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.719417262673854e-05,
      "loss": 1.3077,
      "step": 522
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7021966530414303e-05,
      "loss": 1.2647,
      "step": 523
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.6850450054758092e-05,
      "loss": 1.2924,
      "step": 524
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.667962678643943e-05,
      "loss": 1.2512,
      "step": 525
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.6509500297631787e-05,
      "loss": 1.2396,
      "step": 526
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.6340074145937934e-05,
      "loss": 1.2432,
      "step": 527
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.6171351874315494e-05,
      "loss": 1.2462,
      "step": 528
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.600333701100293e-05,
      "loss": 1.2697,
      "step": 529
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.583603306944572e-05,
      "loss": 1.2251,
      "step": 530
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.566944354822286e-05,
      "loss": 1.2481,
      "step": 531
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5503571930973786e-05,
      "loss": 1.2048,
      "step": 532
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.533842168632541e-05,
      "loss": 1.277,
      "step": 533
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5173996267819695e-05,
      "loss": 1.2525,
      "step": 534
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.5010299113841397e-05,
      "loss": 1.2449,
      "step": 535
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4847333647546113e-05,
      "loss": 1.2571,
      "step": 536
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.468510327678877e-05,
      "loss": 1.2618,
      "step": 537
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4523611394052356e-05,
      "loss": 1.2877,
      "step": 538
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4362861376376896e-05,
      "loss": 1.2563,
      "step": 539
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4202856585288954e-05,
      "loss": 1.3038,
      "step": 540
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4043600366731213e-05,
      "loss": 1.2746,
      "step": 541
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.3885096050992624e-05,
      "loss": 1.2297,
      "step": 542
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.3727346952638703e-05,
      "loss": 1.2217,
      "step": 543
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.3570356370442188e-05,
      "loss": 1.2506,
      "step": 544
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.341412758731413e-05,
      "loss": 1.2465,
      "step": 545
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.325866387023521e-05,
      "loss": 1.2689,
      "step": 546
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.3103968470187384e-05,
      "loss": 1.235,
      "step": 547
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.2950044622085955e-05,
      "loss": 1.2777,
      "step": 548
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.2796895544711929e-05,
      "loss": 1.2638,
      "step": 549
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.2644524440644628e-05,
      "loss": 1.2612,
      "step": 550
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.249293449619483e-05,
      "loss": 1.269,
      "step": 551
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.2342128881338027e-05,
      "loss": 1.2453,
      "step": 552
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.2192110749648233e-05,
      "loss": 1.2294,
      "step": 553
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.2042883238231984e-05,
      "loss": 1.2449,
      "step": 554
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.1894449467662728e-05,
      "loss": 1.2523,
      "step": 555
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.1746812541915608e-05,
      "loss": 1.2162,
      "step": 556
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.1599975548302549e-05,
      "loss": 1.2533,
      "step": 557
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.1453941557407638e-05,
      "loss": 1.2307,
      "step": 558
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.1308713623022987e-05,
      "loss": 1.2713,
      "step": 559
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.1164294782084866e-05,
      "loss": 1.2636,
      "step": 560
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.1020688054610118e-05,
      "loss": 1.2408,
      "step": 561
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.0877896443633117e-05,
      "loss": 1.2865,
      "step": 562
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.0735922935142873e-05,
      "loss": 1.2493,
      "step": 563
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.0594770498020657e-05,
      "loss": 1.2406,
      "step": 564
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.0454442083977912e-05,
      "loss": 1.263,
      "step": 565
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.0314940627494451e-05,
      "loss": 1.2701,
      "step": 566
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.0176269045757202e-05,
      "loss": 1.2196,
      "step": 567
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.0038430238599156e-05,
      "loss": 1.2103,
      "step": 568
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.901427088438675e-06,
      "loss": 1.2255,
      "step": 569
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.76526246021931e-06,
      "loss": 1.2491,
      "step": 570
    },
    {
      "epoch": 0.8,
      "learning_rate": 9.629939201349853e-06,
      "loss": 1.2494,
      "step": 571
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.49546014164474e-06,
      "loss": 1.2868,
      "step": 572
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.361828093264984e-06,
      "loss": 1.2383,
      "step": 573
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.229045850659252e-06,
      "loss": 1.2227,
      "step": 574
    },
    {
      "epoch": 0.81,
      "learning_rate": 9.097116190505516e-06,
      "loss": 1.2423,
      "step": 575
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.966041871652969e-06,
      "loss": 1.233,
      "step": 576
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.835825635064266e-06,
      "loss": 1.2144,
      "step": 577
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.706470203758316e-06,
      "loss": 1.2529,
      "step": 578
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.577978282753274e-06,
      "loss": 1.2622,
      "step": 579
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.45035255900995e-06,
      "loss": 1.2398,
      "step": 580
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.323595701375702e-06,
      "loss": 1.2622,
      "step": 581
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.197710360528571e-06,
      "loss": 1.2585,
      "step": 582
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.072699168921826e-06,
      "loss": 1.2459,
      "step": 583
    },
    {
      "epoch": 0.82,
      "learning_rate": 7.948564740728998e-06,
      "loss": 1.2625,
      "step": 584
    },
    {
      "epoch": 0.82,
      "learning_rate": 7.825309671789128e-06,
      "loss": 1.2628,
      "step": 585
    },
    {
      "epoch": 0.83,
      "learning_rate": 7.70293653955254e-06,
      "loss": 1.2116,
      "step": 586
    },
    {
      "epoch": 0.83,
      "learning_rate": 7.581447903026939e-06,
      "loss": 1.2787,
      "step": 587
    },
    {
      "epoch": 0.83,
      "learning_rate": 7.460846302723845e-06,
      "loss": 1.2558,
      "step": 588
    },
    {
      "epoch": 0.83,
      "learning_rate": 7.341134260605536e-06,
      "loss": 1.2558,
      "step": 589
    },
    {
      "epoch": 0.83,
      "learning_rate": 7.2223142800322775e-06,
      "loss": 1.2198,
      "step": 590
    },
    {
      "epoch": 0.83,
      "learning_rate": 7.104388845709981e-06,
      "loss": 1.2856,
      "step": 591
    },
    {
      "epoch": 0.83,
      "learning_rate": 6.987360423638206e-06,
      "loss": 1.1944,
      "step": 592
    },
    {
      "epoch": 0.84,
      "learning_rate": 6.871231461058658e-06,
      "loss": 1.2197,
      "step": 593
    },
    {
      "epoch": 0.84,
      "learning_rate": 6.756004386403996e-06,
      "loss": 1.2229,
      "step": 594
    },
    {
      "epoch": 0.84,
      "learning_rate": 6.64168160924698e-06,
      "loss": 1.2093,
      "step": 595
    },
    {
      "epoch": 0.84,
      "learning_rate": 6.528265520250182e-06,
      "loss": 1.2403,
      "step": 596
    },
    {
      "epoch": 0.84,
      "learning_rate": 6.415758491115953e-06,
      "loss": 1.2749,
      "step": 597
    },
    {
      "epoch": 0.84,
      "learning_rate": 6.304162874536796e-06,
      "loss": 1.2668,
      "step": 598
    },
    {
      "epoch": 0.84,
      "learning_rate": 6.1934810041462066e-06,
      "loss": 1.2744,
      "step": 599
    },
    {
      "epoch": 0.85,
      "learning_rate": 6.08371519446988e-06,
      "loss": 1.2197,
      "step": 600
    },
    {
      "epoch": 0.85,
      "learning_rate": 5.974867740877283e-06,
      "loss": 1.248,
      "step": 601
    },
    {
      "epoch": 0.85,
      "learning_rate": 5.866940919533642e-06,
      "loss": 1.227,
      "step": 602
    },
    {
      "epoch": 0.85,
      "learning_rate": 5.759936987352399e-06,
      "loss": 1.2888,
      "step": 603
    },
    {
      "epoch": 0.85,
      "learning_rate": 5.65385818194798e-06,
      "loss": 1.2776,
      "step": 604
    },
    {
      "epoch": 0.85,
      "learning_rate": 5.548706721588986e-06,
      "loss": 1.2825,
      "step": 605
    },
    {
      "epoch": 0.85,
      "learning_rate": 5.444484805151856e-06,
      "loss": 1.2633,
      "step": 606
    },
    {
      "epoch": 0.86,
      "learning_rate": 5.341194612074824e-06,
      "loss": 1.2443,
      "step": 607
    },
    {
      "epoch": 0.86,
      "learning_rate": 5.23883830231241e-06,
      "loss": 1.247,
      "step": 608
    },
    {
      "epoch": 0.86,
      "learning_rate": 5.137418016290207e-06,
      "loss": 1.2748,
      "step": 609
    },
    {
      "epoch": 0.86,
      "learning_rate": 5.036935874860111e-06,
      "loss": 1.2337,
      "step": 610
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.937393979256016e-06,
      "loss": 1.2413,
      "step": 611
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.8387944110498685e-06,
      "loss": 1.2295,
      "step": 612
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.7411392321080605e-06,
      "loss": 1.2466,
      "step": 613
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.644430484548428e-06,
      "loss": 1.247,
      "step": 614
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.548670190697485e-06,
      "loss": 1.2306,
      "step": 615
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.453860353048112e-06,
      "loss": 1.2751,
      "step": 616
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.360002954217734e-06,
      "loss": 1.2279,
      "step": 617
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.267099956906828e-06,
      "loss": 1.2359,
      "step": 618
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.175153303857887e-06,
      "loss": 1.2383,
      "step": 619
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.084164917814815e-06,
      "loss": 1.2658,
      "step": 620
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.994136701482659e-06,
      "loss": 1.2115,
      "step": 621
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.905070537487909e-06,
      "loss": 1.2615,
      "step": 622
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.8169682883390565e-06,
      "loss": 1.1924,
      "step": 623
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.729831796387667e-06,
      "loss": 1.2593,
      "step": 624
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.643662883789878e-06,
      "loss": 1.2781,
      "step": 625
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.558463352468272e-06,
      "loss": 1.1864,
      "step": 626
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.474234984074182e-06,
      "loss": 1.2902,
      "step": 627
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.3909795399504783e-06,
      "loss": 1.2285,
      "step": 628
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.3086987610946807e-06,
      "loss": 1.2597,
      "step": 629
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.2273943681225992e-06,
      "loss": 1.2546,
      "step": 630
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.14706806123235e-06,
      "loss": 1.2337,
      "step": 631
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.067721520168748e-06,
      "loss": 1.214,
      "step": 632
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.9893564041882484e-06,
      "loss": 1.2604,
      "step": 633
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.9119743520242217e-06,
      "loss": 1.2261,
      "step": 634
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.835576981852656e-06,
      "loss": 1.2825,
      "step": 635
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7601658912583763e-06,
      "loss": 1.2606,
      "step": 636
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.685742657201601e-06,
      "loss": 1.2474,
      "step": 637
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.61230883598495e-06,
      "loss": 1.2088,
      "step": 638
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.5398659632209552e-06,
      "loss": 1.2612,
      "step": 639
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.4684155537998743e-06,
      "loss": 1.2175,
      "step": 640
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.397959101858083e-06,
      "loss": 1.2022,
      "step": 641
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.3284980807467994e-06,
      "loss": 1.219,
      "step": 642
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.2600339430012442e-06,
      "loss": 1.2495,
      "step": 643
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.1925681203103287e-06,
      "loss": 1.2155,
      "step": 644
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.12610202348667e-06,
      "loss": 1.2359,
      "step": 645
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.060637042437097e-06,
      "loss": 1.2291,
      "step": 646
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9961745461335947e-06,
      "loss": 1.2033,
      "step": 647
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9327158825846848e-06,
      "loss": 1.2563,
      "step": 648
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8702623788072027e-06,
      "loss": 1.2112,
      "step": 649
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8088153407985809e-06,
      "loss": 1.2475,
      "step": 650
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.7483760535095262e-06,
      "loss": 1.2246,
      "step": 651
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6889457808171472e-06,
      "loss": 1.2597,
      "step": 652
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6305257654985361e-06,
      "loss": 1.2363,
      "step": 653
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5731172292047625e-06,
      "loss": 1.2366,
      "step": 654
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5167213724353424e-06,
      "loss": 1.2134,
      "step": 655
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.4613393745131321e-06,
      "loss": 1.2548,
      "step": 656
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4069723935596412e-06,
      "loss": 1.2399,
      "step": 657
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3536215664708586e-06,
      "loss": 1.242,
      "step": 658
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3012880088934532e-06,
      "loss": 1.2886,
      "step": 659
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.2499728152014334e-06,
      "loss": 1.2667,
      "step": 660
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1996770584732919e-06,
      "loss": 1.28,
      "step": 661
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1504017904695296e-06,
      "loss": 1.2598,
      "step": 662
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1021480416106956e-06,
      "loss": 1.2444,
      "step": 663
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.0549168209558312e-06,
      "loss": 1.2365,
      "step": 664
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.008709116181339e-06,
      "loss": 1.235,
      "step": 665
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.635258935603796e-07,
      "loss": 1.2456,
      "step": 666
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.19368097942619e-07,
      "loss": 1.2448,
      "step": 667
    },
    {
      "epoch": 0.94,
      "learning_rate": 8.762366527345022e-07,
      "loss": 1.284,
      "step": 668
    },
    {
      "epoch": 0.94,
      "learning_rate": 8.341324598799216e-07,
      "loss": 1.2056,
      "step": 669
    },
    {
      "epoch": 0.94,
      "learning_rate": 7.930563998413798e-07,
      "loss": 1.175,
      "step": 670
    },
    {
      "epoch": 0.95,
      "learning_rate": 7.530093315815557e-07,
      "loss": 1.2085,
      "step": 671
    },
    {
      "epoch": 0.95,
      "learning_rate": 7.139920925453347e-07,
      "loss": 1.2548,
      "step": 672
    },
    {
      "epoch": 0.95,
      "learning_rate": 6.760054986423459e-07,
      "loss": 1.1877,
      "step": 673
    },
    {
      "epoch": 0.95,
      "learning_rate": 6.390503442298413e-07,
      "loss": 1.2495,
      "step": 674
    },
    {
      "epoch": 0.95,
      "learning_rate": 6.031274020961152e-07,
      "loss": 1.2678,
      "step": 675
    },
    {
      "epoch": 0.95,
      "learning_rate": 5.682374234443344e-07,
      "loss": 1.2285,
      "step": 676
    },
    {
      "epoch": 0.95,
      "learning_rate": 5.343811378768492e-07,
      "loss": 1.2401,
      "step": 677
    },
    {
      "epoch": 0.96,
      "learning_rate": 5.015592533799074e-07,
      "loss": 1.1871,
      "step": 678
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.6977245630886455e-07,
      "loss": 1.2252,
      "step": 679
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.3902141137382444e-07,
      "loss": 1.2074,
      "step": 680
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.0930676162576063e-07,
      "loss": 1.2778,
      "step": 681
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.8062912844302746e-07,
      "loss": 1.2601,
      "step": 682
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.5298911151841475e-07,
      "loss": 1.239,
      "step": 683
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.263872888465691e-07,
      "loss": 1.231,
      "step": 684
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.0082421671192575e-07,
      "loss": 1.2737,
      "step": 685
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.7630042967707327e-07,
      "loss": 1.29,
      "step": 686
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5281644057156826e-07,
      "loss": 1.248,
      "step": 687
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.303727404812217e-07,
      "loss": 1.2008,
      "step": 688
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.0896979873782918e-07,
      "loss": 1.1855,
      "step": 689
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8860806290932897e-07,
      "loss": 1.1995,
      "step": 690
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.692879587904983e-07,
      "loss": 1.245,
      "step": 691
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5100989039399939e-07,
      "loss": 1.2523,
      "step": 692
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.337742399419306e-07,
      "loss": 1.2624,
      "step": 693
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.1758136785788854e-07,
      "loss": 1.2609,
      "step": 694
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0243161275936274e-07,
      "loss": 1.2369,
      "step": 695
    },
    {
      "epoch": 0.98,
      "learning_rate": 8.832529145070267e-08,
      "loss": 1.2945,
      "step": 696
    },
    {
      "epoch": 0.98,
      "learning_rate": 7.526269891646176e-08,
      "loss": 1.2309,
      "step": 697
    },
    {
      "epoch": 0.98,
      "learning_rate": 6.324410831525795e-08,
      "loss": 1.2671,
      "step": 698
    },
    {
      "epoch": 0.99,
      "learning_rate": 5.226977097403385e-08,
      "loss": 1.2309,
      "step": 699
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.233991638281642e-08,
      "loss": 1.1994,
      "step": 700
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.3454752189926444e-08,
      "loss": 1.234,
      "step": 701
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.561446419760416e-08,
      "loss": 1.2064,
      "step": 702
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8819216358156864e-08,
      "loss": 1.1986,
      "step": 703
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.3069150770528282e-08,
      "loss": 1.2388,
      "step": 704
    },
    {
      "epoch": 0.99,
      "learning_rate": 8.364387677295415e-09,
      "loss": 1.2003,
      "step": 705
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.705025462187207e-09,
      "loss": 1.2338,
      "step": 706
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.091140648013967e-09,
      "loss": 1.264,
      "step": 707
    },
    {
      "epoch": 1.0,
      "learning_rate": 5.22787895074206e-10,
      "loss": 1.2477,
      "step": 708
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0,
      "loss": 1.208,
      "step": 709
    },
    {
      "epoch": 1.0,
      "step": 709,
      "total_flos": 0.0,
      "train_loss": 0.7294200115714995,
      "train_runtime": 11040.3966,
      "train_samples_per_second": 7.201,
      "train_steps_per_second": 0.064
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 709,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 100,
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}