{
  "best_metric": 0.7565978169441223,
  "best_model_checkpoint": "autotrain-80y7u-aoweu/checkpoint-1017",
  "epoch": 9.0,
  "eval_steps": 500,
  "global_step": 1017,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.017699115044247787,
      "grad_norm": Infinity,
      "learning_rate": 0.0,
      "loss": 1.8501,
      "step": 2
    },
    {
      "epoch": 0.035398230088495575,
      "grad_norm": 45.10062026977539,
      "learning_rate": 8.849557522123894e-07,
      "loss": 1.857,
      "step": 4
    },
    {
      "epoch": 0.05309734513274336,
      "grad_norm": 58.66633605957031,
      "learning_rate": 1.3274336283185841e-06,
      "loss": 1.8116,
      "step": 6
    },
    {
      "epoch": 0.07079646017699115,
      "grad_norm": 56.49775695800781,
      "learning_rate": 2.2123893805309734e-06,
      "loss": 2.0345,
      "step": 8
    },
    {
      "epoch": 0.08849557522123894,
      "grad_norm": 58.526424407958984,
      "learning_rate": 3.097345132743363e-06,
      "loss": 1.9499,
      "step": 10
    },
    {
      "epoch": 0.10619469026548672,
      "grad_norm": 52.64912796020508,
      "learning_rate": 3.982300884955752e-06,
      "loss": 1.9647,
      "step": 12
    },
    {
      "epoch": 0.12389380530973451,
      "grad_norm": 31.514617919921875,
      "learning_rate": 4.867256637168142e-06,
      "loss": 1.6022,
      "step": 14
    },
    {
      "epoch": 0.1415929203539823,
      "grad_norm": 39.92071533203125,
      "learning_rate": 5.752212389380531e-06,
      "loss": 1.7882,
      "step": 16
    },
    {
      "epoch": 0.1592920353982301,
      "grad_norm": 57.345947265625,
      "learning_rate": 6.6371681415929215e-06,
      "loss": 1.7493,
      "step": 18
    },
    {
      "epoch": 0.17699115044247787,
      "grad_norm": 40.93531799316406,
      "learning_rate": 7.52212389380531e-06,
      "loss": 1.5886,
      "step": 20
    },
    {
      "epoch": 0.19469026548672566,
      "grad_norm": 34.53605651855469,
      "learning_rate": 8.407079646017701e-06,
      "loss": 1.6985,
      "step": 22
    },
    {
      "epoch": 0.21238938053097345,
      "grad_norm": 59.89328384399414,
      "learning_rate": 9.29203539823009e-06,
      "loss": 1.8046,
      "step": 24
    },
    {
      "epoch": 0.23008849557522124,
      "grad_norm": 36.242183685302734,
      "learning_rate": 1.0176991150442479e-05,
      "loss": 1.5316,
      "step": 26
    },
    {
      "epoch": 0.24778761061946902,
      "grad_norm": 19.723634719848633,
      "learning_rate": 1.1061946902654869e-05,
      "loss": 1.612,
      "step": 28
    },
    {
      "epoch": 0.26548672566371684,
      "grad_norm": 33.05853271484375,
      "learning_rate": 1.1946902654867258e-05,
      "loss": 1.7078,
      "step": 30
    },
    {
      "epoch": 0.2831858407079646,
      "grad_norm": 25.069561004638672,
      "learning_rate": 1.2831858407079647e-05,
      "loss": 1.616,
      "step": 32
    },
    {
      "epoch": 0.3008849557522124,
      "grad_norm": 58.163509368896484,
      "learning_rate": 1.3716814159292036e-05,
      "loss": 1.6574,
      "step": 34
    },
    {
      "epoch": 0.3185840707964602,
      "grad_norm": 32.350101470947266,
      "learning_rate": 1.4601769911504426e-05,
      "loss": 1.3009,
      "step": 36
    },
    {
      "epoch": 0.336283185840708,
      "grad_norm": 36.51224899291992,
      "learning_rate": 1.5486725663716813e-05,
      "loss": 1.36,
      "step": 38
    },
    {
      "epoch": 0.35398230088495575,
      "grad_norm": 32.10768508911133,
      "learning_rate": 1.6371681415929206e-05,
      "loss": 1.515,
      "step": 40
    },
    {
      "epoch": 0.37168141592920356,
      "grad_norm": 25.605300903320312,
      "learning_rate": 1.7256637168141594e-05,
      "loss": 1.3663,
      "step": 42
    },
    {
      "epoch": 0.3893805309734513,
      "grad_norm": 146.0785369873047,
      "learning_rate": 1.8141592920353983e-05,
      "loss": 1.3139,
      "step": 44
    },
    {
      "epoch": 0.40707964601769914,
      "grad_norm": 37.9253044128418,
      "learning_rate": 1.9026548672566372e-05,
      "loss": 1.3613,
      "step": 46
    },
    {
      "epoch": 0.4247787610619469,
      "grad_norm": 36.18649673461914,
      "learning_rate": 1.991150442477876e-05,
      "loss": 1.4165,
      "step": 48
    },
    {
      "epoch": 0.4424778761061947,
      "grad_norm": 22.322145462036133,
      "learning_rate": 2.079646017699115e-05,
      "loss": 1.3432,
      "step": 50
    },
    {
      "epoch": 0.46017699115044247,
      "grad_norm": 28.047632217407227,
      "learning_rate": 2.1681415929203542e-05,
      "loss": 1.1527,
      "step": 52
    },
    {
      "epoch": 0.4778761061946903,
      "grad_norm": 34.25035858154297,
      "learning_rate": 2.2566371681415928e-05,
      "loss": 1.0954,
      "step": 54
    },
    {
      "epoch": 0.49557522123893805,
      "grad_norm": 154.13543701171875,
      "learning_rate": 2.345132743362832e-05,
      "loss": 1.289,
      "step": 56
    },
    {
      "epoch": 0.5132743362831859,
      "grad_norm": 34.313114166259766,
      "learning_rate": 2.433628318584071e-05,
      "loss": 1.4081,
      "step": 58
    },
    {
      "epoch": 0.5309734513274337,
      "grad_norm": 31.601171493530273,
      "learning_rate": 2.5221238938053098e-05,
      "loss": 1.1908,
      "step": 60
    },
    {
      "epoch": 0.5486725663716814,
      "grad_norm": 40.591888427734375,
      "learning_rate": 2.610619469026549e-05,
      "loss": 1.3728,
      "step": 62
    },
    {
      "epoch": 0.5663716814159292,
      "grad_norm": 39.01655578613281,
      "learning_rate": 2.6991150442477875e-05,
      "loss": 1.1882,
      "step": 64
    },
    {
      "epoch": 0.584070796460177,
      "grad_norm": 35.67778778076172,
      "learning_rate": 2.7876106194690264e-05,
      "loss": 0.9822,
      "step": 66
    },
    {
      "epoch": 0.6017699115044248,
      "grad_norm": 37.64238357543945,
      "learning_rate": 2.8761061946902656e-05,
      "loss": 0.9998,
      "step": 68
    },
    {
      "epoch": 0.6194690265486725,
      "grad_norm": 29.477062225341797,
      "learning_rate": 2.9203539823008852e-05,
      "loss": 1.3209,
      "step": 70
    },
    {
      "epoch": 0.6371681415929203,
      "grad_norm": 21.486434936523438,
      "learning_rate": 3.008849557522124e-05,
      "loss": 1.2007,
      "step": 72
    },
    {
      "epoch": 0.6548672566371682,
      "grad_norm": 32.44023513793945,
      "learning_rate": 3.097345132743363e-05,
      "loss": 1.2391,
      "step": 74
    },
    {
      "epoch": 0.672566371681416,
      "grad_norm": 87.39546203613281,
      "learning_rate": 3.185840707964602e-05,
      "loss": 1.2293,
      "step": 76
    },
    {
      "epoch": 0.6902654867256637,
      "grad_norm": 22.6693058013916,
      "learning_rate": 3.274336283185841e-05,
      "loss": 1.3051,
      "step": 78
    },
    {
      "epoch": 0.7079646017699115,
      "grad_norm": 22.473268508911133,
      "learning_rate": 3.3628318584070804e-05,
      "loss": 1.1976,
      "step": 80
    },
    {
      "epoch": 0.7256637168141593,
      "grad_norm": 27.98000144958496,
      "learning_rate": 3.451327433628319e-05,
      "loss": 1.172,
      "step": 82
    },
    {
      "epoch": 0.7433628318584071,
      "grad_norm": 30.333744049072266,
      "learning_rate": 3.5398230088495574e-05,
      "loss": 1.4081,
      "step": 84
    },
    {
      "epoch": 0.7610619469026548,
      "grad_norm": 32.933128356933594,
      "learning_rate": 3.628318584070797e-05,
      "loss": 1.0802,
      "step": 86
    },
    {
      "epoch": 0.7787610619469026,
      "grad_norm": 29.343582153320312,
      "learning_rate": 3.716814159292036e-05,
      "loss": 1.0651,
      "step": 88
    },
    {
      "epoch": 0.7964601769911505,
      "grad_norm": 28.437707901000977,
      "learning_rate": 3.8053097345132744e-05,
      "loss": 1.079,
      "step": 90
    },
    {
      "epoch": 0.8141592920353983,
      "grad_norm": 34.3442268371582,
      "learning_rate": 3.893805309734514e-05,
      "loss": 1.1715,
      "step": 92
    },
    {
      "epoch": 0.831858407079646,
      "grad_norm": 24.838363647460938,
      "learning_rate": 3.982300884955752e-05,
      "loss": 1.2568,
      "step": 94
    },
    {
      "epoch": 0.8495575221238938,
      "grad_norm": 19.97493553161621,
      "learning_rate": 4.0707964601769914e-05,
      "loss": 1.0946,
      "step": 96
    },
    {
      "epoch": 0.8672566371681416,
      "grad_norm": 26.681554794311523,
      "learning_rate": 4.15929203539823e-05,
      "loss": 0.9971,
      "step": 98
    },
    {
      "epoch": 0.8849557522123894,
      "grad_norm": 20.66066551208496,
      "learning_rate": 4.247787610619469e-05,
      "loss": 1.205,
      "step": 100
    },
    {
      "epoch": 0.9026548672566371,
      "grad_norm": 30.359079360961914,
      "learning_rate": 4.3362831858407084e-05,
      "loss": 1.0563,
      "step": 102
    },
    {
      "epoch": 0.9203539823008849,
      "grad_norm": 42.22723388671875,
      "learning_rate": 4.4247787610619477e-05,
      "loss": 1.0019,
      "step": 104
    },
    {
      "epoch": 0.9380530973451328,
      "grad_norm": 28.618349075317383,
      "learning_rate": 4.5132743362831855e-05,
      "loss": 1.1699,
      "step": 106
    },
    {
      "epoch": 0.9557522123893806,
      "grad_norm": 47.17839431762695,
      "learning_rate": 4.601769911504425e-05,
      "loss": 1.0793,
      "step": 108
    },
    {
      "epoch": 0.9734513274336283,
      "grad_norm": 28.66024398803711,
      "learning_rate": 4.690265486725664e-05,
      "loss": 1.1115,
      "step": 110
    },
    {
      "epoch": 0.9911504424778761,
      "grad_norm": 24.186906814575195,
      "learning_rate": 4.778761061946903e-05,
      "loss": 0.9751,
      "step": 112
    },
    {
      "epoch": 1.0,
      "eval_loss": 1.1276780366897583,
      "eval_map": 0.1062,
      "eval_map_50": 0.2136,
      "eval_map_75": 0.0953,
      "eval_map_large": 0.1451,
      "eval_map_medium": 0.0275,
      "eval_map_per_class": 0.1062,
      "eval_map_small": 0.0,
      "eval_mar_1": 0.1671,
      "eval_mar_10": 0.3786,
      "eval_mar_100": 0.6103,
      "eval_mar_100_per_class": 0.6103,
      "eval_mar_large": 0.6775,
      "eval_mar_medium": 0.4422,
      "eval_mar_small": 0.0,
      "eval_runtime": 4.6231,
      "eval_samples_per_second": 21.631,
      "eval_steps_per_second": 1.514,
      "step": 113
    },
    {
      "epoch": 1.008849557522124,
      "grad_norm": 29.109996795654297,
      "learning_rate": 4.867256637168142e-05,
      "loss": 1.0909,
      "step": 114
    },
    {
      "epoch": 1.0265486725663717,
      "grad_norm": 36.879417419433594,
      "learning_rate": 4.955752212389381e-05,
      "loss": 1.0213,
      "step": 116
    },
    {
      "epoch": 1.0442477876106195,
      "grad_norm": 25.578012466430664,
      "learning_rate": 4.9950835791543757e-05,
      "loss": 1.0314,
      "step": 118
    },
    {
      "epoch": 1.0619469026548674,
      "grad_norm": 20.469444274902344,
      "learning_rate": 4.985250737463127e-05,
      "loss": 1.0589,
      "step": 120
    },
    {
      "epoch": 1.079646017699115,
      "grad_norm": 18.58279037475586,
      "learning_rate": 4.9754178957718786e-05,
      "loss": 1.0507,
      "step": 122
    },
    {
      "epoch": 1.0973451327433628,
      "grad_norm": 29.6292724609375,
      "learning_rate": 4.9655850540806295e-05,
      "loss": 1.0577,
      "step": 124
    },
    {
      "epoch": 1.1150442477876106,
      "grad_norm": 57.019222259521484,
      "learning_rate": 4.955752212389381e-05,
      "loss": 1.0879,
      "step": 126
    },
    {
      "epoch": 1.1327433628318584,
      "grad_norm": 39.2066650390625,
      "learning_rate": 4.9459193706981325e-05,
      "loss": 1.1121,
      "step": 128
    },
    {
      "epoch": 1.1504424778761062,
      "grad_norm": 26.280122756958008,
      "learning_rate": 4.936086529006883e-05,
      "loss": 1.001,
      "step": 130
    },
    {
      "epoch": 1.168141592920354,
      "grad_norm": 41.13880920410156,
      "learning_rate": 4.926253687315635e-05,
      "loss": 1.0486,
      "step": 132
    },
    {
      "epoch": 1.1858407079646018,
      "grad_norm": 45.9874267578125,
      "learning_rate": 4.9164208456243856e-05,
      "loss": 1.0237,
      "step": 134
    },
    {
      "epoch": 1.2035398230088497,
      "grad_norm": 31.62242317199707,
      "learning_rate": 4.906588003933137e-05,
      "loss": 1.0973,
      "step": 136
    },
    {
      "epoch": 1.2212389380530975,
      "grad_norm": 30.55558967590332,
      "learning_rate": 4.8967551622418886e-05,
      "loss": 1.0251,
      "step": 138
    },
    {
      "epoch": 1.238938053097345,
      "grad_norm": 40.2532844543457,
      "learning_rate": 4.8869223205506394e-05,
      "loss": 0.9459,
      "step": 140
    },
    {
      "epoch": 1.2566371681415929,
      "grad_norm": 77.14492797851562,
      "learning_rate": 4.877089478859391e-05,
      "loss": 1.0721,
      "step": 142
    },
    {
      "epoch": 1.2743362831858407,
      "grad_norm": 30.2625732421875,
      "learning_rate": 4.867256637168142e-05,
      "loss": 1.0249,
      "step": 144
    },
    {
      "epoch": 1.2920353982300885,
      "grad_norm": 39.786590576171875,
      "learning_rate": 4.857423795476893e-05,
      "loss": 1.0735,
      "step": 146
    },
    {
      "epoch": 1.3097345132743363,
      "grad_norm": 33.97883605957031,
      "learning_rate": 4.847590953785645e-05,
      "loss": 0.9914,
      "step": 148
    },
    {
      "epoch": 1.3274336283185841,
      "grad_norm": 27.49061393737793,
      "learning_rate": 4.8377581120943956e-05,
      "loss": 1.2146,
      "step": 150
    },
    {
      "epoch": 1.3451327433628317,
      "grad_norm": 25.95128059387207,
      "learning_rate": 4.827925270403147e-05,
      "loss": 1.048,
      "step": 152
    },
    {
      "epoch": 1.3628318584070795,
      "grad_norm": 25.32373809814453,
      "learning_rate": 4.818092428711898e-05,
      "loss": 1.0952,
      "step": 154
    },
    {
      "epoch": 1.3805309734513274,
      "grad_norm": 29.068500518798828,
      "learning_rate": 4.8082595870206494e-05,
      "loss": 1.0903,
      "step": 156
    },
    {
      "epoch": 1.3982300884955752,
      "grad_norm": 30.730670928955078,
      "learning_rate": 4.7984267453294e-05,
      "loss": 1.0388,
      "step": 158
    },
    {
      "epoch": 1.415929203539823,
      "grad_norm": 34.4339714050293,
      "learning_rate": 4.788593903638152e-05,
      "loss": 1.0369,
      "step": 160
    },
    {
      "epoch": 1.4336283185840708,
      "grad_norm": 74.3895263671875,
      "learning_rate": 4.778761061946903e-05,
      "loss": 1.0317,
      "step": 162
    },
    {
      "epoch": 1.4513274336283186,
      "grad_norm": 25.546483993530273,
      "learning_rate": 4.768928220255654e-05,
      "loss": 0.9741,
      "step": 164
    },
    {
      "epoch": 1.4690265486725664,
      "grad_norm": 26.646549224853516,
      "learning_rate": 4.7590953785644055e-05,
      "loss": 1.1132,
      "step": 166
    },
    {
      "epoch": 1.4867256637168142,
      "grad_norm": 27.06178092956543,
      "learning_rate": 4.749262536873156e-05,
      "loss": 0.9592,
      "step": 168
    },
    {
      "epoch": 1.504424778761062,
      "grad_norm": 27.895265579223633,
      "learning_rate": 4.739429695181908e-05,
      "loss": 1.0061,
      "step": 170
    },
    {
      "epoch": 1.5221238938053099,
      "grad_norm": 28.37946319580078,
      "learning_rate": 4.729596853490659e-05,
      "loss": 0.9069,
      "step": 172
    },
    {
      "epoch": 1.5398230088495575,
      "grad_norm": 22.880842208862305,
      "learning_rate": 4.71976401179941e-05,
      "loss": 1.0585,
      "step": 174
    },
    {
      "epoch": 1.5575221238938053,
      "grad_norm": 27.391437530517578,
      "learning_rate": 4.7099311701081617e-05,
      "loss": 1.1805,
      "step": 176
    },
    {
      "epoch": 1.575221238938053,
      "grad_norm": 76.33792877197266,
      "learning_rate": 4.7000983284169125e-05,
      "loss": 0.93,
      "step": 178
    },
    {
      "epoch": 1.592920353982301,
      "grad_norm": 32.36155700683594,
      "learning_rate": 4.690265486725664e-05,
      "loss": 1.066,
      "step": 180
    },
    {
      "epoch": 1.6106194690265485,
      "grad_norm": 23.659805297851562,
      "learning_rate": 4.680432645034415e-05,
      "loss": 0.9452,
      "step": 182
    },
    {
      "epoch": 1.6283185840707963,
      "grad_norm": 40.60184097290039,
      "learning_rate": 4.670599803343166e-05,
      "loss": 0.9825,
      "step": 184
    },
    {
      "epoch": 1.6460176991150441,
      "grad_norm": 47.14199447631836,
      "learning_rate": 4.660766961651918e-05,
      "loss": 0.9014,
      "step": 186
    },
    {
      "epoch": 1.663716814159292,
      "grad_norm": 26.055511474609375,
      "learning_rate": 4.6509341199606686e-05,
      "loss": 1.0034,
      "step": 188
    },
    {
      "epoch": 1.6814159292035398,
      "grad_norm": 31.73177146911621,
      "learning_rate": 4.64110127826942e-05,
      "loss": 1.0716,
      "step": 190
    },
    {
      "epoch": 1.6991150442477876,
      "grad_norm": 28.099885940551758,
      "learning_rate": 4.631268436578171e-05,
      "loss": 1.1829,
      "step": 192
    },
    {
      "epoch": 1.7168141592920354,
      "grad_norm": 13.378188133239746,
      "learning_rate": 4.6214355948869224e-05,
      "loss": 0.9111,
      "step": 194
    },
    {
      "epoch": 1.7345132743362832,
      "grad_norm": 22.647581100463867,
      "learning_rate": 4.611602753195674e-05,
      "loss": 1.0079,
      "step": 196
    },
    {
      "epoch": 1.752212389380531,
      "grad_norm": 22.10202407836914,
      "learning_rate": 4.601769911504425e-05,
      "loss": 1.0541,
      "step": 198
    },
    {
      "epoch": 1.7699115044247788,
      "grad_norm": 19.46767807006836,
      "learning_rate": 4.591937069813176e-05,
      "loss": 0.8407,
      "step": 200
    },
    {
      "epoch": 1.7876106194690267,
      "grad_norm": 34.163822174072266,
      "learning_rate": 4.582104228121927e-05,
      "loss": 1.157,
      "step": 202
    },
    {
      "epoch": 1.8053097345132745,
      "grad_norm": 24.074146270751953,
      "learning_rate": 4.5722713864306786e-05,
      "loss": 0.9934,
      "step": 204
    },
    {
      "epoch": 1.823008849557522,
      "grad_norm": 150.62364196777344,
      "learning_rate": 4.5624385447394294e-05,
      "loss": 0.9615,
      "step": 206
    },
    {
      "epoch": 1.8407079646017699,
      "grad_norm": 20.424564361572266,
      "learning_rate": 4.552605703048181e-05,
      "loss": 1.0251,
      "step": 208
    },
    {
      "epoch": 1.8584070796460177,
      "grad_norm": 30.88136100769043,
      "learning_rate": 4.5427728613569324e-05,
      "loss": 0.9429,
      "step": 210
    },
    {
      "epoch": 1.8761061946902655,
      "grad_norm": 25.247404098510742,
      "learning_rate": 4.532940019665683e-05,
      "loss": 0.9711,
      "step": 212
    },
    {
      "epoch": 1.893805309734513,
      "grad_norm": 37.985774993896484,
      "learning_rate": 4.523107177974435e-05,
      "loss": 0.9705,
      "step": 214
    },
    {
      "epoch": 1.911504424778761,
      "grad_norm": 22.748384475708008,
      "learning_rate": 4.5132743362831855e-05,
      "loss": 1.0017,
      "step": 216
    },
    {
      "epoch": 1.9292035398230087,
      "grad_norm": 19.267290115356445,
      "learning_rate": 4.503441494591937e-05,
      "loss": 0.9072,
      "step": 218
    },
    {
      "epoch": 1.9469026548672566,
      "grad_norm": 19.34298324584961,
      "learning_rate": 4.4936086529006885e-05,
      "loss": 1.0597,
      "step": 220
    },
    {
      "epoch": 1.9646017699115044,
      "grad_norm": 23.347171783447266,
      "learning_rate": 4.48377581120944e-05,
      "loss": 0.9139,
      "step": 222
    },
    {
      "epoch": 1.9823008849557522,
      "grad_norm": 20.579362869262695,
      "learning_rate": 4.473942969518191e-05,
      "loss": 0.8591,
      "step": 224
    },
    {
      "epoch": 2.0,
      "grad_norm": 64.1324691772461,
      "learning_rate": 4.464110127826942e-05,
      "loss": 1.0063,
      "step": 226
    },
    {
      "epoch": 2.0,
      "eval_loss": 0.9410255551338196,
      "eval_map": 0.2993,
      "eval_map_50": 0.536,
      "eval_map_75": 0.3136,
      "eval_map_large": 0.3631,
      "eval_map_medium": 0.1534,
      "eval_map_per_class": 0.2993,
      "eval_map_small": 0.0,
      "eval_mar_1": 0.2427,
      "eval_mar_10": 0.5261,
      "eval_mar_100": 0.6393,
      "eval_mar_100_per_class": 0.6393,
      "eval_mar_large": 0.7107,
      "eval_mar_medium": 0.4609,
      "eval_mar_small": 0.0,
      "eval_runtime": 3.9582,
      "eval_samples_per_second": 25.264,
      "eval_steps_per_second": 1.768,
      "step": 226
    },
    {
      "epoch": 2.017699115044248,
      "grad_norm": 27.324804306030273,
      "learning_rate": 4.454277286135694e-05,
      "loss": 0.9027,
      "step": 228
    },
    {
      "epoch": 2.0353982300884956,
      "grad_norm": 20.43166732788086,
      "learning_rate": 4.4444444444444447e-05,
      "loss": 0.9257,
      "step": 230
    },
    {
      "epoch": 2.0530973451327434,
      "grad_norm": 18.177186965942383,
      "learning_rate": 4.434611602753196e-05,
      "loss": 0.9379,
      "step": 232
    },
    {
      "epoch": 2.0707964601769913,
      "grad_norm": 24.545679092407227,
      "learning_rate": 4.4247787610619477e-05,
      "loss": 1.0571,
      "step": 234
    },
    {
      "epoch": 2.088495575221239,
      "grad_norm": 21.458559036254883,
      "learning_rate": 4.4149459193706985e-05,
      "loss": 1.0239,
      "step": 236
    },
    {
      "epoch": 2.106194690265487,
      "grad_norm": 20.25364112854004,
      "learning_rate": 4.40511307767945e-05,
      "loss": 0.9194,
      "step": 238
    },
    {
      "epoch": 2.1238938053097347,
      "grad_norm": 24.264408111572266,
      "learning_rate": 4.395280235988201e-05,
      "loss": 0.9601,
      "step": 240
    },
    {
      "epoch": 2.1415929203539825,
      "grad_norm": 17.67040252685547,
      "learning_rate": 4.385447394296952e-05,
      "loss": 1.1069,
      "step": 242
    },
    {
      "epoch": 2.15929203539823,
      "grad_norm": 25.185571670532227,
      "learning_rate": 4.375614552605704e-05,
      "loss": 0.9537,
      "step": 244
    },
    {
      "epoch": 2.1769911504424777,
      "grad_norm": 39.48079299926758,
      "learning_rate": 4.3657817109144546e-05,
      "loss": 1.082,
      "step": 246
    },
    {
      "epoch": 2.1946902654867255,
      "grad_norm": 22.608633041381836,
      "learning_rate": 4.355948869223206e-05,
      "loss": 0.8366,
      "step": 248
    },
    {
      "epoch": 2.2123893805309733,
      "grad_norm": 57.35368347167969,
      "learning_rate": 4.346116027531957e-05,
      "loss": 0.8564,
      "step": 250
    },
    {
      "epoch": 2.230088495575221,
      "grad_norm": 34.98164367675781,
      "learning_rate": 4.3362831858407084e-05,
      "loss": 0.8928,
      "step": 252
    },
    {
      "epoch": 2.247787610619469,
      "grad_norm": 19.134063720703125,
      "learning_rate": 4.326450344149459e-05,
      "loss": 0.9219,
      "step": 254
    },
    {
      "epoch": 2.265486725663717,
      "grad_norm": 33.00546646118164,
      "learning_rate": 4.316617502458211e-05,
      "loss": 0.8816,
      "step": 256
    },
    {
      "epoch": 2.2831858407079646,
      "grad_norm": 26.70623779296875,
      "learning_rate": 4.306784660766962e-05,
      "loss": 0.9003,
      "step": 258
    },
    {
      "epoch": 2.3008849557522124,
      "grad_norm": 16.002517700195312,
      "learning_rate": 4.296951819075713e-05,
      "loss": 0.9086,
      "step": 260
    },
    {
      "epoch": 2.3185840707964602,
      "grad_norm": 26.876773834228516,
      "learning_rate": 4.2871189773844646e-05,
      "loss": 0.8936,
      "step": 262
    },
    {
      "epoch": 2.336283185840708,
      "grad_norm": 19.438966751098633,
      "learning_rate": 4.2772861356932154e-05,
      "loss": 0.886,
      "step": 264
    },
    {
      "epoch": 2.353982300884956,
      "grad_norm": 18.447446823120117,
      "learning_rate": 4.267453294001967e-05,
      "loss": 0.9216,
      "step": 266
    },
    {
      "epoch": 2.3716814159292037,
      "grad_norm": 17.754484176635742,
      "learning_rate": 4.2576204523107184e-05,
      "loss": 0.8503,
      "step": 268
    },
    {
      "epoch": 2.3893805309734515,
      "grad_norm": 21.90740394592285,
      "learning_rate": 4.247787610619469e-05,
      "loss": 0.8937,
      "step": 270
    },
    {
      "epoch": 2.4070796460176993,
      "grad_norm": 17.798200607299805,
      "learning_rate": 4.237954768928221e-05,
      "loss": 0.97,
      "step": 272
    },
    {
      "epoch": 2.4247787610619467,
      "grad_norm": 26.43834114074707,
      "learning_rate": 4.2281219272369715e-05,
      "loss": 0.7791,
      "step": 274
    },
    {
      "epoch": 2.442477876106195,
      "grad_norm": 18.365053176879883,
      "learning_rate": 4.218289085545723e-05,
      "loss": 0.9906,
      "step": 276
    },
    {
      "epoch": 2.4601769911504423,
      "grad_norm": 26.327442169189453,
      "learning_rate": 4.208456243854474e-05,
      "loss": 1.001,
      "step": 278
    },
    {
      "epoch": 2.47787610619469,
      "grad_norm": 22.054363250732422,
      "learning_rate": 4.1986234021632253e-05,
      "loss": 1.0239,
      "step": 280
    },
    {
      "epoch": 2.495575221238938,
      "grad_norm": 39.17606735229492,
      "learning_rate": 4.188790560471977e-05,
      "loss": 0.9184,
      "step": 282
    },
    {
      "epoch": 2.5132743362831858,
      "grad_norm": 18.545665740966797,
      "learning_rate": 4.178957718780728e-05,
      "loss": 0.8107,
      "step": 284
    },
    {
      "epoch": 2.5309734513274336,
      "grad_norm": 15.98513412475586,
      "learning_rate": 4.169124877089479e-05,
      "loss": 1.0371,
      "step": 286
    },
    {
      "epoch": 2.5486725663716814,
      "grad_norm": 23.60863494873047,
      "learning_rate": 4.15929203539823e-05,
      "loss": 0.8471,
      "step": 288
    },
    {
      "epoch": 2.566371681415929,
      "grad_norm": 39.60649108886719,
      "learning_rate": 4.1494591937069815e-05,
      "loss": 0.9184,
      "step": 290
    },
    {
      "epoch": 2.584070796460177,
      "grad_norm": 32.43935012817383,
      "learning_rate": 4.139626352015733e-05,
      "loss": 0.8087,
      "step": 292
    },
    {
      "epoch": 2.601769911504425,
      "grad_norm": 22.511394500732422,
      "learning_rate": 4.129793510324484e-05,
      "loss": 0.8604,
      "step": 294
    },
    {
      "epoch": 2.6194690265486726,
      "grad_norm": 17.795698165893555,
      "learning_rate": 4.119960668633235e-05,
      "loss": 0.9068,
      "step": 296
    },
    {
      "epoch": 2.6371681415929205,
      "grad_norm": 25.883586883544922,
      "learning_rate": 4.110127826941986e-05,
      "loss": 0.7654,
      "step": 298
    },
    {
      "epoch": 2.6548672566371683,
      "grad_norm": 17.679630279541016,
      "learning_rate": 4.1002949852507376e-05,
      "loss": 0.8956,
      "step": 300
    },
    {
      "epoch": 2.672566371681416,
      "grad_norm": 40.860904693603516,
      "learning_rate": 4.0904621435594884e-05,
      "loss": 0.9776,
      "step": 302
    },
    {
      "epoch": 2.6902654867256635,
      "grad_norm": 30.485763549804688,
      "learning_rate": 4.08062930186824e-05,
      "loss": 0.7637,
      "step": 304
    },
    {
      "epoch": 2.7079646017699117,
      "grad_norm": 20.94835090637207,
      "learning_rate": 4.0707964601769914e-05,
      "loss": 0.972,
      "step": 306
    },
    {
      "epoch": 2.725663716814159,
      "grad_norm": 28.85026741027832,
      "learning_rate": 4.060963618485742e-05,
      "loss": 1.0257,
      "step": 308
    },
    {
      "epoch": 2.7433628318584073,
      "grad_norm": 20.580015182495117,
      "learning_rate": 4.051130776794494e-05,
      "loss": 0.9242,
      "step": 310
    },
    {
      "epoch": 2.7610619469026547,
      "grad_norm": 19.40909767150879,
      "learning_rate": 4.0412979351032446e-05,
      "loss": 1.0149,
      "step": 312
    },
    {
      "epoch": 2.7787610619469025,
      "grad_norm": 17.306678771972656,
      "learning_rate": 4.031465093411996e-05,
      "loss": 0.8118,
      "step": 314
    },
    {
      "epoch": 2.7964601769911503,
      "grad_norm": 30.43448829650879,
      "learning_rate": 4.0216322517207476e-05,
      "loss": 0.8623,
      "step": 316
    },
    {
      "epoch": 2.814159292035398,
      "grad_norm": 20.09423828125,
      "learning_rate": 4.0117994100294984e-05,
      "loss": 0.8364,
      "step": 318
    },
    {
      "epoch": 2.831858407079646,
      "grad_norm": 16.069490432739258,
      "learning_rate": 4.00196656833825e-05,
      "loss": 0.812,
      "step": 320
    },
    {
      "epoch": 2.849557522123894,
      "grad_norm": 25.968656539916992,
      "learning_rate": 3.992133726647001e-05,
      "loss": 0.8953,
      "step": 322
    },
    {
      "epoch": 2.8672566371681416,
      "grad_norm": 28.93232536315918,
      "learning_rate": 3.982300884955752e-05,
      "loss": 1.0083,
      "step": 324
    },
    {
      "epoch": 2.8849557522123894,
      "grad_norm": 32.24972152709961,
      "learning_rate": 3.972468043264504e-05,
      "loss": 0.9767,
      "step": 326
    },
    {
      "epoch": 2.9026548672566372,
      "grad_norm": 32.170352935791016,
      "learning_rate": 3.9626352015732545e-05,
      "loss": 0.787,
      "step": 328
    },
    {
      "epoch": 2.920353982300885,
      "grad_norm": 19.10132598876953,
      "learning_rate": 3.952802359882006e-05,
      "loss": 0.8321,
      "step": 330
    },
    {
      "epoch": 2.938053097345133,
      "grad_norm": 30.334962844848633,
      "learning_rate": 3.9429695181907575e-05,
      "loss": 0.8306,
      "step": 332
    },
    {
      "epoch": 2.9557522123893807,
      "grad_norm": 24.146038055419922,
      "learning_rate": 3.9331366764995083e-05,
      "loss": 0.8859,
      "step": 334
    },
    {
      "epoch": 2.9734513274336285,
      "grad_norm": 24.66895294189453,
      "learning_rate": 3.92330383480826e-05,
      "loss": 0.8035,
      "step": 336
    },
    {
      "epoch": 2.991150442477876,
      "grad_norm": 18.740558624267578,
      "learning_rate": 3.9134709931170113e-05,
      "loss": 0.8284,
      "step": 338
    },
    {
      "epoch": 3.0,
      "eval_loss": 0.883878231048584,
      "eval_map": 0.4451,
      "eval_map_50": 0.8057,
      "eval_map_75": 0.399,
      "eval_map_large": 0.5228,
      "eval_map_medium": 0.2384,
      "eval_map_per_class": 0.4451,
      "eval_map_small": 0.0123,
      "eval_mar_1": 0.2662,
      "eval_mar_10": 0.5637,
      "eval_mar_100": 0.6295,
      "eval_mar_100_per_class": 0.6295,
      "eval_mar_large": 0.6976,
      "eval_mar_medium": 0.4547,
      "eval_mar_small": 0.3,
      "eval_runtime": 4.2753,
      "eval_samples_per_second": 23.39,
      "eval_steps_per_second": 1.637,
      "step": 339
    },
    {
      "epoch": 3.0088495575221237,
      "grad_norm": 17.387792587280273,
      "learning_rate": 3.903638151425762e-05,
      "loss": 0.9204,
      "step": 340
    },
    {
      "epoch": 3.0265486725663715,
      "grad_norm": 23.816524505615234,
      "learning_rate": 3.893805309734514e-05,
      "loss": 0.7563,
      "step": 342
    },
    {
      "epoch": 3.0442477876106193,
      "grad_norm": 23.070165634155273,
      "learning_rate": 3.883972468043265e-05,
      "loss": 0.6889,
      "step": 344
    },
    {
      "epoch": 3.061946902654867,
      "grad_norm": 21.282630920410156,
      "learning_rate": 3.874139626352016e-05,
      "loss": 0.7727,
      "step": 346
    },
    {
      "epoch": 3.079646017699115,
      "grad_norm": 32.04922866821289,
      "learning_rate": 3.8643067846607675e-05,
      "loss": 0.7494,
      "step": 348
    },
    {
      "epoch": 3.0973451327433628,
      "grad_norm": 22.994802474975586,
      "learning_rate": 3.854473942969518e-05,
      "loss": 0.9539,
      "step": 350
    },
    {
      "epoch": 3.1150442477876106,
      "grad_norm": 19.214599609375,
      "learning_rate": 3.84464110127827e-05,
      "loss": 0.6915,
      "step": 352
    },
    {
      "epoch": 3.1327433628318584,
      "grad_norm": 15.473613739013672,
      "learning_rate": 3.834808259587021e-05,
      "loss": 0.8087,
      "step": 354
    },
    {
      "epoch": 3.150442477876106,
      "grad_norm": 27.318443298339844,
      "learning_rate": 3.824975417895772e-05,
      "loss": 0.7237,
      "step": 356
    },
    {
      "epoch": 3.168141592920354,
      "grad_norm": 20.30460548400879,
      "learning_rate": 3.8151425762045236e-05,
      "loss": 0.8245,
      "step": 358
    },
    {
      "epoch": 3.185840707964602,
      "grad_norm": 20.886333465576172,
      "learning_rate": 3.8053097345132744e-05,
      "loss": 0.9283,
      "step": 360
    },
    {
      "epoch": 3.2035398230088497,
      "grad_norm": 19.330215454101562,
      "learning_rate": 3.795476892822026e-05,
      "loss": 0.8195,
      "step": 362
    },
    {
      "epoch": 3.2212389380530975,
      "grad_norm": 28.00650405883789,
      "learning_rate": 3.7856440511307774e-05,
      "loss": 0.8749,
      "step": 364
    },
    {
      "epoch": 3.2389380530973453,
      "grad_norm": 23.165494918823242,
      "learning_rate": 3.775811209439528e-05,
      "loss": 0.8077,
      "step": 366
    },
    {
      "epoch": 3.256637168141593,
      "grad_norm": 24.138504028320312,
      "learning_rate": 3.76597836774828e-05,
      "loss": 0.7378,
      "step": 368
    },
    {
      "epoch": 3.274336283185841,
      "grad_norm": 26.703155517578125,
      "learning_rate": 3.7561455260570306e-05,
      "loss": 0.771,
      "step": 370
    },
    {
      "epoch": 3.2920353982300883,
      "grad_norm": 19.84184455871582,
      "learning_rate": 3.746312684365782e-05,
      "loss": 0.8013,
      "step": 372
    },
    {
      "epoch": 3.309734513274336,
      "grad_norm": 23.238510131835938,
      "learning_rate": 3.736479842674533e-05,
      "loss": 0.7144,
      "step": 374
    },
    {
      "epoch": 3.327433628318584,
      "grad_norm": 27.819055557250977,
      "learning_rate": 3.7266470009832844e-05,
      "loss": 0.9366,
      "step": 376
    },
    {
      "epoch": 3.3451327433628317,
      "grad_norm": 18.144052505493164,
      "learning_rate": 3.716814159292036e-05,
      "loss": 0.8564,
      "step": 378
    },
    {
      "epoch": 3.3628318584070795,
      "grad_norm": 16.014623641967773,
      "learning_rate": 3.706981317600787e-05,
      "loss": 0.9854,
      "step": 380
    },
    {
      "epoch": 3.3805309734513274,
      "grad_norm": 19.18610191345215,
      "learning_rate": 3.697148475909538e-05,
      "loss": 0.8153,
      "step": 382
    },
    {
      "epoch": 3.398230088495575,
      "grad_norm": 28.12213134765625,
      "learning_rate": 3.687315634218289e-05,
      "loss": 0.8107,
      "step": 384
    },
    {
      "epoch": 3.415929203539823,
      "grad_norm": 42.245262145996094,
      "learning_rate": 3.6774827925270405e-05,
      "loss": 0.8498,
      "step": 386
    },
    {
      "epoch": 3.433628318584071,
      "grad_norm": 20.642005920410156,
      "learning_rate": 3.667649950835792e-05,
      "loss": 0.8426,
      "step": 388
    },
    {
      "epoch": 3.4513274336283186,
      "grad_norm": 20.72900390625,
      "learning_rate": 3.657817109144543e-05,
      "loss": 0.7979,
      "step": 390
    },
    {
      "epoch": 3.4690265486725664,
      "grad_norm": 36.909027099609375,
      "learning_rate": 3.6479842674532944e-05,
      "loss": 1.0862,
      "step": 392
    },
    {
      "epoch": 3.4867256637168142,
      "grad_norm": 34.34043884277344,
      "learning_rate": 3.638151425762045e-05,
      "loss": 0.9602,
      "step": 394
    },
    {
      "epoch": 3.504424778761062,
      "grad_norm": 17.44400978088379,
      "learning_rate": 3.628318584070797e-05,
      "loss": 0.7522,
      "step": 396
    },
    {
      "epoch": 3.52212389380531,
      "grad_norm": 29.59943962097168,
      "learning_rate": 3.6184857423795475e-05,
      "loss": 0.8881,
      "step": 398
    },
    {
      "epoch": 3.5398230088495577,
      "grad_norm": 33.33137893676758,
      "learning_rate": 3.608652900688299e-05,
      "loss": 0.9806,
      "step": 400
    },
    {
      "epoch": 3.557522123893805,
      "grad_norm": 25.366241455078125,
      "learning_rate": 3.5988200589970505e-05,
      "loss": 1.0134,
      "step": 402
    },
    {
      "epoch": 3.5752212389380533,
      "grad_norm": 28.731691360473633,
      "learning_rate": 3.588987217305801e-05,
      "loss": 0.926,
      "step": 404
    },
    {
      "epoch": 3.5929203539823007,
      "grad_norm": 24.277082443237305,
      "learning_rate": 3.579154375614553e-05,
      "loss": 0.8202,
      "step": 406
    },
    {
      "epoch": 3.6106194690265485,
      "grad_norm": 15.626072883605957,
      "learning_rate": 3.5693215339233036e-05,
      "loss": 0.8415,
      "step": 408
    },
    {
      "epoch": 3.6283185840707963,
      "grad_norm": 35.861610412597656,
      "learning_rate": 3.559488692232055e-05,
      "loss": 0.8739,
      "step": 410
    },
    {
      "epoch": 3.646017699115044,
      "grad_norm": 30.36337661743164,
      "learning_rate": 3.5496558505408066e-05,
      "loss": 0.7045,
      "step": 412
    },
    {
      "epoch": 3.663716814159292,
      "grad_norm": 25.959348678588867,
      "learning_rate": 3.5398230088495574e-05,
      "loss": 0.6826,
      "step": 414
    },
    {
      "epoch": 3.6814159292035398,
      "grad_norm": 20.39558219909668,
      "learning_rate": 3.529990167158309e-05,
      "loss": 0.8847,
      "step": 416
    },
    {
      "epoch": 3.6991150442477876,
      "grad_norm": 13.092119216918945,
      "learning_rate": 3.52015732546706e-05,
      "loss": 1.0101,
      "step": 418
    },
    {
      "epoch": 3.7168141592920354,
      "grad_norm": 32.8824462890625,
      "learning_rate": 3.510324483775811e-05,
      "loss": 0.8229,
      "step": 420
    },
    {
      "epoch": 3.734513274336283,
      "grad_norm": 52.47900390625,
      "learning_rate": 3.500491642084562e-05,
      "loss": 0.8438,
      "step": 422
    },
    {
      "epoch": 3.752212389380531,
      "grad_norm": 40.38644790649414,
      "learning_rate": 3.4906588003933136e-05,
      "loss": 0.8129,
      "step": 424
    },
    {
      "epoch": 3.769911504424779,
      "grad_norm": 24.974903106689453,
      "learning_rate": 3.480825958702065e-05,
      "loss": 0.8478,
      "step": 426
    },
    {
      "epoch": 3.7876106194690267,
      "grad_norm": 26.479198455810547,
      "learning_rate": 3.470993117010816e-05,
      "loss": 0.8803,
      "step": 428
    },
    {
      "epoch": 3.8053097345132745,
      "grad_norm": 20.01055908203125,
      "learning_rate": 3.4611602753195674e-05,
      "loss": 0.8078,
      "step": 430
    },
    {
      "epoch": 3.823008849557522,
      "grad_norm": 34.75425720214844,
      "learning_rate": 3.451327433628319e-05,
      "loss": 0.9117,
      "step": 432
    },
    {
      "epoch": 3.84070796460177,
      "grad_norm": 30.763608932495117,
      "learning_rate": 3.44149459193707e-05,
      "loss": 0.7776,
      "step": 434
    },
    {
      "epoch": 3.8584070796460175,
      "grad_norm": 38.77566909790039,
      "learning_rate": 3.431661750245821e-05,
      "loss": 0.8284,
      "step": 436
    },
    {
      "epoch": 3.8761061946902657,
      "grad_norm": 24.3690128326416,
      "learning_rate": 3.421828908554573e-05,
      "loss": 0.7673,
      "step": 438
    },
    {
      "epoch": 3.893805309734513,
      "grad_norm": 22.95048713684082,
      "learning_rate": 3.4119960668633235e-05,
      "loss": 0.7641,
      "step": 440
    },
    {
      "epoch": 3.911504424778761,
      "grad_norm": 21.6431827545166,
      "learning_rate": 3.402163225172075e-05,
      "loss": 0.832,
      "step": 442
    },
    {
      "epoch": 3.9292035398230087,
      "grad_norm": 30.03487777709961,
      "learning_rate": 3.3923303834808265e-05,
      "loss": 0.8857,
      "step": 444
    },
    {
      "epoch": 3.9469026548672566,
      "grad_norm": 32.90857696533203,
      "learning_rate": 3.3824975417895774e-05,
      "loss": 0.8063,
      "step": 446
    },
    {
      "epoch": 3.9646017699115044,
      "grad_norm": 18.506166458129883,
      "learning_rate": 3.372664700098329e-05,
      "loss": 0.8189,
      "step": 448
    },
    {
      "epoch": 3.982300884955752,
      "grad_norm": 21.59197235107422,
      "learning_rate": 3.3628318584070804e-05,
      "loss": 0.711,
      "step": 450
    },
    {
      "epoch": 4.0,
      "grad_norm": 55.65620422363281,
      "learning_rate": 3.352999016715831e-05,
      "loss": 0.8934,
      "step": 452
    },
    {
      "epoch": 4.0,
      "eval_loss": 0.8967828154563904,
      "eval_map": 0.4488,
      "eval_map_50": 0.8568,
      "eval_map_75": 0.4271,
      "eval_map_large": 0.5174,
      "eval_map_medium": 0.264,
      "eval_map_per_class": 0.4488,
      "eval_map_small": 0.0,
      "eval_mar_1": 0.2483,
      "eval_mar_10": 0.5628,
      "eval_mar_100": 0.6197,
      "eval_mar_100_per_class": 0.6197,
      "eval_mar_large": 0.6834,
      "eval_mar_medium": 0.4609,
      "eval_mar_small": 0.0,
      "eval_runtime": 3.9666,
      "eval_samples_per_second": 25.21,
      "eval_steps_per_second": 1.765,
      "step": 452
    },
    {
      "epoch": 4.017699115044247,
      "grad_norm": 27.480350494384766,
      "learning_rate": 3.343166175024583e-05,
      "loss": 0.8197,
      "step": 454
    },
    {
      "epoch": 4.035398230088496,
      "grad_norm": 17.332786560058594,
      "learning_rate": 3.3333333333333335e-05,
      "loss": 0.8171,
      "step": 456
    },
    {
      "epoch": 4.053097345132743,
      "grad_norm": 14.860098838806152,
      "learning_rate": 3.323500491642085e-05,
      "loss": 0.7993,
      "step": 458
    },
    {
      "epoch": 4.070796460176991,
      "grad_norm": 16.08732795715332,
      "learning_rate": 3.313667649950836e-05,
      "loss": 0.7328,
      "step": 460
    },
    {
      "epoch": 4.088495575221239,
      "grad_norm": 24.5137939453125,
      "learning_rate": 3.303834808259587e-05,
      "loss": 0.7629,
      "step": 462
    },
    {
      "epoch": 4.106194690265487,
      "grad_norm": 28.86885643005371,
      "learning_rate": 3.294001966568339e-05,
      "loss": 0.9142,
      "step": 464
    },
    {
      "epoch": 4.123893805309734,
      "grad_norm": 26.640892028808594,
      "learning_rate": 3.2841691248770896e-05,
      "loss": 0.8952,
      "step": 466
    },
    {
      "epoch": 4.1415929203539825,
      "grad_norm": 16.20265769958496,
      "learning_rate": 3.274336283185841e-05,
      "loss": 0.7812,
      "step": 468
    },
    {
      "epoch": 4.15929203539823,
      "grad_norm": 22.30803871154785,
      "learning_rate": 3.264503441494592e-05,
      "loss": 0.8641,
      "step": 470
    },
    {
      "epoch": 4.176991150442478,
      "grad_norm": 17.450302124023438,
      "learning_rate": 3.2546705998033434e-05,
      "loss": 0.8061,
      "step": 472
    },
    {
      "epoch": 4.1946902654867255,
      "grad_norm": 20.605792999267578,
      "learning_rate": 3.244837758112095e-05,
      "loss": 0.8872,
      "step": 474
    },
    {
      "epoch": 4.212389380530974,
      "grad_norm": 21.550548553466797,
      "learning_rate": 3.235004916420846e-05,
      "loss": 0.8599,
      "step": 476
    },
    {
      "epoch": 4.230088495575221,
      "grad_norm": 14.21965217590332,
      "learning_rate": 3.225172074729597e-05,
      "loss": 0.7412,
      "step": 478
    },
    {
      "epoch": 4.247787610619469,
      "grad_norm": 24.305540084838867,
      "learning_rate": 3.215339233038348e-05,
      "loss": 0.7362,
      "step": 480
    },
    {
      "epoch": 4.265486725663717,
      "grad_norm": 26.687883377075195,
      "learning_rate": 3.2055063913470996e-05,
      "loss": 0.8364,
      "step": 482
    },
    {
      "epoch": 4.283185840707965,
      "grad_norm": 23.61547088623047,
      "learning_rate": 3.1956735496558504e-05,
      "loss": 0.6807,
      "step": 484
    },
    {
      "epoch": 4.300884955752212,
      "grad_norm": 36.08306121826172,
      "learning_rate": 3.185840707964602e-05,
      "loss": 0.7484,
      "step": 486
    },
    {
      "epoch": 4.31858407079646,
      "grad_norm": 24.663066864013672,
      "learning_rate": 3.1760078662733534e-05,
      "loss": 0.7722,
      "step": 488
    },
    {
      "epoch": 4.336283185840708,
      "grad_norm": 25.614458084106445,
      "learning_rate": 3.166175024582104e-05,
      "loss": 0.9276,
      "step": 490
    },
    {
      "epoch": 4.353982300884955,
      "grad_norm": 22.014490127563477,
      "learning_rate": 3.156342182890856e-05,
      "loss": 0.8552,
      "step": 492
    },
    {
      "epoch": 4.371681415929204,
      "grad_norm": 16.620935440063477,
      "learning_rate": 3.1465093411996065e-05,
      "loss": 0.8883,
      "step": 494
    },
    {
      "epoch": 4.389380530973451,
      "grad_norm": 19.76045036315918,
      "learning_rate": 3.136676499508358e-05,
      "loss": 0.5994,
      "step": 496
    },
    {
      "epoch": 4.407079646017699,
      "grad_norm": 27.53520393371582,
      "learning_rate": 3.1268436578171095e-05,
      "loss": 0.7127,
      "step": 498
    },
    {
      "epoch": 4.424778761061947,
      "grad_norm": 17.578168869018555,
      "learning_rate": 3.1170108161258604e-05,
      "loss": 0.8606,
      "step": 500
    },
    {
      "epoch": 4.442477876106195,
      "grad_norm": 25.594207763671875,
      "learning_rate": 3.107177974434612e-05,
      "loss": 0.736,
      "step": 502
    },
    {
      "epoch": 4.460176991150442,
      "grad_norm": 17.3150691986084,
      "learning_rate": 3.097345132743363e-05,
      "loss": 0.8871,
      "step": 504
    },
    {
      "epoch": 4.477876106194691,
      "grad_norm": 34.62860107421875,
      "learning_rate": 3.087512291052114e-05,
      "loss": 0.8176,
      "step": 506
    },
    {
      "epoch": 4.495575221238938,
      "grad_norm": 21.978734970092773,
      "learning_rate": 3.077679449360865e-05,
      "loss": 0.8523,
      "step": 508
    },
    {
      "epoch": 4.513274336283186,
      "grad_norm": 37.72077941894531,
      "learning_rate": 3.0678466076696165e-05,
      "loss": 0.6939,
      "step": 510
    },
    {
      "epoch": 4.530973451327434,
      "grad_norm": 29.62040901184082,
      "learning_rate": 3.058013765978368e-05,
      "loss": 0.9062,
      "step": 512
    },
    {
      "epoch": 4.548672566371682,
      "grad_norm": 43.172821044921875,
      "learning_rate": 3.048180924287119e-05,
      "loss": 0.9389,
      "step": 514
    },
    {
      "epoch": 4.566371681415929,
      "grad_norm": 19.374061584472656,
      "learning_rate": 3.0383480825958703e-05,
      "loss": 0.8577,
      "step": 516
    },
    {
      "epoch": 4.584070796460177,
      "grad_norm": 18.854230880737305,
      "learning_rate": 3.0285152409046215e-05,
      "loss": 0.7861,
      "step": 518
    },
    {
      "epoch": 4.601769911504425,
      "grad_norm": 27.53508949279785,
      "learning_rate": 3.018682399213373e-05,
      "loss": 0.8935,
      "step": 520
    },
    {
      "epoch": 4.619469026548672,
      "grad_norm": 25.085142135620117,
      "learning_rate": 3.008849557522124e-05,
      "loss": 0.6619,
      "step": 522
    },
    {
      "epoch": 4.6371681415929205,
      "grad_norm": 22.648717880249023,
      "learning_rate": 2.9990167158308753e-05,
      "loss": 0.7065,
      "step": 524
    },
    {
      "epoch": 4.654867256637168,
      "grad_norm": 19.692138671875,
      "learning_rate": 2.9891838741396268e-05,
      "loss": 0.8524,
      "step": 526
    },
    {
      "epoch": 4.672566371681416,
      "grad_norm": 17.703353881835938,
      "learning_rate": 2.9793510324483776e-05,
      "loss": 0.8405,
      "step": 528
    },
    {
      "epoch": 4.6902654867256635,
      "grad_norm": 19.434688568115234,
      "learning_rate": 2.969518190757129e-05,
      "loss": 0.7932,
      "step": 530
    },
    {
      "epoch": 4.707964601769912,
      "grad_norm": 19.711090087890625,
      "learning_rate": 2.95968534906588e-05,
      "loss": 0.6708,
      "step": 532
    },
    {
      "epoch": 4.725663716814159,
      "grad_norm": 32.07950973510742,
      "learning_rate": 2.9498525073746314e-05,
      "loss": 0.8732,
      "step": 534
    },
    {
      "epoch": 4.743362831858407,
      "grad_norm": 26.682344436645508,
      "learning_rate": 2.940019665683383e-05,
      "loss": 0.8234,
      "step": 536
    },
    {
      "epoch": 4.761061946902655,
      "grad_norm": 20.226280212402344,
      "learning_rate": 2.9301868239921338e-05,
      "loss": 0.6903,
      "step": 538
    },
    {
      "epoch": 4.778761061946903,
      "grad_norm": 28.748722076416016,
      "learning_rate": 2.9203539823008852e-05,
      "loss": 0.8606,
      "step": 540
    },
    {
      "epoch": 4.79646017699115,
      "grad_norm": 21.78730583190918,
      "learning_rate": 2.910521140609636e-05,
      "loss": 0.7924,
      "step": 542
    },
    {
      "epoch": 4.814159292035399,
      "grad_norm": 24.920272827148438,
      "learning_rate": 2.9006882989183876e-05,
      "loss": 0.7917,
      "step": 544
    },
    {
      "epoch": 4.831858407079646,
      "grad_norm": 32.964298248291016,
      "learning_rate": 2.890855457227139e-05,
      "loss": 0.8017,
      "step": 546
    },
    {
      "epoch": 4.849557522123893,
      "grad_norm": 27.990001678466797,
      "learning_rate": 2.88102261553589e-05,
      "loss": 0.8392,
      "step": 548
    },
    {
      "epoch": 4.867256637168142,
      "grad_norm": 24.096538543701172,
      "learning_rate": 2.8711897738446414e-05,
      "loss": 0.9032,
      "step": 550
    },
    {
      "epoch": 4.88495575221239,
      "grad_norm": 22.516420364379883,
      "learning_rate": 2.8613569321533922e-05,
      "loss": 0.8411,
      "step": 552
    },
    {
      "epoch": 4.902654867256637,
      "grad_norm": 32.133506774902344,
      "learning_rate": 2.8515240904621437e-05,
      "loss": 0.824,
      "step": 554
    },
    {
      "epoch": 4.920353982300885,
      "grad_norm": 20.993709564208984,
      "learning_rate": 2.8416912487708945e-05,
      "loss": 0.6233,
      "step": 556
    },
    {
      "epoch": 4.938053097345133,
      "grad_norm": 17.405899047851562,
      "learning_rate": 2.831858407079646e-05,
      "loss": 0.8157,
      "step": 558
    },
    {
      "epoch": 4.95575221238938,
      "grad_norm": 18.684974670410156,
      "learning_rate": 2.8220255653883975e-05,
      "loss": 0.889,
      "step": 560
    },
    {
      "epoch": 4.9734513274336285,
      "grad_norm": 22.140762329101562,
      "learning_rate": 2.8121927236971483e-05,
      "loss": 0.7555,
      "step": 562
    },
    {
      "epoch": 4.991150442477876,
      "grad_norm": 93.78765869140625,
      "learning_rate": 2.8023598820059e-05,
      "loss": 0.8166,
      "step": 564
    },
    {
      "epoch": 5.0,
      "eval_loss": 0.8204176425933838,
      "eval_map": 0.4963,
      "eval_map_50": 0.8784,
      "eval_map_75": 0.5214,
      "eval_map_large": 0.56,
      "eval_map_medium": 0.3309,
      "eval_map_per_class": 0.4963,
      "eval_map_small": 0.0209,
      "eval_mar_1": 0.2598,
      "eval_mar_10": 0.5983,
      "eval_mar_100": 0.6581,
      "eval_mar_100_per_class": 0.6581,
      "eval_mar_large": 0.7207,
      "eval_mar_medium": 0.4969,
      "eval_mar_small": 0.4,
      "eval_runtime": 4.466,
      "eval_samples_per_second": 22.392,
      "eval_steps_per_second": 1.567,
      "step": 565
    },
    {
      "epoch": 5.008849557522124,
      "grad_norm": 17.966142654418945,
      "learning_rate": 2.792527040314651e-05,
      "loss": 0.7743,
      "step": 566
    },
    {
      "epoch": 5.0265486725663715,
      "grad_norm": 17.44831657409668,
      "learning_rate": 2.782694198623402e-05,
      "loss": 0.8002,
      "step": 568
    },
    {
      "epoch": 5.04424778761062,
      "grad_norm": 16.89792823791504,
      "learning_rate": 2.7728613569321537e-05,
      "loss": 0.8238,
      "step": 570
    },
    {
      "epoch": 5.061946902654867,
      "grad_norm": 15.499351501464844,
      "learning_rate": 2.7630285152409048e-05,
      "loss": 0.5594,
      "step": 572
    },
    {
      "epoch": 5.079646017699115,
      "grad_norm": 26.665563583374023,
      "learning_rate": 2.753195673549656e-05,
      "loss": 0.7994,
      "step": 574
    },
    {
      "epoch": 5.097345132743363,
      "grad_norm": 19.839536666870117,
      "learning_rate": 2.743362831858407e-05,
      "loss": 0.6578,
      "step": 576
    },
    {
      "epoch": 5.115044247787611,
      "grad_norm": 24.534902572631836,
      "learning_rate": 2.7335299901671586e-05,
      "loss": 0.7748,
      "step": 578
    },
    {
      "epoch": 5.132743362831858,
      "grad_norm": 23.61812400817871,
      "learning_rate": 2.7236971484759095e-05,
      "loss": 0.7771,
      "step": 580
    },
    {
      "epoch": 5.150442477876107,
      "grad_norm": 17.933223724365234,
      "learning_rate": 2.713864306784661e-05,
      "loss": 0.7349,
      "step": 582
    },
    {
      "epoch": 5.168141592920354,
      "grad_norm": 18.23980712890625,
      "learning_rate": 2.7040314650934125e-05,
      "loss": 0.7771,
      "step": 584
    },
    {
      "epoch": 5.185840707964601,
      "grad_norm": 20.185016632080078,
      "learning_rate": 2.6941986234021633e-05,
      "loss": 0.8009,
      "step": 586
    },
    {
      "epoch": 5.20353982300885,
      "grad_norm": 26.6129093170166,
      "learning_rate": 2.6843657817109148e-05,
      "loss": 0.8614,
      "step": 588
    },
    {
      "epoch": 5.221238938053097,
      "grad_norm": 27.28291893005371,
      "learning_rate": 2.6745329400196656e-05,
      "loss": 0.8594,
      "step": 590
    },
    {
      "epoch": 5.238938053097345,
      "grad_norm": 19.851408004760742,
      "learning_rate": 2.664700098328417e-05,
      "loss": 0.6753,
      "step": 592
    },
    {
      "epoch": 5.256637168141593,
      "grad_norm": 15.925873756408691,
      "learning_rate": 2.6548672566371686e-05,
      "loss": 0.7364,
      "step": 594
    },
    {
      "epoch": 5.274336283185841,
      "grad_norm": 16.969070434570312,
      "learning_rate": 2.6450344149459194e-05,
      "loss": 0.7607,
      "step": 596
    },
    {
      "epoch": 5.292035398230088,
      "grad_norm": 29.575332641601562,
      "learning_rate": 2.635201573254671e-05,
      "loss": 0.6162,
      "step": 598
    },
    {
      "epoch": 5.3097345132743365,
      "grad_norm": 27.425867080688477,
      "learning_rate": 2.6253687315634217e-05,
      "loss": 0.7873,
      "step": 600
    },
    {
      "epoch": 5.327433628318584,
      "grad_norm": 36.430233001708984,
      "learning_rate": 2.6155358898721732e-05,
      "loss": 0.7635,
      "step": 602
    },
    {
      "epoch": 5.345132743362832,
      "grad_norm": 23.08193588256836,
      "learning_rate": 2.605703048180924e-05,
      "loss": 0.8998,
      "step": 604
    },
    {
      "epoch": 5.3628318584070795,
      "grad_norm": 19.06671714782715,
      "learning_rate": 2.5958702064896756e-05,
      "loss": 0.6907,
      "step": 606
    },
    {
      "epoch": 5.380530973451328,
      "grad_norm": 17.793241500854492,
      "learning_rate": 2.586037364798427e-05,
      "loss": 0.7653,
      "step": 608
    },
    {
      "epoch": 5.398230088495575,
      "grad_norm": 18.47113800048828,
      "learning_rate": 2.576204523107178e-05,
      "loss": 0.7344,
      "step": 610
    },
    {
      "epoch": 5.415929203539823,
      "grad_norm": 23.754230499267578,
      "learning_rate": 2.5663716814159294e-05,
      "loss": 0.7098,
      "step": 612
    },
    {
      "epoch": 5.433628318584071,
      "grad_norm": 17.286996841430664,
      "learning_rate": 2.5565388397246802e-05,
      "loss": 0.7242,
      "step": 614
    },
    {
      "epoch": 5.451327433628318,
      "grad_norm": 33.41328811645508,
      "learning_rate": 2.5467059980334317e-05,
      "loss": 0.8577,
      "step": 616
    },
    {
      "epoch": 5.469026548672566,
      "grad_norm": 23.295421600341797,
      "learning_rate": 2.5368731563421832e-05,
      "loss": 0.765,
      "step": 618
    },
    {
      "epoch": 5.486725663716814,
      "grad_norm": 29.50783348083496,
      "learning_rate": 2.527040314650934e-05,
      "loss": 0.7021,
      "step": 620
    },
    {
      "epoch": 5.504424778761062,
      "grad_norm": 20.60035514831543,
      "learning_rate": 2.5172074729596855e-05,
      "loss": 0.8331,
      "step": 622
    },
    {
      "epoch": 5.522123893805309,
      "grad_norm": 28.808502197265625,
      "learning_rate": 2.5073746312684367e-05,
      "loss": 0.9017,
      "step": 624
    },
    {
      "epoch": 5.539823008849558,
      "grad_norm": 63.102195739746094,
      "learning_rate": 2.4975417895771878e-05,
      "loss": 0.7724,
      "step": 626
    },
    {
      "epoch": 5.557522123893805,
      "grad_norm": 31.407791137695312,
      "learning_rate": 2.4877089478859393e-05,
      "loss": 0.775,
      "step": 628
    },
    {
      "epoch": 5.575221238938053,
      "grad_norm": 27.927444458007812,
      "learning_rate": 2.4778761061946905e-05,
      "loss": 0.6807,
      "step": 630
    },
    {
      "epoch": 5.592920353982301,
      "grad_norm": 31.49755096435547,
      "learning_rate": 2.4680432645034416e-05,
      "loss": 0.7617,
      "step": 632
    },
    {
      "epoch": 5.610619469026549,
      "grad_norm": 24.49897003173828,
      "learning_rate": 2.4582104228121928e-05,
      "loss": 0.7275,
      "step": 634
    },
    {
      "epoch": 5.628318584070796,
      "grad_norm": 29.279388427734375,
      "learning_rate": 2.4483775811209443e-05,
      "loss": 0.6349,
      "step": 636
    },
    {
      "epoch": 5.646017699115045,
      "grad_norm": 18.994455337524414,
      "learning_rate": 2.4385447394296955e-05,
      "loss": 0.6773,
      "step": 638
    },
    {
      "epoch": 5.663716814159292,
      "grad_norm": 21.28229522705078,
      "learning_rate": 2.4287118977384466e-05,
      "loss": 0.8142,
      "step": 640
    },
    {
      "epoch": 5.68141592920354,
      "grad_norm": 28.956829071044922,
      "learning_rate": 2.4188790560471978e-05,
      "loss": 0.7215,
      "step": 642
    },
    {
      "epoch": 5.699115044247788,
      "grad_norm": 35.786991119384766,
      "learning_rate": 2.409046214355949e-05,
      "loss": 0.8148,
      "step": 644
    },
    {
      "epoch": 5.716814159292035,
      "grad_norm": 17.956323623657227,
      "learning_rate": 2.3992133726647e-05,
      "loss": 0.9103,
      "step": 646
    },
    {
      "epoch": 5.734513274336283,
      "grad_norm": 21.570697784423828,
      "learning_rate": 2.3893805309734516e-05,
      "loss": 0.6904,
      "step": 648
    },
    {
      "epoch": 5.752212389380531,
      "grad_norm": 13.703173637390137,
      "learning_rate": 2.3795476892822028e-05,
      "loss": 0.6716,
      "step": 650
    },
    {
      "epoch": 5.769911504424779,
      "grad_norm": 19.16945457458496,
      "learning_rate": 2.369714847590954e-05,
      "loss": 0.8447,
      "step": 652
    },
    {
      "epoch": 5.787610619469026,
      "grad_norm": 15.37072467803955,
      "learning_rate": 2.359882005899705e-05,
      "loss": 0.7729,
      "step": 654
    },
    {
      "epoch": 5.8053097345132745,
      "grad_norm": 16.8475341796875,
      "learning_rate": 2.3500491642084562e-05,
      "loss": 0.8331,
      "step": 656
    },
    {
      "epoch": 5.823008849557522,
      "grad_norm": 23.530323028564453,
      "learning_rate": 2.3402163225172074e-05,
      "loss": 0.7384,
      "step": 658
    },
    {
      "epoch": 5.84070796460177,
      "grad_norm": 29.106124877929688,
      "learning_rate": 2.330383480825959e-05,
      "loss": 0.7302,
      "step": 660
    },
    {
      "epoch": 5.8584070796460175,
      "grad_norm": 13.785088539123535,
      "learning_rate": 2.32055063913471e-05,
      "loss": 0.6624,
      "step": 662
    },
    {
      "epoch": 5.876106194690266,
      "grad_norm": 24.307811737060547,
      "learning_rate": 2.3107177974434612e-05,
      "loss": 0.8129,
      "step": 664
    },
    {
      "epoch": 5.893805309734513,
      "grad_norm": 32.765541076660156,
      "learning_rate": 2.3008849557522124e-05,
      "loss": 0.7463,
      "step": 666
    },
    {
      "epoch": 5.911504424778761,
      "grad_norm": 15.36600112915039,
      "learning_rate": 2.2910521140609635e-05,
      "loss": 0.8945,
      "step": 668
    },
    {
      "epoch": 5.929203539823009,
      "grad_norm": 21.934234619140625,
      "learning_rate": 2.2812192723697147e-05,
      "loss": 0.7991,
      "step": 670
    },
    {
      "epoch": 5.946902654867257,
      "grad_norm": 18.248523712158203,
      "learning_rate": 2.2713864306784662e-05,
      "loss": 0.7406,
      "step": 672
    },
    {
      "epoch": 5.964601769911504,
      "grad_norm": 18.006677627563477,
      "learning_rate": 2.2615535889872174e-05,
      "loss": 0.7947,
      "step": 674
    },
    {
      "epoch": 5.982300884955752,
      "grad_norm": 34.55229568481445,
      "learning_rate": 2.2517207472959685e-05,
      "loss": 0.7464,
      "step": 676
    },
    {
      "epoch": 6.0,
      "grad_norm": 36.384239196777344,
      "learning_rate": 2.24188790560472e-05,
      "loss": 0.6889,
      "step": 678
    },
    {
      "epoch": 6.0,
      "eval_loss": 0.7955648303031921,
      "eval_map": 0.5081,
      "eval_map_50": 0.8984,
      "eval_map_75": 0.5276,
      "eval_map_large": 0.5756,
      "eval_map_medium": 0.3127,
      "eval_map_per_class": 0.5081,
      "eval_map_small": 0.0226,
      "eval_mar_1": 0.2658,
      "eval_mar_10": 0.5991,
      "eval_mar_100": 0.6538,
      "eval_mar_100_per_class": 0.6538,
      "eval_mar_large": 0.713,
      "eval_mar_medium": 0.5016,
      "eval_mar_small": 0.4,
      "eval_runtime": 4.0138,
      "eval_samples_per_second": 24.914,
      "eval_steps_per_second": 1.744,
      "step": 678
    },
    {
      "epoch": 6.017699115044247,
      "grad_norm": 30.158843994140625,
      "learning_rate": 2.232055063913471e-05,
      "loss": 0.7447,
      "step": 680
    },
    {
      "epoch": 6.035398230088496,
      "grad_norm": 21.382171630859375,
      "learning_rate": 2.2222222222222223e-05,
      "loss": 0.7444,
      "step": 682
    },
    {
      "epoch": 6.053097345132743,
      "grad_norm": 16.47271156311035,
      "learning_rate": 2.2123893805309738e-05,
      "loss": 0.7564,
      "step": 684
    },
    {
      "epoch": 6.070796460176991,
      "grad_norm": 26.326805114746094,
      "learning_rate": 2.202556538839725e-05,
      "loss": 0.7716,
      "step": 686
    },
    {
      "epoch": 6.088495575221239,
      "grad_norm": 21.000934600830078,
      "learning_rate": 2.192723697148476e-05,
      "loss": 0.9769,
      "step": 688
    },
    {
      "epoch": 6.106194690265487,
      "grad_norm": 14.34850025177002,
      "learning_rate": 2.1828908554572273e-05,
      "loss": 0.6075,
      "step": 690
    },
    {
      "epoch": 6.123893805309734,
      "grad_norm": 14.81210994720459,
      "learning_rate": 2.1730580137659785e-05,
      "loss": 0.6219,
      "step": 692
    },
    {
      "epoch": 6.1415929203539825,
      "grad_norm": 15.356616973876953,
      "learning_rate": 2.1632251720747296e-05,
      "loss": 0.6604,
      "step": 694
    },
    {
      "epoch": 6.15929203539823,
      "grad_norm": 25.215105056762695,
      "learning_rate": 2.153392330383481e-05,
      "loss": 0.8016,
      "step": 696
    },
    {
      "epoch": 6.176991150442478,
      "grad_norm": 29.563343048095703,
      "learning_rate": 2.1435594886922323e-05,
      "loss": 0.6862,
      "step": 698
    },
    {
      "epoch": 6.1946902654867255,
      "grad_norm": 21.926513671875,
      "learning_rate": 2.1337266470009834e-05,
      "loss": 0.7859,
      "step": 700
    },
    {
      "epoch": 6.212389380530974,
      "grad_norm": 23.07330322265625,
      "learning_rate": 2.1238938053097346e-05,
      "loss": 0.6345,
      "step": 702
    },
    {
      "epoch": 6.230088495575221,
      "grad_norm": 17.459535598754883,
      "learning_rate": 2.1140609636184858e-05,
      "loss": 0.7799,
      "step": 704
    },
    {
      "epoch": 6.247787610619469,
      "grad_norm": 16.414749145507812,
      "learning_rate": 2.104228121927237e-05,
      "loss": 0.8234,
      "step": 706
    },
    {
      "epoch": 6.265486725663717,
      "grad_norm": 30.402633666992188,
      "learning_rate": 2.0943952802359884e-05,
      "loss": 0.8438,
      "step": 708
    },
    {
      "epoch": 6.283185840707965,
      "grad_norm": 36.2429084777832,
      "learning_rate": 2.0845624385447396e-05,
      "loss": 0.8062,
      "step": 710
    },
    {
      "epoch": 6.300884955752212,
      "grad_norm": 29.136394500732422,
      "learning_rate": 2.0747295968534907e-05,
      "loss": 0.7717,
      "step": 712
    },
    {
      "epoch": 6.31858407079646,
      "grad_norm": 22.905851364135742,
      "learning_rate": 2.064896755162242e-05,
      "loss": 0.7359,
      "step": 714
    },
    {
      "epoch": 6.336283185840708,
      "grad_norm": 20.239849090576172,
      "learning_rate": 2.055063913470993e-05,
      "loss": 0.6981,
      "step": 716
    },
    {
      "epoch": 6.353982300884955,
      "grad_norm": 24.594646453857422,
      "learning_rate": 2.0452310717797442e-05,
      "loss": 0.6381,
      "step": 718
    },
    {
      "epoch": 6.371681415929204,
      "grad_norm": 26.90498161315918,
      "learning_rate": 2.0353982300884957e-05,
      "loss": 0.8903,
      "step": 720
    },
    {
      "epoch": 6.389380530973451,
      "grad_norm": 29.696321487426758,
      "learning_rate": 2.025565388397247e-05,
      "loss": 0.6345,
      "step": 722
    },
    {
      "epoch": 6.407079646017699,
      "grad_norm": 30.505664825439453,
      "learning_rate": 2.015732546705998e-05,
      "loss": 0.7814,
      "step": 724
    },
    {
      "epoch": 6.424778761061947,
      "grad_norm": 36.33932876586914,
      "learning_rate": 2.0058997050147492e-05,
      "loss": 0.7754,
      "step": 726
    },
    {
      "epoch": 6.442477876106195,
      "grad_norm": 18.5018310546875,
      "learning_rate": 1.9960668633235004e-05,
      "loss": 0.7061,
      "step": 728
    },
    {
      "epoch": 6.460176991150442,
      "grad_norm": 16.27545166015625,
      "learning_rate": 1.986234021632252e-05,
      "loss": 0.7759,
      "step": 730
    },
    {
      "epoch": 6.477876106194691,
      "grad_norm": 19.90315055847168,
      "learning_rate": 1.976401179941003e-05,
      "loss": 0.7227,
      "step": 732
    },
    {
      "epoch": 6.495575221238938,
      "grad_norm": 23.48577880859375,
      "learning_rate": 1.9665683382497542e-05,
      "loss": 0.7967,
      "step": 734
    },
    {
      "epoch": 6.513274336283186,
      "grad_norm": 25.51618003845215,
      "learning_rate": 1.9567354965585057e-05,
      "loss": 0.7768,
      "step": 736
    },
    {
      "epoch": 6.530973451327434,
      "grad_norm": 17.899826049804688,
      "learning_rate": 1.946902654867257e-05,
      "loss": 0.6657,
      "step": 738
    },
    {
      "epoch": 6.548672566371682,
      "grad_norm": 19.610185623168945,
      "learning_rate": 1.937069813176008e-05,
      "loss": 0.8086,
      "step": 740
    },
    {
      "epoch": 6.566371681415929,
      "grad_norm": 25.385841369628906,
      "learning_rate": 1.927236971484759e-05,
      "loss": 0.6468,
      "step": 742
    },
    {
      "epoch": 6.584070796460177,
      "grad_norm": 15.946438789367676,
      "learning_rate": 1.9174041297935107e-05,
      "loss": 0.676,
      "step": 744
    },
    {
      "epoch": 6.601769911504425,
      "grad_norm": 17.454030990600586,
      "learning_rate": 1.9075712881022618e-05,
      "loss": 0.7586,
      "step": 746
    },
    {
      "epoch": 6.619469026548672,
      "grad_norm": 24.09688949584961,
      "learning_rate": 1.897738446411013e-05,
      "loss": 0.5866,
      "step": 748
    },
    {
      "epoch": 6.6371681415929205,
      "grad_norm": 20.896167755126953,
      "learning_rate": 1.887905604719764e-05,
      "loss": 0.6377,
      "step": 750
    },
    {
      "epoch": 6.654867256637168,
      "grad_norm": 13.922218322753906,
      "learning_rate": 1.8780727630285153e-05,
      "loss": 0.6401,
      "step": 752
    },
    {
      "epoch": 6.672566371681416,
      "grad_norm": 22.96921157836914,
      "learning_rate": 1.8682399213372664e-05,
      "loss": 0.7318,
      "step": 754
    },
    {
      "epoch": 6.6902654867256635,
      "grad_norm": 24.575101852416992,
      "learning_rate": 1.858407079646018e-05,
      "loss": 0.7185,
      "step": 756
    },
    {
      "epoch": 6.707964601769912,
      "grad_norm": 19.46567726135254,
      "learning_rate": 1.848574237954769e-05,
      "loss": 0.7333,
      "step": 758
    },
    {
      "epoch": 6.725663716814159,
      "grad_norm": 16.00038719177246,
      "learning_rate": 1.8387413962635203e-05,
      "loss": 0.6284,
      "step": 760
    },
    {
      "epoch": 6.743362831858407,
      "grad_norm": 19.083911895751953,
      "learning_rate": 1.8289085545722714e-05,
      "loss": 0.7064,
      "step": 762
    },
    {
      "epoch": 6.761061946902655,
      "grad_norm": 21.226545333862305,
      "learning_rate": 1.8190757128810226e-05,
      "loss": 0.8701,
      "step": 764
    },
    {
      "epoch": 6.778761061946903,
      "grad_norm": 20.60919189453125,
      "learning_rate": 1.8092428711897737e-05,
      "loss": 0.8677,
      "step": 766
    },
    {
      "epoch": 6.79646017699115,
      "grad_norm": 26.225242614746094,
      "learning_rate": 1.7994100294985252e-05,
      "loss": 0.7429,
      "step": 768
    },
    {
      "epoch": 6.814159292035399,
      "grad_norm": 23.108654022216797,
      "learning_rate": 1.7895771878072764e-05,
      "loss": 0.6417,
      "step": 770
    },
    {
      "epoch": 6.831858407079646,
      "grad_norm": 21.556283950805664,
      "learning_rate": 1.7797443461160276e-05,
      "loss": 0.6965,
      "step": 772
    },
    {
      "epoch": 6.849557522123893,
      "grad_norm": 14.243925094604492,
      "learning_rate": 1.7699115044247787e-05,
      "loss": 0.6269,
      "step": 774
    },
    {
      "epoch": 6.867256637168142,
      "grad_norm": 25.8636417388916,
      "learning_rate": 1.76007866273353e-05,
      "loss": 0.707,
      "step": 776
    },
    {
      "epoch": 6.88495575221239,
      "grad_norm": 15.025083541870117,
      "learning_rate": 1.750245821042281e-05,
      "loss": 0.5885,
      "step": 778
    },
    {
      "epoch": 6.902654867256637,
      "grad_norm": 12.319948196411133,
      "learning_rate": 1.7404129793510325e-05,
      "loss": 0.6618,
      "step": 780
    },
    {
      "epoch": 6.920353982300885,
      "grad_norm": 18.02228355407715,
      "learning_rate": 1.7305801376597837e-05,
      "loss": 0.7721,
      "step": 782
    },
    {
      "epoch": 6.938053097345133,
      "grad_norm": 13.130449295043945,
      "learning_rate": 1.720747295968535e-05,
      "loss": 0.7447,
      "step": 784
    },
    {
      "epoch": 6.95575221238938,
      "grad_norm": 30.51601219177246,
      "learning_rate": 1.7109144542772864e-05,
      "loss": 0.7415,
      "step": 786
    },
    {
      "epoch": 6.9734513274336285,
      "grad_norm": 26.357257843017578,
      "learning_rate": 1.7010816125860375e-05,
      "loss": 0.6869,
      "step": 788
    },
    {
      "epoch": 6.991150442477876,
      "grad_norm": 23.887100219726562,
      "learning_rate": 1.6912487708947887e-05,
      "loss": 0.673,
      "step": 790
    },
    {
      "epoch": 7.0,
      "eval_loss": 0.7656337022781372,
      "eval_map": 0.5423,
      "eval_map_50": 0.9234,
      "eval_map_75": 0.5491,
      "eval_map_large": 0.6107,
      "eval_map_medium": 0.361,
      "eval_map_per_class": 0.5423,
      "eval_map_small": 0.025,
      "eval_mar_1": 0.2774,
      "eval_mar_10": 0.6252,
      "eval_mar_100": 0.6808,
      "eval_mar_100_per_class": 0.6808,
      "eval_mar_large": 0.7473,
      "eval_mar_medium": 0.5094,
      "eval_mar_small": 0.4,
      "eval_runtime": 3.6757,
      "eval_samples_per_second": 27.206,
      "eval_steps_per_second": 1.904,
      "step": 791
    },
    {
      "epoch": 7.008849557522124,
      "grad_norm": 16.251768112182617,
      "learning_rate": 1.6814159292035402e-05,
      "loss": 0.6063,
      "step": 792
    },
    {
      "epoch": 7.0265486725663715,
      "grad_norm": 15.13443660736084,
      "learning_rate": 1.6715830875122913e-05,
      "loss": 0.7639,
      "step": 794
    },
    {
      "epoch": 7.04424778761062,
      "grad_norm": 18.05428123474121,
      "learning_rate": 1.6617502458210425e-05,
      "loss": 0.6947,
      "step": 796
    },
    {
      "epoch": 7.061946902654867,
      "grad_norm": 18.904369354248047,
      "learning_rate": 1.6519174041297937e-05,
      "loss": 0.6764,
      "step": 798
    },
    {
      "epoch": 7.079646017699115,
      "grad_norm": 25.891990661621094,
      "learning_rate": 1.6420845624385448e-05,
      "loss": 0.6985,
      "step": 800
    },
    {
      "epoch": 7.097345132743363,
      "grad_norm": 21.057519912719727,
      "learning_rate": 1.632251720747296e-05,
      "loss": 0.7379,
      "step": 802
    },
    {
      "epoch": 7.115044247787611,
      "grad_norm": 19.18006706237793,
      "learning_rate": 1.6224188790560475e-05,
      "loss": 0.6897,
      "step": 804
    },
    {
      "epoch": 7.132743362831858,
      "grad_norm": 18.494470596313477,
      "learning_rate": 1.6125860373647986e-05,
      "loss": 0.6482,
      "step": 806
    },
    {
      "epoch": 7.150442477876107,
      "grad_norm": 30.54767608642578,
      "learning_rate": 1.6027531956735498e-05,
      "loss": 0.7792,
      "step": 808
    },
    {
      "epoch": 7.168141592920354,
      "grad_norm": 26.467973709106445,
      "learning_rate": 1.592920353982301e-05,
      "loss": 0.6576,
      "step": 810
    },
    {
      "epoch": 7.185840707964601,
      "grad_norm": 24.232133865356445,
      "learning_rate": 1.583087512291052e-05,
      "loss": 0.7487,
      "step": 812
    },
    {
      "epoch": 7.20353982300885,
      "grad_norm": 22.034992218017578,
      "learning_rate": 1.5732546705998033e-05,
      "loss": 0.7208,
      "step": 814
    },
    {
      "epoch": 7.221238938053097,
      "grad_norm": 17.85079574584961,
      "learning_rate": 1.5634218289085548e-05,
      "loss": 0.7738,
      "step": 816
    },
    {
      "epoch": 7.238938053097345,
      "grad_norm": 27.755950927734375,
      "learning_rate": 1.553588987217306e-05,
      "loss": 0.6782,
      "step": 818
    },
    {
      "epoch": 7.256637168141593,
      "grad_norm": 28.554393768310547,
      "learning_rate": 1.543756145526057e-05,
      "loss": 0.8659,
      "step": 820
    },
    {
      "epoch": 7.274336283185841,
      "grad_norm": 17.34409523010254,
      "learning_rate": 1.5339233038348082e-05,
      "loss": 0.6944,
      "step": 822
    },
    {
      "epoch": 7.292035398230088,
      "grad_norm": 17.795570373535156,
      "learning_rate": 1.5240904621435596e-05,
      "loss": 0.804,
      "step": 824
    },
    {
      "epoch": 7.3097345132743365,
      "grad_norm": 23.85445213317871,
      "learning_rate": 1.5142576204523107e-05,
      "loss": 0.7226,
      "step": 826
    },
    {
      "epoch": 7.327433628318584,
      "grad_norm": 22.983423233032227,
      "learning_rate": 1.504424778761062e-05,
      "loss": 0.7246,
      "step": 828
    },
    {
      "epoch": 7.345132743362832,
      "grad_norm": 22.66045570373535,
      "learning_rate": 1.4945919370698134e-05,
      "loss": 0.596,
      "step": 830
    },
    {
      "epoch": 7.3628318584070795,
      "grad_norm": 21.034364700317383,
      "learning_rate": 1.4847590953785646e-05,
      "loss": 0.709,
      "step": 832
    },
    {
      "epoch": 7.380530973451328,
      "grad_norm": 22.207504272460938,
      "learning_rate": 1.4749262536873157e-05,
      "loss": 0.6881,
      "step": 834
    },
    {
      "epoch": 7.398230088495575,
      "grad_norm": 32.008148193359375,
      "learning_rate": 1.4650934119960669e-05,
      "loss": 0.5796,
      "step": 836
    },
    {
      "epoch": 7.415929203539823,
      "grad_norm": 20.24285125732422,
      "learning_rate": 1.455260570304818e-05,
      "loss": 0.6238,
      "step": 838
    },
    {
      "epoch": 7.433628318584071,
      "grad_norm": 21.326202392578125,
      "learning_rate": 1.4454277286135695e-05,
      "loss": 0.7368,
      "step": 840
    },
    {
      "epoch": 7.451327433628318,
      "grad_norm": 30.127544403076172,
      "learning_rate": 1.4355948869223207e-05,
      "loss": 0.7563,
      "step": 842
    },
    {
      "epoch": 7.469026548672566,
      "grad_norm": 21.667497634887695,
      "learning_rate": 1.4257620452310719e-05,
      "loss": 0.7823,
      "step": 844
    },
    {
      "epoch": 7.486725663716814,
      "grad_norm": 25.011871337890625,
      "learning_rate": 1.415929203539823e-05,
      "loss": 0.7403,
      "step": 846
    },
    {
      "epoch": 7.504424778761062,
      "grad_norm": 22.891185760498047,
      "learning_rate": 1.4060963618485742e-05,
      "loss": 0.765,
      "step": 848
    },
    {
      "epoch": 7.522123893805309,
      "grad_norm": 20.289926528930664,
      "learning_rate": 1.3962635201573255e-05,
      "loss": 0.6539,
      "step": 850
    },
    {
      "epoch": 7.539823008849558,
      "grad_norm": 34.45566177368164,
      "learning_rate": 1.3864306784660768e-05,
      "loss": 0.6781,
      "step": 852
    },
    {
      "epoch": 7.557522123893805,
      "grad_norm": 16.83544921875,
      "learning_rate": 1.376597836774828e-05,
      "loss": 0.6451,
      "step": 854
    },
    {
      "epoch": 7.575221238938053,
      "grad_norm": 23.303050994873047,
      "learning_rate": 1.3667649950835793e-05,
      "loss": 0.6929,
      "step": 856
    },
    {
      "epoch": 7.592920353982301,
      "grad_norm": 22.088911056518555,
      "learning_rate": 1.3569321533923305e-05,
      "loss": 0.5961,
      "step": 858
    },
    {
      "epoch": 7.610619469026549,
      "grad_norm": 14.59677791595459,
      "learning_rate": 1.3470993117010816e-05,
      "loss": 0.6993,
      "step": 860
    },
    {
      "epoch": 7.628318584070796,
      "grad_norm": 23.7571964263916,
      "learning_rate": 1.3372664700098328e-05,
      "loss": 0.7123,
      "step": 862
    },
    {
      "epoch": 7.646017699115045,
      "grad_norm": 22.422454833984375,
      "learning_rate": 1.3274336283185843e-05,
      "loss": 0.8684,
      "step": 864
    },
    {
      "epoch": 7.663716814159292,
      "grad_norm": 18.472415924072266,
      "learning_rate": 1.3176007866273355e-05,
      "loss": 0.7778,
      "step": 866
    },
    {
      "epoch": 7.68141592920354,
      "grad_norm": 18.021146774291992,
      "learning_rate": 1.3077679449360866e-05,
      "loss": 0.8397,
      "step": 868
    },
    {
      "epoch": 7.699115044247788,
      "grad_norm": 19.535381317138672,
      "learning_rate": 1.2979351032448378e-05,
      "loss": 0.7257,
      "step": 870
    },
    {
      "epoch": 7.716814159292035,
      "grad_norm": 20.843671798706055,
      "learning_rate": 1.288102261553589e-05,
      "loss": 0.7802,
      "step": 872
    },
    {
      "epoch": 7.734513274336283,
      "grad_norm": 14.444415092468262,
      "learning_rate": 1.2782694198623401e-05,
      "loss": 0.6855,
      "step": 874
    },
    {
      "epoch": 7.752212389380531,
      "grad_norm": 15.192424774169922,
      "learning_rate": 1.2684365781710916e-05,
      "loss": 0.7157,
      "step": 876
    },
    {
      "epoch": 7.769911504424779,
      "grad_norm": 14.531706809997559,
      "learning_rate": 1.2586037364798428e-05,
      "loss": 0.7733,
      "step": 878
    },
    {
      "epoch": 7.787610619469026,
      "grad_norm": 22.154911041259766,
      "learning_rate": 1.2487708947885939e-05,
      "loss": 0.6404,
      "step": 880
    },
    {
      "epoch": 7.8053097345132745,
      "grad_norm": 18.85184669494629,
      "learning_rate": 1.2389380530973452e-05,
      "loss": 0.5891,
      "step": 882
    },
    {
      "epoch": 7.823008849557522,
      "grad_norm": 18.25426483154297,
      "learning_rate": 1.2291052114060964e-05,
      "loss": 0.8318,
      "step": 884
    },
    {
      "epoch": 7.84070796460177,
      "grad_norm": 15.124105453491211,
      "learning_rate": 1.2192723697148477e-05,
      "loss": 0.728,
      "step": 886
    },
    {
      "epoch": 7.8584070796460175,
      "grad_norm": 18.860633850097656,
      "learning_rate": 1.2094395280235989e-05,
      "loss": 0.8461,
      "step": 888
    },
    {
      "epoch": 7.876106194690266,
      "grad_norm": 21.146745681762695,
      "learning_rate": 1.19960668633235e-05,
      "loss": 0.7089,
      "step": 890
    },
    {
      "epoch": 7.893805309734513,
      "grad_norm": 19.12373924255371,
      "learning_rate": 1.1897738446411014e-05,
      "loss": 0.668,
      "step": 892
    },
    {
      "epoch": 7.911504424778761,
      "grad_norm": 12.939081192016602,
      "learning_rate": 1.1799410029498525e-05,
      "loss": 0.8069,
      "step": 894
    },
    {
      "epoch": 7.929203539823009,
      "grad_norm": 14.852071762084961,
      "learning_rate": 1.1701081612586037e-05,
      "loss": 0.6977,
      "step": 896
    },
    {
      "epoch": 7.946902654867257,
      "grad_norm": 16.5046329498291,
      "learning_rate": 1.160275319567355e-05,
      "loss": 0.6836,
      "step": 898
    },
    {
      "epoch": 7.964601769911504,
      "grad_norm": 27.28510284423828,
      "learning_rate": 1.1504424778761062e-05,
      "loss": 0.7441,
      "step": 900
    },
    {
      "epoch": 7.982300884955752,
      "grad_norm": 20.414901733398438,
      "learning_rate": 1.1406096361848573e-05,
      "loss": 0.7189,
      "step": 902
    },
    {
      "epoch": 8.0,
      "grad_norm": 15.717076301574707,
      "learning_rate": 1.1307767944936087e-05,
      "loss": 0.7197,
      "step": 904
    },
    {
      "epoch": 8.0,
      "eval_loss": 0.7693684101104736,
      "eval_map": 0.5471,
      "eval_map_50": 0.9308,
      "eval_map_75": 0.5913,
      "eval_map_large": 0.6096,
      "eval_map_medium": 0.3715,
      "eval_map_per_class": 0.5471,
      "eval_map_small": 0.0188,
      "eval_mar_1": 0.2833,
      "eval_mar_10": 0.6235,
      "eval_mar_100": 0.6761,
      "eval_mar_100_per_class": 0.6761,
      "eval_mar_large": 0.7278,
      "eval_mar_medium": 0.5453,
      "eval_mar_small": 0.3,
      "eval_runtime": 3.669,
      "eval_samples_per_second": 27.255,
      "eval_steps_per_second": 1.908,
      "step": 904
    },
    {
      "epoch": 8.017699115044248,
      "grad_norm": 19.057044982910156,
      "learning_rate": 1.12094395280236e-05,
      "loss": 0.6378,
      "step": 906
    },
    {
      "epoch": 8.035398230088495,
      "grad_norm": 26.174707412719727,
      "learning_rate": 1.1111111111111112e-05,
      "loss": 0.6774,
      "step": 908
    },
    {
      "epoch": 8.053097345132743,
      "grad_norm": 23.515504837036133,
      "learning_rate": 1.1012782694198625e-05,
      "loss": 0.7632,
      "step": 910
    },
    {
      "epoch": 8.070796460176991,
      "grad_norm": 32.522216796875,
      "learning_rate": 1.0914454277286137e-05,
      "loss": 0.7866,
      "step": 912
    },
    {
      "epoch": 8.08849557522124,
      "grad_norm": 17.90611457824707,
      "learning_rate": 1.0816125860373648e-05,
      "loss": 0.6607,
      "step": 914
    },
    {
      "epoch": 8.106194690265486,
      "grad_norm": 17.308259963989258,
      "learning_rate": 1.0717797443461161e-05,
      "loss": 0.9028,
      "step": 916
    },
    {
      "epoch": 8.123893805309734,
      "grad_norm": 23.00870704650879,
      "learning_rate": 1.0619469026548673e-05,
      "loss": 0.6444,
      "step": 918
    },
    {
      "epoch": 8.141592920353983,
      "grad_norm": 23.513565063476562,
      "learning_rate": 1.0521140609636185e-05,
      "loss": 0.7422,
      "step": 920
    },
    {
      "epoch": 8.15929203539823,
      "grad_norm": 22.787883758544922,
      "learning_rate": 1.0422812192723698e-05,
      "loss": 0.8156,
      "step": 922
    },
    {
      "epoch": 8.176991150442477,
      "grad_norm": 26.117023468017578,
      "learning_rate": 1.032448377581121e-05,
      "loss": 0.538,
      "step": 924
    },
    {
      "epoch": 8.194690265486726,
      "grad_norm": 21.87485122680664,
      "learning_rate": 1.0226155358898721e-05,
      "loss": 0.6871,
      "step": 926
    },
    {
      "epoch": 8.212389380530974,
      "grad_norm": 21.752988815307617,
      "learning_rate": 1.0127826941986234e-05,
      "loss": 0.7181,
      "step": 928
    },
    {
      "epoch": 8.230088495575222,
      "grad_norm": 24.260828018188477,
      "learning_rate": 1.0029498525073746e-05,
      "loss": 0.7211,
      "step": 930
    },
    {
      "epoch": 8.247787610619469,
      "grad_norm": 12.92242431640625,
      "learning_rate": 9.93117010816126e-06,
      "loss": 0.8347,
      "step": 932
    },
    {
      "epoch": 8.265486725663717,
      "grad_norm": 16.984403610229492,
      "learning_rate": 9.832841691248771e-06,
      "loss": 0.7955,
      "step": 934
    },
    {
      "epoch": 8.283185840707965,
      "grad_norm": 36.490318298339844,
      "learning_rate": 9.734513274336284e-06,
      "loss": 0.7464,
      "step": 936
    },
    {
      "epoch": 8.300884955752213,
      "grad_norm": 16.81266975402832,
      "learning_rate": 9.636184857423796e-06,
      "loss": 0.5796,
      "step": 938
    },
    {
      "epoch": 8.31858407079646,
      "grad_norm": 14.971821784973145,
      "learning_rate": 9.537856440511309e-06,
      "loss": 0.6221,
      "step": 940
    },
    {
      "epoch": 8.336283185840708,
      "grad_norm": 21.590499877929688,
      "learning_rate": 9.43952802359882e-06,
      "loss": 0.6271,
      "step": 942
    },
    {
      "epoch": 8.353982300884956,
      "grad_norm": 31.352338790893555,
      "learning_rate": 9.341199606686332e-06,
      "loss": 0.7592,
      "step": 944
    },
    {
      "epoch": 8.371681415929203,
      "grad_norm": 23.651779174804688,
      "learning_rate": 9.242871189773846e-06,
      "loss": 0.6352,
      "step": 946
    },
    {
      "epoch": 8.389380530973451,
      "grad_norm": 20.58877182006836,
      "learning_rate": 9.144542772861357e-06,
      "loss": 0.6383,
      "step": 948
    },
    {
      "epoch": 8.4070796460177,
      "grad_norm": 24.20329475402832,
      "learning_rate": 9.046214355948869e-06,
      "loss": 0.7669,
      "step": 950
    },
    {
      "epoch": 8.424778761061948,
      "grad_norm": 18.367490768432617,
      "learning_rate": 8.947885939036382e-06,
      "loss": 0.6304,
      "step": 952
    },
    {
      "epoch": 8.442477876106194,
      "grad_norm": 27.963157653808594,
      "learning_rate": 8.849557522123894e-06,
      "loss": 0.7402,
      "step": 954
    },
    {
      "epoch": 8.460176991150442,
      "grad_norm": 19.442922592163086,
      "learning_rate": 8.751229105211405e-06,
      "loss": 0.6278,
      "step": 956
    },
    {
      "epoch": 8.47787610619469,
      "grad_norm": 13.04371166229248,
      "learning_rate": 8.652900688298919e-06,
      "loss": 0.7028,
      "step": 958
    },
    {
      "epoch": 8.495575221238939,
      "grad_norm": 25.378374099731445,
      "learning_rate": 8.554572271386432e-06,
      "loss": 0.6889,
      "step": 960
    },
    {
      "epoch": 8.513274336283185,
      "grad_norm": 19.712556838989258,
      "learning_rate": 8.456243854473943e-06,
      "loss": 0.581,
      "step": 962
    },
    {
      "epoch": 8.530973451327434,
      "grad_norm": 20.15717315673828,
      "learning_rate": 8.357915437561457e-06,
      "loss": 0.6877,
      "step": 964
    },
    {
      "epoch": 8.548672566371682,
      "grad_norm": 24.568340301513672,
      "learning_rate": 8.259587020648968e-06,
      "loss": 0.5734,
      "step": 966
    },
    {
      "epoch": 8.56637168141593,
      "grad_norm": 20.580244064331055,
      "learning_rate": 8.16125860373648e-06,
      "loss": 0.695,
      "step": 968
    },
    {
      "epoch": 8.584070796460177,
      "grad_norm": 25.639179229736328,
      "learning_rate": 8.062930186823993e-06,
      "loss": 0.7367,
      "step": 970
    },
    {
      "epoch": 8.601769911504425,
      "grad_norm": 14.882777214050293,
      "learning_rate": 7.964601769911505e-06,
      "loss": 0.7085,
      "step": 972
    },
    {
      "epoch": 8.619469026548673,
      "grad_norm": 25.450916290283203,
      "learning_rate": 7.866273352999016e-06,
      "loss": 0.7158,
      "step": 974
    },
    {
      "epoch": 8.63716814159292,
      "grad_norm": 17.252824783325195,
      "learning_rate": 7.76794493608653e-06,
      "loss": 0.722,
      "step": 976
    },
    {
      "epoch": 8.654867256637168,
      "grad_norm": 17.026081085205078,
      "learning_rate": 7.669616519174041e-06,
      "loss": 0.7374,
      "step": 978
    },
    {
      "epoch": 8.672566371681416,
      "grad_norm": 17.954788208007812,
      "learning_rate": 7.571288102261554e-06,
      "loss": 0.7943,
      "step": 980
    },
    {
      "epoch": 8.690265486725664,
      "grad_norm": 14.818877220153809,
      "learning_rate": 7.472959685349067e-06,
      "loss": 0.687,
      "step": 982
    },
    {
      "epoch": 8.70796460176991,
      "grad_norm": 14.548257827758789,
      "learning_rate": 7.374631268436579e-06,
      "loss": 0.6824,
      "step": 984
    },
    {
      "epoch": 8.725663716814159,
      "grad_norm": 26.486650466918945,
      "learning_rate": 7.27630285152409e-06,
      "loss": 0.7428,
      "step": 986
    },
    {
      "epoch": 8.743362831858407,
      "grad_norm": 14.338964462280273,
      "learning_rate": 7.1779744346116035e-06,
      "loss": 0.6337,
      "step": 988
    },
    {
      "epoch": 8.761061946902656,
      "grad_norm": 11.065580368041992,
      "learning_rate": 7.079646017699115e-06,
      "loss": 0.7577,
      "step": 990
    },
    {
      "epoch": 8.778761061946902,
      "grad_norm": 20.75477409362793,
      "learning_rate": 6.9813176007866275e-06,
      "loss": 0.8219,
      "step": 992
    },
    {
      "epoch": 8.79646017699115,
      "grad_norm": 29.077869415283203,
      "learning_rate": 6.88298918387414e-06,
      "loss": 0.6982,
      "step": 994
    },
    {
      "epoch": 8.814159292035399,
      "grad_norm": 19.636323928833008,
      "learning_rate": 6.784660766961652e-06,
      "loss": 0.7256,
      "step": 996
    },
    {
      "epoch": 8.831858407079647,
      "grad_norm": 34.08713912963867,
      "learning_rate": 6.686332350049164e-06,
      "loss": 0.6156,
      "step": 998
    },
    {
      "epoch": 8.849557522123893,
      "grad_norm": 21.24959945678711,
      "learning_rate": 6.588003933136677e-06,
      "loss": 0.7501,
      "step": 1000
    },
    {
      "epoch": 8.867256637168142,
      "grad_norm": 24.227680206298828,
      "learning_rate": 6.489675516224189e-06,
      "loss": 0.7819,
      "step": 1002
    },
    {
      "epoch": 8.88495575221239,
      "grad_norm": 19.83379364013672,
      "learning_rate": 6.3913470993117005e-06,
      "loss": 0.6253,
      "step": 1004
    },
    {
      "epoch": 8.902654867256636,
      "grad_norm": 17.381193161010742,
      "learning_rate": 6.293018682399214e-06,
      "loss": 0.7019,
      "step": 1006
    },
    {
      "epoch": 8.920353982300885,
      "grad_norm": 19.214319229125977,
      "learning_rate": 6.194690265486726e-06,
      "loss": 0.7062,
      "step": 1008
    },
    {
      "epoch": 8.938053097345133,
      "grad_norm": 13.928613662719727,
      "learning_rate": 6.096361848574239e-06,
      "loss": 0.6825,
      "step": 1010
    },
    {
      "epoch": 8.955752212389381,
      "grad_norm": 29.10097885131836,
      "learning_rate": 5.99803343166175e-06,
      "loss": 0.6878,
      "step": 1012
    },
    {
      "epoch": 8.973451327433628,
      "grad_norm": 24.121362686157227,
      "learning_rate": 5.899705014749263e-06,
      "loss": 0.7148,
      "step": 1014
    },
    {
      "epoch": 8.991150442477876,
      "grad_norm": 23.926654815673828,
      "learning_rate": 5.801376597836775e-06,
      "loss": 0.7509,
      "step": 1016
    },
    {
      "epoch": 9.0,
      "eval_loss": 0.7565978169441223,
      "eval_map": 0.5582,
      "eval_map_50": 0.9326,
      "eval_map_75": 0.6075,
      "eval_map_large": 0.618,
      "eval_map_medium": 0.3937,
      "eval_map_per_class": 0.5582,
      "eval_map_small": 0.025,
      "eval_mar_1": 0.2923,
      "eval_mar_10": 0.6329,
      "eval_mar_100": 0.697,
      "eval_mar_100_per_class": 0.697,
      "eval_mar_large": 0.7497,
      "eval_mar_medium": 0.5641,
      "eval_mar_small": 0.3,
      "eval_runtime": 3.7908,
      "eval_samples_per_second": 26.38,
      "eval_steps_per_second": 1.847,
      "step": 1017
    }
  ],
  "logging_steps": 2,
  "max_steps": 1130,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 5,
        "early_stopping_threshold": 0.01
      },
      "attributes": {
        "early_stopping_patience_counter": 0
      }
    },
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 3.3929164854288e+17,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}