{
  "best_metric": 2.7611162662506104,
  "best_model_checkpoint": "models/opt-babylm2-clean-20-epochs-earlystop_seed-42_1e-3/checkpoint-43960",
  "epoch": 19.995451444166477,
  "eval_steps": 500,
  "global_step": 43960,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.45485558335228565,
      "grad_norm": 0.429352343082428,
      "learning_rate": 3.125e-05,
      "loss": 5.6789,
      "step": 1000
    },
    {
      "epoch": 0.9097111667045713,
      "grad_norm": 0.6156922578811646,
      "learning_rate": 6.25e-05,
      "loss": 4.24,
      "step": 2000
    },
    {
      "epoch": 0.9997725722083238,
      "eval_accuracy": 0.3447452490930147,
      "eval_loss": 3.9528262615203857,
      "eval_runtime": 118.313,
      "eval_samples_per_second": 500.3,
      "eval_steps_per_second": 7.818,
      "step": 2198
    },
    {
      "epoch": 1.364566750056857,
      "grad_norm": 0.531734824180603,
      "learning_rate": 9.375e-05,
      "loss": 3.8229,
      "step": 3000
    },
    {
      "epoch": 1.8194223334091426,
      "grad_norm": 0.5823336839675903,
      "learning_rate": 0.000125,
      "loss": 3.5581,
      "step": 4000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.39479023397438445,
      "eval_loss": 3.411952018737793,
      "eval_runtime": 118.0185,
      "eval_samples_per_second": 501.549,
      "eval_steps_per_second": 7.838,
      "step": 4397
    },
    {
      "epoch": 2.274277916761428,
      "grad_norm": 0.533871591091156,
      "learning_rate": 0.00015625,
      "loss": 3.3475,
      "step": 5000
    },
    {
      "epoch": 2.729133500113714,
      "grad_norm": 0.48165056109428406,
      "learning_rate": 0.0001875,
      "loss": 3.2191,
      "step": 6000
    },
    {
      "epoch": 2.999772572208324,
      "eval_accuracy": 0.4179818284929866,
      "eval_loss": 3.180582046508789,
      "eval_runtime": 118.1479,
      "eval_samples_per_second": 500.999,
      "eval_steps_per_second": 7.829,
      "step": 6595
    },
    {
      "epoch": 3.1839890834659994,
      "grad_norm": 0.4450700879096985,
      "learning_rate": 0.00021875,
      "loss": 3.1207,
      "step": 7000
    },
    {
      "epoch": 3.6388446668182852,
      "grad_norm": 0.3985724449157715,
      "learning_rate": 0.00025,
      "loss": 3.0495,
      "step": 8000
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.42872771625206374,
      "eval_loss": 3.0720248222351074,
      "eval_runtime": 118.2962,
      "eval_samples_per_second": 500.371,
      "eval_steps_per_second": 7.819,
      "step": 8794
    },
    {
      "epoch": 4.093700250170571,
      "grad_norm": 0.3782784938812256,
      "learning_rate": 0.00028125000000000003,
      "loss": 2.9985,
      "step": 9000
    },
    {
      "epoch": 4.548555833522856,
      "grad_norm": 0.3812876343727112,
      "learning_rate": 0.0003125,
      "loss": 2.9458,
      "step": 10000
    },
    {
      "epoch": 4.9997725722083235,
      "eval_accuracy": 0.4353868037281138,
      "eval_loss": 3.005370616912842,
      "eval_runtime": 118.0467,
      "eval_samples_per_second": 501.428,
      "eval_steps_per_second": 7.836,
      "step": 10992
    },
    {
      "epoch": 5.003411416875142,
      "grad_norm": 0.34280985593795776,
      "learning_rate": 0.00034375,
      "loss": 2.9247,
      "step": 11000
    },
    {
      "epoch": 5.458267000227428,
      "grad_norm": 0.33273234963417053,
      "learning_rate": 0.000375,
      "loss": 2.8732,
      "step": 12000
    },
    {
      "epoch": 5.913122583579714,
      "grad_norm": 0.3201596736907959,
      "learning_rate": 0.00040621875,
      "loss": 2.8669,
      "step": 13000
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.43990927496826543,
      "eval_loss": 2.966292381286621,
      "eval_runtime": 117.0736,
      "eval_samples_per_second": 505.597,
      "eval_steps_per_second": 7.901,
      "step": 13191
    },
    {
      "epoch": 6.367978166931999,
      "grad_norm": 0.3063608705997467,
      "learning_rate": 0.00043746875000000003,
      "loss": 2.8263,
      "step": 14000
    },
    {
      "epoch": 6.822833750284285,
      "grad_norm": 0.2880852520465851,
      "learning_rate": 0.0004686875,
      "loss": 2.8256,
      "step": 15000
    },
    {
      "epoch": 6.9997725722083235,
      "eval_accuracy": 0.4429723545047158,
      "eval_loss": 2.938157081604004,
      "eval_runtime": 117.2958,
      "eval_samples_per_second": 504.639,
      "eval_steps_per_second": 7.886,
      "step": 15389
    },
    {
      "epoch": 7.2776893336365704,
      "grad_norm": 0.26923319697380066,
      "learning_rate": 0.0004999375,
      "loss": 2.7938,
      "step": 16000
    },
    {
      "epoch": 7.732544916988856,
      "grad_norm": 0.25492221117019653,
      "learning_rate": 0.00053115625,
      "loss": 2.79,
      "step": 17000
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.4452082157366258,
      "eval_loss": 2.919893741607666,
      "eval_runtime": 117.0581,
      "eval_samples_per_second": 505.664,
      "eval_steps_per_second": 7.902,
      "step": 17588
    },
    {
      "epoch": 8.187400500341141,
      "grad_norm": 0.251396507024765,
      "learning_rate": 0.0005623749999999999,
      "loss": 2.7715,
      "step": 18000
    },
    {
      "epoch": 8.642256083693427,
      "grad_norm": 0.2381746470928192,
      "learning_rate": 0.000593625,
      "loss": 2.7624,
      "step": 19000
    },
    {
      "epoch": 8.999772572208323,
      "eval_accuracy": 0.4468282677309334,
      "eval_loss": 2.9052329063415527,
      "eval_runtime": 117.1996,
      "eval_samples_per_second": 505.053,
      "eval_steps_per_second": 7.893,
      "step": 19786
    },
    {
      "epoch": 9.097111667045713,
      "grad_norm": 0.2604175806045532,
      "learning_rate": 0.000624875,
      "loss": 2.7571,
      "step": 20000
    },
    {
      "epoch": 9.551967250397999,
      "grad_norm": 0.22635479271411896,
      "learning_rate": 0.000656125,
      "loss": 2.7361,
      "step": 21000
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.44817423658205,
      "eval_loss": 2.891494035720825,
      "eval_runtime": 117.1018,
      "eval_samples_per_second": 505.475,
      "eval_steps_per_second": 7.899,
      "step": 21985
    },
    {
      "epoch": 10.006822833750284,
      "grad_norm": 0.23204845190048218,
      "learning_rate": 0.0006873125,
      "loss": 2.7513,
      "step": 22000
    },
    {
      "epoch": 10.46167841710257,
      "grad_norm": 0.22187474370002747,
      "learning_rate": 0.0007185625,
      "loss": 2.714,
      "step": 23000
    },
    {
      "epoch": 10.916534000454856,
      "grad_norm": 0.20946632325649261,
      "learning_rate": 0.0007498125,
      "loss": 2.7354,
      "step": 24000
    },
    {
      "epoch": 10.999772572208323,
      "eval_accuracy": 0.44908744954935614,
      "eval_loss": 2.8843252658843994,
      "eval_runtime": 117.0503,
      "eval_samples_per_second": 505.697,
      "eval_steps_per_second": 7.903,
      "step": 24183
    },
    {
      "epoch": 11.371389583807142,
      "grad_norm": 0.2115669995546341,
      "learning_rate": 0.0007810625,
      "loss": 2.7054,
      "step": 25000
    },
    {
      "epoch": 11.826245167159428,
      "grad_norm": 0.19731654226779938,
      "learning_rate": 0.00081228125,
      "loss": 2.7225,
      "step": 26000
    },
    {
      "epoch": 12.0,
      "eval_accuracy": 0.4499743606051692,
      "eval_loss": 2.8777432441711426,
      "eval_runtime": 117.2326,
      "eval_samples_per_second": 504.911,
      "eval_steps_per_second": 7.89,
      "step": 26382
    },
    {
      "epoch": 12.281100750511712,
      "grad_norm": 0.1893465369939804,
      "learning_rate": 0.0008435000000000001,
      "loss": 2.6994,
      "step": 27000
    },
    {
      "epoch": 12.735956333863998,
      "grad_norm": 0.2169094681739807,
      "learning_rate": 0.0008747500000000001,
      "loss": 2.7092,
      "step": 28000
    },
    {
      "epoch": 12.999772572208323,
      "eval_accuracy": 0.4505424023914201,
      "eval_loss": 2.870769739151001,
      "eval_runtime": 117.0845,
      "eval_samples_per_second": 505.549,
      "eval_steps_per_second": 7.9,
      "step": 28580
    },
    {
      "epoch": 13.190811917216283,
      "grad_norm": 0.1946067363023758,
      "learning_rate": 0.000906,
      "loss": 2.6995,
      "step": 29000
    },
    {
      "epoch": 13.64566750056857,
      "grad_norm": 0.18853700160980225,
      "learning_rate": 0.00093725,
      "loss": 2.6987,
      "step": 30000
    },
    {
      "epoch": 14.0,
      "eval_accuracy": 0.4508735944709009,
      "eval_loss": 2.8687899112701416,
      "eval_runtime": 117.3614,
      "eval_samples_per_second": 504.356,
      "eval_steps_per_second": 7.882,
      "step": 30779
    },
    {
      "epoch": 14.100523083920855,
      "grad_norm": 0.19028180837631226,
      "learning_rate": 0.0009685000000000001,
      "loss": 2.7009,
      "step": 31000
    },
    {
      "epoch": 14.555378667273141,
      "grad_norm": 0.1784777194261551,
      "learning_rate": 0.00099971875,
      "loss": 2.6894,
      "step": 32000
    },
    {
      "epoch": 14.999772572208323,
      "eval_accuracy": 0.4526506629141723,
      "eval_loss": 2.854236364364624,
      "eval_runtime": 117.0791,
      "eval_samples_per_second": 505.573,
      "eval_steps_per_second": 7.901,
      "step": 32977
    },
    {
      "epoch": 15.010234250625427,
      "grad_norm": 0.1875610202550888,
      "learning_rate": 0.0009171404682274248,
      "loss": 2.6972,
      "step": 33000
    },
    {
      "epoch": 15.465089833977713,
      "grad_norm": 0.18635216355323792,
      "learning_rate": 0.0008336120401337793,
      "loss": 2.647,
      "step": 34000
    },
    {
      "epoch": 15.919945417329998,
      "grad_norm": 0.19111545383930206,
      "learning_rate": 0.00075,
      "loss": 2.6561,
      "step": 35000
    },
    {
      "epoch": 16.0,
      "eval_accuracy": 0.4563634062896682,
      "eval_loss": 2.825805187225342,
      "eval_runtime": 117.2142,
      "eval_samples_per_second": 504.99,
      "eval_steps_per_second": 7.892,
      "step": 35176
    },
    {
      "epoch": 16.374801000682282,
      "grad_norm": 0.1833820343017578,
      "learning_rate": 0.0006663879598662207,
      "loss": 2.6017,
      "step": 36000
    },
    {
      "epoch": 16.82965658403457,
      "grad_norm": 0.18850316107273102,
      "learning_rate": 0.0005828595317725752,
      "loss": 2.6055,
      "step": 37000
    },
    {
      "epoch": 16.999772572208325,
      "eval_accuracy": 0.45953467479707116,
      "eval_loss": 2.800510883331299,
      "eval_runtime": 118.167,
      "eval_samples_per_second": 500.918,
      "eval_steps_per_second": 7.828,
      "step": 37374
    },
    {
      "epoch": 17.284512167386854,
      "grad_norm": 0.19247354567050934,
      "learning_rate": 0.000499247491638796,
      "loss": 2.557,
      "step": 38000
    },
    {
      "epoch": 17.73936775073914,
      "grad_norm": 0.18034116923809052,
      "learning_rate": 0.0004156354515050167,
      "loss": 2.5464,
      "step": 39000
    },
    {
      "epoch": 18.0,
      "eval_accuracy": 0.4627209824327082,
      "eval_loss": 2.7813873291015625,
      "eval_runtime": 118.2688,
      "eval_samples_per_second": 500.487,
      "eval_steps_per_second": 7.821,
      "step": 39573
    },
    {
      "epoch": 18.194223334091426,
      "grad_norm": 0.19630157947540283,
      "learning_rate": 0.0003320234113712375,
      "loss": 2.5107,
      "step": 40000
    },
    {
      "epoch": 18.649078917443713,
      "grad_norm": 0.20289246737957,
      "learning_rate": 0.000248494983277592,
      "loss": 2.4778,
      "step": 41000
    },
    {
      "epoch": 18.999772572208325,
      "eval_accuracy": 0.4658446822437584,
      "eval_loss": 2.7629892826080322,
      "eval_runtime": 118.2031,
      "eval_samples_per_second": 500.765,
      "eval_steps_per_second": 7.826,
      "step": 41771
    },
    {
      "epoch": 19.103934500795997,
      "grad_norm": 0.19990454614162445,
      "learning_rate": 0.00016496655518394648,
      "loss": 2.458,
      "step": 42000
    },
    {
      "epoch": 19.55879008414828,
      "grad_norm": 0.2090277075767517,
      "learning_rate": 8.135451505016722e-05,
      "loss": 2.4036,
      "step": 43000
    },
    {
      "epoch": 19.995451444166477,
      "eval_accuracy": 0.46725332517112805,
      "eval_loss": 2.7611162662506104,
      "eval_runtime": 118.1837,
      "eval_samples_per_second": 500.847,
      "eval_steps_per_second": 7.827,
      "step": 43960
    },
    {
      "epoch": 19.995451444166477,
      "step": 43960,
      "total_flos": 1.470253944471552e+18,
      "train_loss": 2.893489753646781,
      "train_runtime": 41720.7377,
      "train_samples_per_second": 269.801,
      "train_steps_per_second": 1.054
    }
  ],
  "logging_steps": 1000,
  "max_steps": 43960,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 20,
  "save_steps": 500,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 3,
        "early_stopping_threshold": 0.0
      },
      "attributes": {
        "early_stopping_patience_counter": 0
      }
    },
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 1.470253944471552e+18,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}