{
  "best_metric": 0.6108968257904053,
  "best_model_checkpoint": "BERT-WMM/run-10/checkpoint-1066",
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 2132,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.94,
      "grad_norm": 11.003177642822266,
      "learning_rate": 3.5116101047118276e-06,
      "loss": 0.7883,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7370892018779343,
      "eval_loss": 0.6418591737747192,
      "eval_runtime": 2.0865,
      "eval_samples_per_second": 1020.838,
      "eval_steps_per_second": 64.222,
      "step": 533
    },
    {
      "epoch": 1.88,
      "grad_norm": 22.481332778930664,
      "learning_rate": 2.8608298777452175e-06,
      "loss": 0.5687,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7577464788732394,
      "eval_loss": 0.6108968257904053,
      "eval_runtime": 2.0305,
      "eval_samples_per_second": 1049.022,
      "eval_steps_per_second": 65.995,
      "step": 1066
    },
    {
      "epoch": 2.81,
      "grad_norm": 32.06578063964844,
      "learning_rate": 2.2100496507786074e-06,
      "loss": 0.4792,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7553990610328638,
      "eval_loss": 0.6331790089607239,
      "eval_runtime": 2.1296,
      "eval_samples_per_second": 1000.171,
      "eval_steps_per_second": 62.922,
      "step": 1599
    },
    {
      "epoch": 3.75,
      "grad_norm": 21.596628189086914,
      "learning_rate": 1.5592694238119975e-06,
      "loss": 0.4182,
      "step": 2000
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.7530516431924883,
      "eval_loss": 0.6648078560829163,
      "eval_runtime": 2.7647,
      "eval_samples_per_second": 770.434,
      "eval_steps_per_second": 48.469,
      "step": 2132
    }
  ],
  "logging_steps": 500,
  "max_steps": 3198,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 500,
  "total_flos": 867010100753160.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 4.162390331678438e-06,
    "num_train_epochs": 6,
    "per_device_train_batch_size": 16,
    "weight_decay": 0.000561727996440909
  }
}