{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 672,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.23,
      "learning_rate": 1.2797619047619047e-05,
      "loss": 2.2489,
      "step": 500
    },
    {
      "epoch": 3.0,
      "step": 672,
      "total_flos": 923602387009536.0,
      "train_loss": 2.0268579437619163,
      "train_runtime": 167.7019,
      "train_samples_per_second": 39.928,
      "train_steps_per_second": 4.007
    }
  ],
  "max_steps": 672,
  "num_train_epochs": 3,
  "total_flos": 923602387009536.0,
  "trial_name": null,
  "trial_params": null
}