trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 729,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.06,
      "learning_rate": 1.570644718792867e-05,
      "loss": 2.9439,
      "step": 500
    },
    {
      "epoch": 3.0,
      "step": 729,
      "total_flos": 6806063508836352.0,
      "train_loss": 2.8308688191229425,
      "train_runtime": 860.89,
      "train_samples_per_second": 216.428,
      "train_steps_per_second": 0.847
    }
  ],
  "max_steps": 729,
  "num_train_epochs": 3,
  "total_flos": 6806063508836352.0,
  "trial_name": null,
  "trial_params": null
}