trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 30.0,
  "global_step": 660,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 22.73,
      "learning_rate": 1.2121212121212122e-05,
      "loss": 1.5729,
      "step": 500
    },
    {
      "epoch": 30.0,
      "step": 660,
      "total_flos": 2460546318827520.0,
      "train_loss": 1.454706689083215,
      "train_runtime": 652.8678,
      "train_samples_per_second": 4.044,
      "train_steps_per_second": 1.011
    }
  ],
  "max_steps": 660,
  "num_train_epochs": 30,
  "total_flos": 2460546318827520.0,
  "trial_name": null,
  "trial_params": null
}