trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 690,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.17,
      "learning_rate": 1.3768115942028985e-05,
      "loss": 0.3782,
      "step": 500
    },
    {
      "epoch": 3.0,
      "step": 690,
      "total_flos": 723818513295360.0,
      "train_loss": 0.29736072982566947,
      "train_runtime": 72.3445,
      "train_samples_per_second": 152.106,
      "train_steps_per_second": 9.538
    }
  ],
  "max_steps": 690,
  "num_train_epochs": 3,
  "total_flos": 723818513295360.0,
  "trial_name": null,
  "trial_params": null
}