trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "eval_steps": 500,
  "global_step": 690,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 7.25,
      "learning_rate": 1.3768115942028985e-05,
      "loss": 0.0529,
      "step": 500
    },
    {
      "epoch": 10.0,
      "step": 690,
      "total_flos": 3359758649241600.0,
      "train_loss": 0.04075576412504998,
      "train_runtime": 147.9128,
      "train_samples_per_second": 18.66,
      "train_steps_per_second": 4.665
    }
  ],
  "logging_steps": 500,
  "max_steps": 690,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 3359758649241600.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}