trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 29.85576923076923,
  "eval_steps": 500,
  "global_step": 690,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 21.63,
      "learning_rate": 0.0002753623188405797,
      "loss": 0.0731,
      "step": 500
    },
    {
      "epoch": 29.86,
      "step": 690,
      "total_flos": 1.9441172348928e+16,
      "train_loss": 0.05997911328854768,
      "train_runtime": 6173.8501,
      "train_samples_per_second": 6.055,
      "train_steps_per_second": 0.112
    }
  ],
  "logging_steps": 500,
  "max_steps": 690,
  "num_train_epochs": 30,
  "save_steps": 500,
  "total_flos": 1.9441172348928e+16,
  "trial_name": null,
  "trial_params": null
}