trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 40.0,
  "eval_steps": 500,
  "global_step": 1160,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 17.24,
      "learning_rate": 2.844827586206897e-05,
      "loss": 2.1876,
      "step": 500
    },
    {
      "epoch": 34.48,
      "learning_rate": 6.896551724137932e-06,
      "loss": 1.4227,
      "step": 1000
    },
    {
      "epoch": 40.0,
      "step": 1160,
      "total_flos": 2362079969280000.0,
      "train_loss": 1.7243760996851427,
      "train_runtime": 861.0101,
      "train_samples_per_second": 5.25,
      "train_steps_per_second": 1.347
    }
  ],
  "logging_steps": 500,
  "max_steps": 1160,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 40,
  "save_steps": 500,
  "total_flos": 2362079969280000.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}