trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 930,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.69,
      "learning_rate": 2.3118279569892472e-05,
      "loss": 1.9451,
      "step": 500
    },
    {
      "epoch": 5.0,
      "step": 930,
      "total_flos": 426680294150880.0,
      "train_loss": 1.813177949638777,
      "train_runtime": 4079.9381,
      "train_samples_per_second": 1.824,
      "train_steps_per_second": 0.228
    }
  ],
  "logging_steps": 500,
  "max_steps": 930,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 426680294150880.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}