trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 804,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.87,
      "learning_rate": 1.1343283582089553e-05,
      "loss": 0.4457,
      "step": 500
    },
    {
      "epoch": 3.0,
      "step": 804,
      "total_flos": 843698487893760.0,
      "train_loss": 0.344569457704155,
      "train_runtime": 294.7167,
      "train_samples_per_second": 87.043,
      "train_steps_per_second": 2.728
    }
  ],
  "max_steps": 804,
  "num_train_epochs": 3,
  "total_flos": 843698487893760.0,
  "trial_name": null,
  "trial_params": null
}