trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 50.0,
  "eval_steps": 500,
  "global_step": 1200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 20.83,
      "learning_rate": 2.916666666666667e-05,
      "loss": 0.5624,
      "step": 500
    },
    {
      "epoch": 41.67,
      "learning_rate": 8.333333333333334e-06,
      "loss": 0.0146,
      "step": 1000
    },
    {
      "epoch": 50.0,
      "step": 1200,
      "total_flos": 2.93958851328e+16,
      "train_loss": 0.24115766083200774,
      "train_runtime": 850.2688,
      "train_samples_per_second": 176.415,
      "train_steps_per_second": 1.411
    }
  ],
  "logging_steps": 500,
  "max_steps": 1200,
  "num_train_epochs": 50,
  "save_steps": 500,
  "total_flos": 2.93958851328e+16,
  "trial_name": null,
  "trial_params": null
}