trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "global_step": 18,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.11,
      "learning_rate": 8.888888888888889e-05,
      "loss": 0.5675,
      "step": 10
    },
    {
      "epoch": 2.0,
      "step": 18,
      "total_flos": 2.014809787920384e+16,
      "train_loss": 0.39612122376759845,
      "train_runtime": 584.0146,
      "train_samples_per_second": 0.445,
      "train_steps_per_second": 0.031
    }
  ],
  "max_steps": 18,
  "num_train_epochs": 2,
  "total_flos": 2.014809787920384e+16,
  "trial_name": null,
  "trial_params": null
}