trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "global_step": 20,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.0,
      "learning_rate": 0.0001,
      "loss": 1.1677,
      "step": 10
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0,
      "loss": 0.5814,
      "step": 20
    },
    {
      "epoch": 4.0,
      "step": 20,
      "total_flos": 2.4178150845628416e+16,
      "train_loss": 0.8745415687561036,
      "train_runtime": 612.1032,
      "train_samples_per_second": 0.51,
      "train_steps_per_second": 0.033
    }
  ],
  "max_steps": 20,
  "num_train_epochs": 4,
  "total_flos": 2.4178150845628416e+16,
  "trial_name": null,
  "trial_params": null
}