trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9863013698630136,
  "global_step": 27,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.73,
      "learning_rate": 2.5925925925925925e-05,
      "loss": 2.2144,
      "step": 20
    },
    {
      "epoch": 0.99,
      "step": 27,
      "total_flos": 1.5437027289858048e+17,
      "train_loss": 2.189213894031666,
      "train_runtime": 168.693,
      "train_samples_per_second": 331.964,
      "train_steps_per_second": 0.16
    }
  ],
  "max_steps": 27,
  "num_train_epochs": 1,
  "total_flos": 1.5437027289858048e+17,
  "trial_name": null,
  "trial_params": null
}