trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 135,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.22,
      "learning_rate": 1.4074074074074075e-05,
      "loss": 4.9042,
      "step": 100
    },
    {
      "epoch": 3.0,
      "step": 135,
      "total_flos": 719870106009600.0,
      "train_loss": 4.75333997938368,
      "train_runtime": 38.4881,
      "train_samples_per_second": 779.461,
      "train_steps_per_second": 3.508
    }
  ],
  "logging_steps": 100,
  "max_steps": 135,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 719870106009600.0,
  "trial_name": null,
  "trial_params": null
}