trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 540,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.85,
      "learning_rate": 3.148148148148148e-05,
      "loss": 3.4743,
      "step": 200
    },
    {
      "epoch": 3.7,
      "learning_rate": 1.2962962962962962e-05,
      "loss": 3.4332,
      "step": 400
    },
    {
      "epoch": 5.0,
      "step": 540,
      "total_flos": 0.0,
      "train_loss": 3.448119721589265,
      "train_runtime": 810.4417,
      "train_samples_per_second": 10.642,
      "train_steps_per_second": 0.666
    }
  ],
  "logging_steps": 200,
  "max_steps": 540,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 0.0,
  "trial_name": null,
  "trial_params": null
}