trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.9205607476635516,
  "global_step": 1250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.17,
      "learning_rate": 5e-06,
      "loss": 2.3429,
      "step": 500
    },
    {
      "epoch": 2.34,
      "learning_rate": 5e-06,
      "loss": 2.0258,
      "step": 1000
    }
  ],
  "max_steps": 2000,
  "num_train_epochs": 5,
  "total_flos": 2.91632922258047e+17,
  "trial_name": null,
  "trial_params": null
}