trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.996605054696341,
  "global_step": 993,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.51,
      "learning_rate": 1.0238836967808932e-05,
      "loss": 0.1087,
      "step": 500
    },
    {
      "epoch": 3.0,
      "step": 993,
      "total_flos": 3.075776576319652e+17,
      "train_loss": 0.059486828541827706,
      "train_runtime": 1614.0773,
      "train_samples_per_second": 39.414,
      "train_steps_per_second": 0.615
    }
  ],
  "max_steps": 993,
  "num_train_epochs": 3,
  "total_flos": 3.075776576319652e+17,
  "trial_name": null,
  "trial_params": null
}