trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 1250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8,
      "grad_norm": 35.02853012084961,
      "learning_rate": 3.68e-05,
      "loss": 0.4341,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8725,
      "eval_loss": 0.44710320234298706,
      "eval_runtime": 2.8718,
      "eval_samples_per_second": 278.576,
      "eval_steps_per_second": 34.822,
      "step": 625
    },
    {
      "epoch": 1.6,
      "grad_norm": 1.4543651342391968,
      "learning_rate": 2.3466666666666667e-05,
      "loss": 0.2212,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.89125,
      "eval_loss": 0.45788365602493286,
      "eval_runtime": 2.5514,
      "eval_samples_per_second": 313.549,
      "eval_steps_per_second": 39.194,
      "step": 1250
    }
  ],
  "logging_steps": 500,
  "max_steps": 1875,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 325271042188800.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}