trainer_state.json

{
  "best_metric": 0.8025827407836914,
  "best_model_checkpoint": "sean_test_out/checkpoint-903",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 903,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.55,
      "grad_norm": 418741.09375,
      "learning_rate": 2.6777408637873754e-05,
      "loss": 0.975,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6483,
      "eval_loss": 0.8025827407836914,
      "eval_runtime": 33.2114,
      "eval_samples_per_second": 301.102,
      "eval_steps_per_second": 4.185,
      "step": 903
    }
  ],
  "logging_steps": 500,
  "max_steps": 903,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 500,
  "total_flos": 2.650663586377728e+16,
  "train_batch_size": 72,
  "trial_name": null,
  "trial_params": null
}