trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "global_step": 880,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.45,
      "eval_HasAns_exact": 43.43891402714932,
      "eval_HasAns_f1": 69.1661607072733,
      "eval_HasAns_total": 221,
      "eval_best_exact": 43.43891402714932,
      "eval_best_exact_thresh": 0.0,
      "eval_best_f1": 69.1661607072733,
      "eval_best_f1_thresh": 0.0,
      "eval_exact": 43.43891402714932,
      "eval_f1": 69.1661607072733,
      "eval_total": 221,
      "step": 200
    },
    {
      "epoch": 0.91,
      "eval_HasAns_exact": 49.321266968325794,
      "eval_HasAns_f1": 70.42596830206043,
      "eval_HasAns_total": 221,
      "eval_best_exact": 49.321266968325794,
      "eval_best_exact_thresh": 0.0,
      "eval_best_f1": 70.42596830206043,
      "eval_best_f1_thresh": 0.0,
      "eval_exact": 49.321266968325794,
      "eval_f1": 70.42596830206043,
      "eval_total": 221,
      "step": 400
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.0795454545454547e-05,
      "loss": 0.8745,
      "step": 500
    },
    {
      "epoch": 1.36,
      "eval_HasAns_exact": 45.70135746606335,
      "eval_HasAns_f1": 67.08475754056293,
      "eval_HasAns_total": 221,
      "eval_best_exact": 45.70135746606335,
      "eval_best_exact_thresh": 0.0,
      "eval_best_f1": 67.08475754056293,
      "eval_best_f1_thresh": 0.0,
      "eval_exact": 45.70135746606335,
      "eval_f1": 67.08475754056293,
      "eval_total": 221,
      "step": 600
    },
    {
      "epoch": 1.82,
      "eval_HasAns_exact": 45.248868778280546,
      "eval_HasAns_f1": 67.36800189264167,
      "eval_HasAns_total": 221,
      "eval_best_exact": 45.248868778280546,
      "eval_best_exact_thresh": 0.0,
      "eval_best_f1": 67.36800189264167,
      "eval_best_f1_thresh": 0.0,
      "eval_exact": 45.248868778280546,
      "eval_f1": 67.36800189264167,
      "eval_total": 221,
      "step": 800
    },
    {
      "epoch": 2.0,
      "step": 880,
      "total_flos": 1378079095025664.0,
      "train_loss": 0.7235306653109463,
      "train_runtime": 418.985,
      "train_samples_per_second": 16.783,
      "train_steps_per_second": 2.1
    }
  ],
  "max_steps": 880,
  "num_train_epochs": 2,
  "total_flos": 1378079095025664.0,
  "trial_name": null,
  "trial_params": null
}