{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.351206434316354,
  "eval_steps": 500,
  "global_step": 2500,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.6702412868632708,
      "grad_norm": 20.911739349365234,
      "learning_rate": 1.9406388835210846e-05,
      "loss": 0.5519,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7670130729675293,
      "eval_loss": 0.4711463153362274,
      "eval_runtime": 8.8245,
      "eval_samples_per_second": 338.035,
      "eval_steps_per_second": 21.191,
      "step": 746
    },
    {
      "epoch": 1.3404825737265416,
      "grad_norm": 12.272890090942383,
      "learning_rate": 1.801225170624455e-05,
      "loss": 0.3843,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7767348289489746,
      "eval_loss": 0.5532987117767334,
      "eval_runtime": 8.9725,
      "eval_samples_per_second": 332.46,
      "eval_steps_per_second": 20.841,
      "step": 1492
    },
    {
      "epoch": 2.0107238605898123,
      "grad_norm": 15.580121994018555,
      "learning_rate": 1.6618114577278254e-05,
      "loss": 0.2751,
      "step": 1500
    },
    {
      "epoch": 2.680965147453083,
      "grad_norm": 8.400450706481934,
      "learning_rate": 1.5223977448311958e-05,
      "loss": 0.1119,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7800871729850769,
      "eval_loss": 0.8922383189201355,
      "eval_runtime": 8.9532,
      "eval_samples_per_second": 333.175,
      "eval_steps_per_second": 20.886,
      "step": 2238
    },
    {
      "epoch": 3.351206434316354,
      "grad_norm": 42.92605972290039,
      "learning_rate": 1.382984031934566e-05,
      "loss": 0.0723,
      "step": 2500
    }
  ],
  "logging_steps": 500,
  "max_steps": 7460,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 5932068918206580.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 2.0800525964177143e-05,
    "per_device_train_batch_size": 16
  }
}