{
  "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": 1.8121609687805176,
      "learning_rate": 4.400558387583305e-05,
      "loss": 0.6264,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5072075128555298,
      "eval_loss": 0.6931472420692444,
      "eval_runtime": 9.1548,
      "eval_samples_per_second": 325.839,
      "eval_steps_per_second": 20.426,
      "step": 746
    },
    {
      "epoch": 1.3404825737265416,
      "grad_norm": 1.8386740684509277,
      "learning_rate": 4.0844263195097915e-05,
      "loss": 0.7039,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.49882668256759644,
      "eval_loss": 0.6931472420692444,
      "eval_runtime": 9.1028,
      "eval_samples_per_second": 327.702,
      "eval_steps_per_second": 20.543,
      "step": 1492
    },
    {
      "epoch": 2.0107238605898123,
      "grad_norm": 1.775593876838684,
      "learning_rate": 3.768294251436278e-05,
      "loss": 0.7008,
      "step": 1500
    },
    {
      "epoch": 2.680965147453083,
      "grad_norm": 1.6979082822799683,
      "learning_rate": 3.452162183362765e-05,
      "loss": 0.7034,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.5015085339546204,
      "eval_loss": 0.6931472420692444,
      "eval_runtime": 9.1903,
      "eval_samples_per_second": 324.583,
      "eval_steps_per_second": 20.348,
      "step": 2238
    },
    {
      "epoch": 3.351206434316354,
      "grad_norm": 1.7350631952285767,
      "learning_rate": 3.136030115289252e-05,
      "loss": 0.7017,
      "step": 2500
    }
  ],
  "logging_steps": 500,
  "max_steps": 7460,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 6517632906564732.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 4.716690455656818e-05,
    "per_device_train_batch_size": 16
  }
}