{
  "best_metric": 1.5045615434646606,
  "best_model_checkpoint": "outputs/checkpoint-46",
  "epoch": 0.9932523616734144,
  "eval_steps": 500,
  "global_step": 46,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.4318488529014845,
      "grad_norm": 0.5840024948120117,
      "learning_rate": 4e-05,
      "loss": 1.9547,
      "step": 20
    },
    {
      "epoch": 0.863697705802969,
      "grad_norm": 0.4871944487094879,
      "learning_rate": 8e-05,
      "loss": 1.8203,
      "step": 40
    },
    {
      "epoch": 0.9932523616734144,
      "eval_loss": 1.5045615434646606,
      "eval_runtime": 13.3054,
      "eval_samples_per_second": 27.959,
      "eval_steps_per_second": 3.532,
      "step": 46
    }
  ],
  "logging_steps": 20,
  "max_steps": 782,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 17,
  "save_steps": 500,
  "total_flos": 3531518210801664.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}