{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "eval_steps": 500,
  "global_step": 1570,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 3.18,
      "learning_rate": 3.407643312101911e-05,
      "loss": 1.5984,
      "step": 500
    },
    {
      "epoch": 6.37,
      "learning_rate": 1.8152866242038215e-05,
      "loss": 0.2991,
      "step": 1000
    },
    {
      "epoch": 9.55,
      "learning_rate": 2.229299363057325e-06,
      "loss": 0.1001,
      "step": 1500
    },
    {
      "epoch": 10.0,
      "step": 1570,
      "total_flos": 4899491251200000.0,
      "train_loss": 0.6390929455969744,
      "train_runtime": 233.8699,
      "train_samples_per_second": 213.794,
      "train_steps_per_second": 6.713
    }
  ],
  "logging_steps": 500,
  "max_steps": 1570,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 4899491251200000.0,
  "trial_name": null,
  "trial_params": null
}