{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.3714285714285714,
  "eval_steps": 500,
  "global_step": 39,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.2,
      "learning_rate": 0.00012727272727272728,
      "loss": 1.3064,
      "step": 7
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.00019361702127659576,
      "loss": 1.2383,
      "step": 14
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00017872340425531915,
      "loss": 1.0564,
      "step": 21
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.00016382978723404257,
      "loss": 0.9892,
      "step": 28
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.00014893617021276596,
      "loss": 0.8992,
      "step": 35
    }
  ],
  "logging_steps": 7,
  "max_steps": 105,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 6821844363509760.0,
  "trial_name": null,
  "trial_params": null
}