{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 12,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.25,
      "learning_rate": 2e-05,
      "loss": 6.1431,
      "step": 1
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9594929736144978e-05,
      "loss": 6.2302,
      "step": 2
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.8412535328311813e-05,
      "loss": 4.7669,
      "step": 3
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.6548607339452853e-05,
      "loss": 4.3465,
      "step": 4
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.4154150130018867e-05,
      "loss": 5.9263,
      "step": 5
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.1423148382732854e-05,
      "loss": 4.6402,
      "step": 6
    },
    {
      "epoch": 1.75,
      "learning_rate": 8.576851617267151e-06,
      "loss": 3.9222,
      "step": 7
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.845849869981137e-06,
      "loss": 3.6988,
      "step": 8
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.4513926605471504e-06,
      "loss": 3.4682,
      "step": 9
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.587464671688187e-06,
      "loss": 3.4927,
      "step": 10
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.0507026385502747e-07,
      "loss": 3.4117,
      "step": 11
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0,
      "loss": 3.3472,
      "step": 12
    },
    {
      "epoch": 3.0,
      "step": 12,
      "total_flos": 993619813072896.0,
      "train_loss": 4.44948794444402,
      "train_runtime": 46.3553,
      "train_samples_per_second": 0.971,
      "train_steps_per_second": 0.259
    }
  ],
  "max_steps": 12,
  "num_train_epochs": 3,
  "total_flos": 993619813072896.0,
  "trial_name": null,
  "trial_params": null
}