trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2379,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.63,
      "learning_rate": 3.949138293400589e-05,
      "loss": 6.5299,
      "step": 500
    },
    {
      "epoch": 1.26,
      "learning_rate": 2.898276586801177e-05,
      "loss": 3.4523,
      "step": 1000
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.8474148802017654e-05,
      "loss": 3.2633,
      "step": 1500
    },
    {
      "epoch": 2.52,
      "learning_rate": 7.965531736023538e-06,
      "loss": 3.1001,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "step": 2379,
      "total_flos": 621613744128000.0,
      "train_loss": 3.9267082967994695,
      "train_runtime": 876.5017,
      "train_samples_per_second": 2.714,
      "train_steps_per_second": 2.714
    }
  ],
  "max_steps": 2379,
  "num_train_epochs": 3,
  "total_flos": 621613744128000.0,
  "trial_name": null,
  "trial_params": null
}