trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 100.0,
  "eval_steps": 500,
  "global_step": 2400,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 20.83,
      "learning_rate": 3.958333333333333e-05,
      "loss": 0.2896,
      "step": 500
    },
    {
      "epoch": 41.67,
      "learning_rate": 2.916666666666667e-05,
      "loss": 0.0171,
      "step": 1000
    },
    {
      "epoch": 62.5,
      "learning_rate": 1.8750000000000002e-05,
      "loss": 0.0055,
      "step": 1500
    },
    {
      "epoch": 83.33,
      "learning_rate": 8.333333333333334e-06,
      "loss": 0.0028,
      "step": 2000
    },
    {
      "epoch": 100.0,
      "step": 2400,
      "total_flos": 4968855705600000.0,
      "train_loss": 0.06646478484074275,
      "train_runtime": 1714.3663,
      "train_samples_per_second": 174.992,
      "train_steps_per_second": 1.4
    }
  ],
  "logging_steps": 500,
  "max_steps": 2400,
  "num_train_epochs": 100,
  "save_steps": 500,
  "total_flos": 4968855705600000.0,
  "trial_name": null,
  "trial_params": null
}