trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 2931,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.51,
      "learning_rate": 4.14704878880928e-05,
      "loss": 1.6585,
      "step": 500
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.294097577618561e-05,
      "loss": 1.534,
      "step": 1000
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.4411463664278406e-05,
      "loss": 1.504,
      "step": 1500
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.5881951552371206e-05,
      "loss": 1.4679,
      "step": 2000
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.3524394404640055e-06,
      "loss": 1.442,
      "step": 2500
    },
    {
      "epoch": 3.0,
      "step": 2931,
      "total_flos": 118513687363584.0,
      "train_loss": 1.504677425346778,
      "train_runtime": 1613.5134,
      "train_samples_per_second": 7.264,
      "train_steps_per_second": 1.817
    }
  ],
  "logging_steps": 500,
  "max_steps": 2931,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 118513687363584.0,
  "trial_name": null,
  "trial_params": null
}