trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 3357,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.45,
      "learning_rate": 4.2552874590408106e-05,
      "loss": 0.5766,
      "step": 500
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.5105749180816204e-05,
      "loss": 0.2461,
      "step": 1000
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.7658623771224308e-05,
      "loss": 0.2071,
      "step": 1500
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.021149836163241e-05,
      "loss": 0.1864,
      "step": 2000
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.2764372952040512e-05,
      "loss": 0.1738,
      "step": 2500
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.317247542448615e-06,
      "loss": 0.167,
      "step": 3000
    },
    {
      "epoch": 3.0,
      "step": 3357,
      "total_flos": 1.9225322988957696e+16,
      "train_loss": 0.2496061805177381,
      "train_runtime": 2390.8058,
      "train_samples_per_second": 22.461,
      "train_steps_per_second": 1.404
    }
  ],
  "max_steps": 3357,
  "num_train_epochs": 3,
  "total_flos": 1.9225322988957696e+16,
  "trial_name": null,
  "trial_params": null
}