trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 3399,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.44,
      "learning_rate": 4.265960576640188e-05,
      "loss": 2.3769,
      "step": 500
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.5304501323918806e-05,
      "loss": 1.6449,
      "step": 1000
    },
    {
      "epoch": 1.32,
      "learning_rate": 2.794939688143572e-05,
      "loss": 1.3448,
      "step": 1500
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.0594292438952634e-05,
      "loss": 1.2324,
      "step": 2000
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.3239187996469552e-05,
      "loss": 1.08,
      "step": 2500
    },
    {
      "epoch": 2.65,
      "learning_rate": 5.884083553986467e-06,
      "loss": 0.9944,
      "step": 3000
    },
    {
      "epoch": 3.0,
      "step": 3399,
      "total_flos": 1.7867117154729984e+16,
      "train_loss": 0.10813238958149456,
      "train_runtime": 76.6572,
      "train_samples_per_second": 2481.45,
      "train_steps_per_second": 44.34
    }
  ],
  "logging_steps": 500,
  "max_steps": 3399,
  "num_train_epochs": 3,
  "save_steps": 1000,
  "total_flos": 1.7867117154729984e+16,
  "trial_name": null,
  "trial_params": null
}