trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.936857562408223,
  "eval_steps": 500,
  "global_step": 2000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.73,
      "grad_norm": 0.14444780349731445,
      "learning_rate": 5e-05,
      "loss": 0.0033,
      "step": 500
    },
    {
      "epoch": 1.47,
      "grad_norm": 0.22277051210403442,
      "learning_rate": 3.3797796500324045e-05,
      "loss": 0.0028,
      "step": 1000
    },
    {
      "epoch": 2.2,
      "grad_norm": 0.0007193220080807805,
      "learning_rate": 1.7595593000648088e-05,
      "loss": 0.0015,
      "step": 1500
    },
    {
      "epoch": 2.94,
      "grad_norm": 0.040156736969947815,
      "learning_rate": 1.3933895009721324e-06,
      "loss": 0.0007,
      "step": 2000
    }
  ],
  "logging_steps": 500,
  "max_steps": 2043,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 4179595948087296.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}