checkpoint-2000/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.074688796680498,
  "eval_steps": 500,
  "global_step": 2000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.52,
      "grad_norm": 7.744676113128662,
      "learning_rate": 4.135546334716459e-05,
      "loss": 0.9335,
      "step": 500
    },
    {
      "epoch": 1.04,
      "grad_norm": 0.05482616648077965,
      "learning_rate": 3.271092669432919e-05,
      "loss": 0.1947,
      "step": 1000
    },
    {
      "epoch": 1.56,
      "grad_norm": 28.68220329284668,
      "learning_rate": 2.406639004149378e-05,
      "loss": 0.0911,
      "step": 1500
    },
    {
      "epoch": 2.07,
      "grad_norm": 0.026414938271045685,
      "learning_rate": 1.5421853388658368e-05,
      "loss": 0.0557,
      "step": 2000
    }
  ],
  "logging_steps": 500,
  "max_steps": 2892,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 2119250816409600.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}