generation_config.json

{
  "align_inputs_to_blocks": true,
  "block_size": 4,
  "bos_token_id": 50256,
  "confidence_based_noising": true,
  "confidence_margin_based_noising": false,
  "confidence_threshold": 1000000.0,
  "eos_token_id": 50256,
  "first_hitting": true,
  "min_t": 1e-05,
  "num_steps": 4,
  "pad_token_id": 50256,
  "sampling_strategy": "predict_and_noise",
  "transformers_version": "4.52.3",
  "use_model_output_cache": true
}