generation_config.json

{
  "_from_model_config": true,
  "bos_token_id": 2,
  "cache_implementation": "hybrid",
  "do_sample": true,
  "eos_token_id": 1,
  "max_length": 8192,
  "max_new_tokens": 512,
  "pad_token_id": 0,
  "temperature": 0.5,
  "top_k": 30,
  "top_p": 0.9,
  "transformers_version": "4.44.0"
}