generation_config.json

{
  "device": "cuda",
  "do_sample": true,
  "eos_token_id": [
    151329,
    151336,
    151338
  ],
  "max_length": 128000,
  "pad_token_id": 151329,
  "temperature": 0.8,
  "top_p": 0.8,
  "transformers_version": "4.47.0"
}