generation_config.json

{
  "attn_implementation": "flash_attention_2",
  "bos_token_id": 151643,
  "do_sample": true,
  "eos_token_id": [
    151645,
    151643
  ],
  "pad_token_id": 151643,
  "repetition_penalty": 1.08,
  "rope_theta": 5000000,
  "temperature": 0.65,
  "top_k": 20,
  "top_p": 0.8,
  "transformers_version": "4.46.2",
  "use_cache": true
}