generation_config.json

{
  "chat_format": "chatml",
  "do_sample": false,                  // Disable sampling for faster inference
  "eos_token_id": 151643,
  "max_new_tokens": 256,               // Reduce the number of tokens to generate
  "max_window_size": 6144,
  "pad_token_id": 151643,
  "top_k": 50,                         // Use a reasonable value for top_k
  "top_p": 0.9,                        // Increase top_p to allow more tokens
  "transformers_version": "4.31.0"
}