diff --git a/python/llm/src/ipex_llm/serving/fastchat/ipex_llm_worker.py b/python/llm/src/ipex_llm/serving/fastchat/ipex_llm_worker.py index b357c8d3..9fbe1cb7 100644 --- a/python/llm/src/ipex_llm/serving/fastchat/ipex_llm_worker.py +++ b/python/llm/src/ipex_llm/serving/fastchat/ipex_llm_worker.py @@ -100,7 +100,9 @@ class BigDLLLMWorker(BaseModelWorker): temperature = float(params.get("temperature", 1.0)) repetition_penalty = float(params.get("repetition_penalty", 1.0)) top_p = float(params.get("top_p", 1.0)) - top_k = int(params.get("top_k", 0)) # 0 means disable + top_k = int(params.get("top_k", 1)) + if top_k == -1: + top_k = 1 max_new_tokens = int(params.get("max_new_tokens", 256)) echo = bool(params.get("echo", True)) stop_str = params.get("stop", None)