remove usage of stop_token_ids

Signed-off-by: Ceng23333 <441651826@qq.com>

python/infinilm/llm/llm.py

@@ -320,12 +320,6 @@ class LLMEngine:
                 req.finish_reason = FinishReason.STOP_STRING
                 return True
 
-        # Check stop token IDs
-        stop_token_ids = req.sampling_params.stop_token_ids or []
-        if stop_token_ids and token_id in stop_token_ids:
-            req.finish_reason = FinishReason.STOP_STRING
-            return True
-
         return False
 
     def tokenize(self, text: str) -> List[int]:

python/infinilm/llm/sampling_params.py

@@ -15,7 +15,7 @@ class SamplingParams:
     top_k: int = 1
     max_tokens: Optional[int] = None
     stop: Optional[List[str]] = None
-    stop_token_ids: Optional[List[int]] = None
+    stop_token_ids: Optional[List[int]] = None  # Placeholder for future usage, not currently handled
 
     def __post_init__(self):
         if self.stop is None:

python/infinilm/server/inference_server.py

@@ -233,9 +233,6 @@ class InferenceServer:
         if isinstance(stop, str):
             stop = [stop]
 
-        stop_token_ids = pick("stop_token_ids", None)
-        if isinstance(stop_token_ids, int):
-            stop_token_ids = [stop_token_ids]
 
         return SamplingParams(
             temperature=float(pick("temperature", self.temperature)),
@@ -243,7 +240,6 @@ class InferenceServer:
             top_k=int(pick("top_k", self.top_k)),
             max_tokens=int(max_tokens) if max_tokens is not None else None,
             stop=stop,
-            stop_token_ids=stop_token_ids,
         )
 
     async def _stream_chat(self, request_id: str, data: dict, http_request: Request):