[Bugfix] fixed inconsistent finish_reason handling between V0 and V1 engines (#27555)

Signed-off-by: chaunceyjiang <[email protected]>

Author: chaunceyjiang

vllm/v1/core/sched/utils.py

@@ -42,13 +42,6 @@ def remove_all(lst: list, items_to_remove: set) -> list:
 def check_stop(
     request: Request, max_model_len: int, pooler_output: torch.Tensor | None = None
 ) -> bool:
-    if (
-        request.num_tokens >= max_model_len
-        or request.num_output_tokens >= request.max_tokens
-    ):
-        request.status = RequestStatus.FINISHED_LENGTH_CAPPED
-        return True
-
     if request.pooling_params:
         if pooler_output is not None:
             request.status = RequestStatus.FINISHED_STOPPED
@@ -70,4 +63,10 @@ def check_stop(
         request.status = RequestStatus.FINISHED_STOPPED
         request.stop_reason = last_token_id
         return True
+    if (
+        request.num_tokens >= max_model_len
+        or request.num_output_tokens >= request.max_tokens
+    ):
+        request.status = RequestStatus.FINISHED_LENGTH_CAPPED
+        return True
     return False