A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Frontend] Improve the performance of `is_reasoning_end` (#25735)
[Frontend] Improve the performance of `is_reasoning_end` (#25735) Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
Chauncey committed
Oct 11, 2025 at 10:43 UTC
be067861c6ab4df53a45e332caf53c163330290d
1 file changed
+2
-1
vllm/reasoning/basic_parsers.py
+2
-1
@@ -59,7 +59,8 @@ class BaseThinkingReasoningParser(ReasoningParser):
59
)
60
61
def is_reasoning_end(self, input_ids: list[int]) -> bool:
62
- return self.end_token_id in input_ids
62
+ end_token_id = self.end_token_id
63
+ return any(input_id == end_token_id for input_id in reversed(input_ids))
64
65
def extract_content_ids(self, input_ids: list[int]) -> list[int]:
66
"""