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 15, 16:00 UTC
[BugFix] make sure socket close (#15875)
Signed-off-by: yihong0618 <zouzou0208@gmail.com>
yihong committed
Apr 2, 2025 at 04:10 UTC
93491aefc7b98770e0a361859eb15660138a3abb
1 file changed
+4
-3
vllm/entrypoints/openai/api_server.py
+4
-3
@@ -1098,9 +1098,10 @@ async def run_server(args, **uvicorn_kwargs) -> None:
1098
)
1099
1100
# NB: Await server shutdown only after the backend context is exited
1101
- await shutdown_task
1102
-
1103
- sock.close()
1101
+ try:
1102
+ await shutdown_task
1103
+ finally:
1104
+ sock.close()
1105
1106
1107
if __name__ == "__main__":