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 10, 06:30 UTC
[Perf] Optimize requests abort (#32211)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Wentao Ye committed
Jan 12, 2026 at 23:11 UTC
2a719e0865d68ef930c53f9b46f718c31ed39377
1 file changed
+4
-3
vllm/v1/engine/async_llm.py
+4
-3
@@ -525,9 +525,10 @@ class AsyncLLM(EngineClient):
525
await asyncio.sleep(0)
526
527
# 3) Abort any reqs that finished due to stop strings.
528
- await engine_core.abort_requests_async(
529
- processed_outputs.reqs_to_abort
530
- )
528
+ if processed_outputs.reqs_to_abort:
529
+ await engine_core.abort_requests_async(
530
+ processed_outputs.reqs_to_abort
531
+ )
532
533
output_processor.update_scheduler_stats(outputs.scheduler_stats)
534