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
[Bugfix][Core] running queue index leakage exception (#26754)
Signed-off-by: CLFutureX <chenyongqyl@163.com>
PiteXChen committed
Oct 23, 2025 at 12:40 UTC
243ed7d32e94f00a9a32fbbc51be932f6277a55d
1 file changed
+1
vllm/v1/core/sched/scheduler.py
+1
@@ -278,6 +278,7 @@ class Scheduler(SchedulerInterface):
278
token_budget += num_scheduled_tokens[preempted_req.request_id]
279
req_to_new_blocks.pop(preempted_req.request_id)
280
num_scheduled_tokens.pop(preempted_req.request_id)
281
+ req_index -= 1
282
else:
283
preempted_req = self.running.pop()
284