@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 7852e50e4d
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

[docs] Document --scheduler-cls base class requirement (extend AsyncScheduler, not Scheduler) (#43724)

Signed-off-by: Georgii Kliukovkin <kliukovkin@gmail.com> Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com> Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

Georgii Kliukovkin committed Jun 11, 2026 at 02:49 UTC 7852e50e4dc4f42a67e9ce8471b177282326145c
1 file changed +6 -5
vllm/config/scheduler.py
+6 -5
@@ -175,12 +175,13 @@ class SchedulerConfig:
175
176 return Scheduler
177
178 - # This warning can be removed once the Scheduler interface is
179 - # finalized and we can maintain support for scheduler classes that
180 - # implement it
178 + # The first half of this warning can be removed once the Scheduler interface is
179 + # finalized and we can maintain support for scheduler classes that implement it
180 logger.warning_once(
182 - "Using custom scheduler class %s. This scheduler interface is "
183 - "not public and compatibility may not be maintained.",
181 + "Using custom scheduler class %s. This scheduler interface is not public "
182 + "and compatibility may not be maintained. If you have subclassed Scheduler "
183 + "instead of AsyncScheduler, you will see degraded performance due to async "
184 + "scheduling being disabled.",
185 self.scheduler_cls, # type: ignore[arg-type]
186 )
187 if not isinstance(self.scheduler_cls, str):