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 9, 20:45 UTC
[Docs] Remove outdated async_scheduling limitation with speculative decoding (#32775)
Signed-off-by: Ifta Khairul Alam Adil <ikaadil007@gmail.com> Signed-off-by: Ifta khairul Alam Adil <25082512+ikaadil@users.noreply.github.com>
Ifta khairul Alam Adil committed
Jan 22, 2026 at 05:19 UTC
a1d82466ea2b174768e20874d9eb1a689a0e675d
1 file changed
-3
vllm/config/scheduler.py
-3
@@ -133,9 +133,6 @@ class SchedulerConfig:
133
async_scheduling: bool = Field(default=None)
134
"""If set to False, disable async scheduling. Async scheduling helps to
135
avoid gaps in GPU utilization, leading to better latency and throughput.
136
- It is currently not supported with some features such as
137
- speculative decoding and pipeline parallelism, and will be automatically
138
- disabled in those cases.
136
"""
137
138
stream_interval: int = Field(default=1, ge=1)