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 13, 17:00 UTC
[Bug] Fix cpu disable shared_experts `VLLM_DISABLE_SHARED_EXPERTS_STREAM` (#28157)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Wentao Ye committed
Nov 5, 2025 at 20:03 UTC
d79d9f078079e4108ec9d28b261a8e002e763338
1 file changed
+1
-1
vllm/platforms/cpu.py
+1
-1
@@ -299,7 +299,7 @@ class CpuPlatform(Platform):
299
os.environ["TORCHINDUCTOR_COMPILE_THREADS"] = "1"
300
301
# Disable multi-stream for shared experts as no Stream on CPU
302
- os.environ["VLLM_DISABLE_SHARED_EXPERTS_STREAM"] = "0"
302
+ os.environ["VLLM_DISABLE_SHARED_EXPERTS_STREAM"] = "1"
303
304
# Intel OpenMP setting
305
ld_preload_str = os.getenv("LD_PRELOAD", "")