xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-02 20:57:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix][CPU] Disable dual stream execution for experts on CPU (#27320)

Browse Source 
Signed-off-by: jiang1.li <jiang1.li@intel.com>
This commit is contained in:
Li, Jiang 2025-10-22 19:02:27 +08:00 committed by GitHub
parent 1f633b8632
commit 843af7f7fc
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/platforms/cpu.py
View File

@ -297,6 +297,9 @@ class CpuPlatform(Platform):
# Disable torch async compiling which won't work with daemonic processes # Disable torch async compiling which won't work with daemonic processes
os.environ["TORCHINDUCTOR_COMPILE_THREADS"] = "1" os.environ["TORCHINDUCTOR_COMPILE_THREADS"] = "1"
# Disable multi-stream for shared experts as no Stream on CPU
os.environ["VLLM_DISABLE_SHARED_EXPERTS_STREAM"] = "0"
# Intel OpenMP setting # Intel OpenMP setting
ld_prealod_str = os.getenv("LD_PRELOAD", "") ld_prealod_str = os.getenv("LD_PRELOAD", "")
if "libiomp5.so" in ld_prealod_str: if "libiomp5.so" in ld_prealod_str: