Skip to content

GitLab

Unverified Commit 3322e264 authored by Li, Jiang's avatar Li, Jiang Committed by GitHub
Browse files

[Bugfix] Avoid more OpenMP thread reallocation in CPU torch compile (#37538) 


Signed-off-by: default avatarjiang1.li <jiang1.li@intel.com>
parent 765e4610
Show whitespace changes
Inline Side-by-side
Showing with 3 additions and 0 deletions
vllm/platforms/cpu.py
View file @ 3322e264
......@@ -281,6 +281,9 @@ class CpuPlatform(Platform):
# Disable multi-stream for shared experts as no Stream on CPU
os.environ["VLLM_DISABLE_SHARED_EXPERTS_STREAM"] = "1"
# Avoid inductor generates num_thread() and breaks the thread binding
os.environ["TORCHINDUCTOR_CPP_DYNAMIC_THREADS"] = "1"
# Intel OpenMP setting
ld_preload_str = os.getenv("LD_PRELOAD", "")
if "libiomp5.so" in ld_preload_str:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment