biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Feature] NUMA binding support for GPU workers (#38635)

Browse Source 
Signed-off-by: Shengqi Chen <harry-chen@outlook.com>
Co-authored-by: Jason Li <jasonlizhengjian@gmail.com>
Co-authored-by: Roger Wang <hey@rogerw.io>
This commit is contained in:
Shengqi Chen
2026-04-09 00:55:24 +08:00
committed by GitHub
parent 512c5eb455
commit 75e01a39a1
13 changed files with 817 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
tests/utils_/test_system_utils.py
View File

@@ -1,10 +1,11 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import os
import tempfile
from pathlib import Path
from vllm.utils.system_utils import unique_filepath
from vllm.utils.system_utils import _maybe_force_spawn, unique_filepath
def test_unique_filepath():
@@ -17,3 +18,10 @@ def test_unique_filepath():
paths.add(path)
assert len(paths) == 10
assert len(list(Path(temp_dir).glob("*.txt"))) == 10
def test_numa_bind_forces_spawn(monkeypatch):
monkeypatch.delenv("VLLM_WORKER_MULTIPROC_METHOD", raising=False)
monkeypatch.setattr("sys.argv", ["vllm", "serve", "--numa-bind"])
_maybe_force_spawn()
assert os.environ["VLLM_WORKER_MULTIPROC_METHOD"] == "spawn"