20231088/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[CI/Build] respect the common environment variable MAX_JOBS (#3600)

Browse Source
This commit is contained in:
youkaichao 2024-03-24 17:04:00 -07:00 committed by GitHub
parent 8b268a46a7
commit 42bc386129
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 22 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
docs/source/getting_started/installation.rst
View File

@ -60,6 +60,15 @@ You can also build and install vLLM from source:
$ cd vllm $ cd vllm
$ pip install -e . # This may take 5-10 minutes. $ pip install -e . # This may take 5-10 minutes.
.. tip::
To avoid your system being overloaded, you can limit the number of compilation jobs
to be run simultaneously, via the environment variable `MAX_JOBS`. For example:
.. code-block:: console
$ export MAX_JOBS=6
$ pip install -e .
.. tip:: .. tip::
If you have trouble building vLLM, we recommend using the NVIDIA PyTorch Docker image. If you have trouble building vLLM, we recommend using the NVIDIA PyTorch Docker image.

19
setup.py
View File

@ -1,6 +1,7 @@
import io import io
import os import os
import re import re
import logging
import subprocess import subprocess
import sys import sys
from typing import List from typing import List
@ -13,6 +14,7 @@ import torch
from torch.utils.cpp_extension import CUDA_HOME from torch.utils.cpp_extension import CUDA_HOME
ROOT_DIR = os.path.dirname(__file__) ROOT_DIR = os.path.dirname(__file__)
logger = logging.getLogger(__name__)
# vLLM only supports Linux platform # vLLM only supports Linux platform
assert sys.platform.startswith( assert sys.platform.startswith(
@ -54,12 +56,17 @@ class cmake_build_ext(build_ext):
# Determine number of compilation jobs and optionally nvcc compile threads. # Determine number of compilation jobs and optionally nvcc compile threads.
# #
def compute_num_jobs(self): def compute_num_jobs(self):
try: num_jobs = os.environ.get("MAX_JOBS", None)
# os.sched_getaffinity() isn't universally available, so fall back if num_jobs is not None:
# to os.cpu_count() if we get an error here. num_jobs = int(num_jobs)
num_jobs = len(os.sched_getaffinity(0)) logger.info(f"Using MAX_JOBS={num_jobs} as the number of jobs.")
except AttributeError: else:
num_jobs = os.cpu_count() try:
# os.sched_getaffinity() isn't universally available, so fall
# back to os.cpu_count() if we get an error here.
num_jobs = len(os.sched_getaffinity(0))
except AttributeError:
num_jobs = os.cpu_count()
nvcc_threads = None nvcc_threads = None
if _is_cuda(): if _is_cuda():