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 16, 10:45 UTC
[CI/Build] setuptools-scm fixes (#8900)
Daniele committed
Oct 14, 2024 at 20:34 UTC
203ab8f80f780baf899a8bc4b5c38a9929fa88ca
7 files changed
+47
-41
.buildkite/release-pipeline.yaml
+2
-2
@@ -3,7 +3,7 @@ steps:
3
agents:
4
queue: cpu_queue
5
commands:
6
- - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg buildkite_commit=$BUILDKITE_COMMIT --build-arg USE_SCCACHE=1 --build-arg CUDA_VERSION=12.1.0 --tag vllm-ci:build-image --target build --progress plain ."
6
+ - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg CUDA_VERSION=12.1.0 --tag vllm-ci:build-image --target build --progress plain ."
7
- "mkdir artifacts"
8
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
9
# rename the files to change linux -> manylinux1
@@ -22,7 +22,7 @@ steps:
22
agents:
23
queue: cpu_queue
24
commands:
25
- - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg buildkite_commit=$BUILDKITE_COMMIT --build-arg USE_SCCACHE=1 --build-arg CUDA_VERSION=11.8.0 --tag vllm-ci:build-image --target build --progress plain ."
25
+ - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg CUDA_VERSION=11.8.0 --tag vllm-ci:build-image --target build --progress plain ."
26
- "mkdir artifacts"
27
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
28
# rename the files to change linux -> manylinux1
.dockerignore
+29
-1
@@ -2,5 +2,33 @@
2
/.venv
3
/build
4
dist
5
-Dockerfile*
5
vllm/*.so
6
+
7
+# Byte-compiled / optimized / DLL files
8
+__pycache__/
9
+*.py[cod]
10
+*$py.class
11
+
12
+.mypy_cache
13
+
14
+# Distribution / packaging
15
+.Python
16
+/build/
17
+cmake-build-*/
18
+CMakeUserPresets.json
19
+develop-eggs/
20
+/dist/
21
+downloads/
22
+eggs/
23
+.eggs/
24
+lib/
25
+lib64/
26
+parts/
27
+sdist/
28
+var/
29
+wheels/
30
+share/python-wheels/
31
+*.egg-info/
32
+.installed.cfg
33
+*.egg
34
+MANIFEST
.github/workflows/scripts/build.sh
+1
-2
@@ -8,8 +8,7 @@ PATH=${cuda_home}/bin:$PATH
8
LD_LIBRARY_PATH=${cuda_home}/lib64:$LD_LIBRARY_PATH
9
10
# Install requirements
11
-$python_executable -m pip install wheel packaging 'setuptools-scm>=8'
12
-$python_executable -m pip install -r requirements-cuda.txt
11
+$python_executable -m pip install -r requirements-build.txt -r requirements-cuda.txt
12
13
# Limit the number of parallel jobs to avoid OOM
14
export MAX_JOBS=1
Dockerfile
+1
-9
@@ -71,15 +71,7 @@ RUN --mount=type=cache,target=/root/.cache/pip \
71
python3 -m pip install -r requirements-build.txt
72
73
# files and directories related to build wheels
74
-COPY csrc csrc
75
-COPY setup.py setup.py
76
-COPY cmake cmake
77
-COPY CMakeLists.txt CMakeLists.txt
78
-COPY README.md README.md
79
-COPY requirements-common.txt requirements-common.txt
80
-COPY requirements-cuda.txt requirements-cuda.txt
81
-COPY pyproject.toml pyproject.toml
82
-COPY vllm vllm
74
+COPY . .
75
76
# max jobs used by Ninja to build extensions
77
ARG max_jobs=2
Dockerfile.openvino
+1
-10
@@ -9,16 +9,7 @@ RUN apt-get update -y && \
9
ffmpeg libsm6 libxext6 libgl1
10
WORKDIR /workspace
11
12
-# copy requirements
13
-COPY requirements-build.txt /workspace/vllm/
14
-COPY requirements-common.txt /workspace/vllm/
15
-COPY requirements-openvino.txt /workspace/vllm/
16
-
17
-COPY vllm/ /workspace/vllm/vllm
18
-COPY csrc/core /workspace/vllm/csrc/core
19
-COPY cmake/utils.cmake /workspace/vllm/cmake/
20
-COPY CMakeLists.txt /workspace/vllm/
21
-COPY setup.py /workspace/vllm/
12
+COPY . .
13
14
# install build requirements
15
RUN PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu" python3 -m pip install -r /workspace/vllm/requirements-build.txt
collect_env.py
+10
-17
@@ -267,23 +267,16 @@ def get_neuron_sdk_version(run_lambda):
267
268
269
def get_vllm_version():
270
- version = ""
271
- try:
272
- import vllm
273
- version = vllm.__version__
274
- except Exception:
275
- pass
276
- commit = ""
277
- try:
278
- import vllm
279
- commit = vllm.__commit__
280
- except Exception:
281
- pass
282
- if version != "" and commit != "":
283
- return f"{version}@{commit}"
284
- if version == "" and commit == "":
285
- return "N/A"
286
- return version or commit
270
+ from vllm import __version__, __version_tuple__
271
+
272
+ if __version__ == "dev":
273
+ return "N/A (dev)"
274
+
275
+ if len(__version_tuple__) == 4: # dev build
276
+ git_sha = __version_tuple__[-1][1:] # type: ignore
277
+ return f"{__version__} (git sha: {git_sha}"
278
+
279
+ return __version__
280
281
def summarize_vllm_build_flags():
282
# This could be a static method if the flags are constant, or dynamic if you need to check environment variables, etc.
pyproject.toml
+3
@@ -12,6 +12,9 @@ requires = [
12
]
13
build-backend = "setuptools.build_meta"
14
15
+[tool.setuptools_scm]
16
+# version_file = "vllm/_version.py" # currently handled by `setup.py:get_version()`
17
+
18
[tool.ruff]
19
# Allow lines to be as long as 80.
20
line-length = 80