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] remove .github from .dockerignore, add dirty repo check (#9375)
Daniele committed
Oct 17, 2024 at 19:25 UTC
a2c71c5405fdd8822956bcd785e72149c1cfb655
12 files changed
+54
-17
.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 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 GIT_REPO_CHECK=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 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 GIT_REPO_CHECK=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
-1
@@ -1,4 +1,3 @@
1
-/.github/
1
/.venv
2
/build
3
dist
.github/workflows/scripts/build.sh
+4
@@ -1,4 +1,5 @@
1
#!/bin/bash
2
+set -eux
3
4
python_executable=python$1
5
cuda_home=/usr/local/cuda-$2
@@ -15,5 +16,8 @@ export MAX_JOBS=1
16
# Make sure release wheels are built for the following architectures
17
export TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 8.9 9.0+PTX"
18
export VLLM_FA_CMAKE_GPU_ARCHES="80-real;90-real"
19
+
20
+bash tools/check_repo.sh
21
+
22
# Build
23
$python_executable setup.py bdist_wheel --dist-dir=dist
Dockerfile
+3
-1
@@ -70,8 +70,10 @@ COPY requirements-build.txt requirements-build.txt
70
RUN --mount=type=cache,target=/root/.cache/pip \
71
python3 -m pip install -r requirements-build.txt
72
73
-# files and directories related to build wheels
73
COPY . .
74
+ARG GIT_REPO_CHECK=0
75
+RUN --mount=type=bind,source=.git,target=.git \
76
+ if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh ; fi
77
78
# max jobs used by Ninja to build extensions
79
ARG max_jobs=2
Dockerfile.cpu
+4
-1
@@ -42,7 +42,10 @@ RUN --mount=type=cache,target=/root/.cache/pip \
42
--mount=type=bind,src=requirements-cpu.txt,target=requirements-cpu.txt \
43
pip install -v -r requirements-cpu.txt
44
45
-COPY ./ ./
45
+COPY . .
46
+ARG GIT_REPO_CHECK=0
47
+RUN --mount=type=bind,source=.git,target=.git \
48
+ if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh ; fi
49
50
# Support for building with non-AVX512 vLLM: docker build --build-arg VLLM_CPU_DISABLE_AVX512="true" ...
51
ARG VLLM_CPU_DISABLE_AVX512
Dockerfile.neuron
+7
-7
@@ -17,7 +17,7 @@ RUN apt-get update && \
17
# When launching the container, mount the code directory to /app
18
ARG APP_MOUNT=/app
19
VOLUME [ ${APP_MOUNT} ]
20
-WORKDIR ${APP_MOUNT}
20
+WORKDIR ${APP_MOUNT}/vllm
21
22
RUN python3 -m pip install --upgrade pip
23
RUN python3 -m pip install --no-cache-dir fastapi ninja tokenizers pandas
@@ -25,17 +25,17 @@ RUN python3 -m pip install sentencepiece transformers==4.36.2 -U
25
RUN python3 -m pip install transformers-neuronx --extra-index-url=https://pip.repos.neuron.amazonaws.com -U
26
RUN python3 -m pip install --pre neuronx-cc==2.15.* --extra-index-url=https://pip.repos.neuron.amazonaws.com -U
27
28
-COPY . /app/vllm
28
+COPY . .
29
+ARG GIT_REPO_CHECK=0
30
+RUN --mount=type=bind,source=.git,target=.git \
31
+ if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh ; fi
32
30
-RUN cd /app/vllm \
31
- && python3 -m pip install -U \
33
+RUN python3 -m pip install -U \
34
cmake>=3.26 ninja packaging setuptools-scm>=8 wheel jinja2 \
35
-r requirements-neuron.txt
36
37
ENV VLLM_TARGET_DEVICE neuron
38
RUN --mount=type=bind,source=.git,target=.git \
37
- cd /app/vllm \
38
- && pip install --no-build-isolation -v -e . \
39
- && cd ..
39
+ pip install --no-build-isolation -v -e . \
40
41
CMD ["/bin/bash"]
Dockerfile.openvino
+3
@@ -10,6 +10,9 @@ RUN apt-get update -y && \
10
WORKDIR /workspace
11
12
COPY . .
13
+ARG GIT_REPO_CHECK=0
14
+RUN --mount=type=bind,source=.git,target=.git \
15
+ if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh ; fi
16
17
# install build requirements
18
RUN PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu" python3 -m pip install -r /workspace/vllm/requirements-build.txt
Dockerfile.ppc64le
+3
@@ -14,6 +14,9 @@ RUN micromamba install -y -n base -c https://ftp.osuosl.org/pub/open-ce/1.11.0-p
14
COPY ./ /workspace/vllm
15
16
WORKDIR /workspace/vllm
17
+ARG GIT_REPO_CHECK=0
18
+RUN --mount=type=bind,source=.git,target=.git \
19
+ if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh; fi
20
21
# These packages will be in rocketce eventually
22
RUN --mount=type=cache,target=/root/.cache/pip \
Dockerfile.rocm
+3
@@ -117,6 +117,9 @@ RUN --mount=type=cache,target=${CCACHE_DIR} \
117
FROM base AS final
118
# Import the vLLM development directory from the build context
119
COPY . .
120
+ARG GIT_REPO_CHECK=0
121
+RUN --mount=type=bind,source=.git,target=.git \
122
+ if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh ; fi
123
124
# Package upgrades for useful functionality or to avoid dependency issues
125
RUN --mount=type=cache,target=/root/.cache/pip \
Dockerfile.tpu
+7
-4
@@ -2,7 +2,7 @@ ARG NIGHTLY_DATE="20240828"
2
ARG BASE_IMAGE="us-central1-docker.pkg.dev/tpu-pytorch-releases/docker/xla:nightly_3.10_tpuvm_$NIGHTLY_DATE"
3
4
FROM $BASE_IMAGE
5
-WORKDIR /workspace
5
+WORKDIR /workspace/vllm
6
7
# Install some basic utilities
8
RUN apt-get update && apt-get install -y \
@@ -16,14 +16,17 @@ RUN --mount=type=cache,target=/root/.cache/pip \
16
python3 -m pip install torch_xla[pallas] -f https://storage.googleapis.com/jax-releases/jax_nightly_releases.html -f https://storage.googleapis.com/jax-releases/jaxlib_nightly_releases.html
17
18
# Build vLLM.
19
-COPY . /workspace/vllm
19
+COPY . .
20
+ARG GIT_REPO_CHECK=0
21
+RUN --mount=type=bind,source=.git,target=.git \
22
+ if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh; fi
23
+
24
ENV VLLM_TARGET_DEVICE="tpu"
25
RUN --mount=type=cache,target=/root/.cache/pip \
26
--mount=type=bind,source=.git,target=.git \
23
- cd /workspace/vllm && \
27
python3 -m pip install \
28
cmake>=3.26 ninja packaging setuptools-scm>=8 wheel jinja2 \
29
-r requirements-tpu.txt
27
-RUN cd /workspace/vllm && python3 setup.py develop
30
+RUN python3 setup.py develop
31
32
CMD ["/bin/bash"]
Dockerfile.xpu
+4
-1
@@ -33,7 +33,10 @@ RUN --mount=type=cache,target=/root/.cache/pip \
33
--extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/ \
34
-r requirements-xpu.txt
35
36
-COPY ./ /workspace/vllm
36
+COPY . .
37
+ARG GIT_REPO_CHECK
38
+RUN --mount=type=bind,source=.git,target=.git \
39
+ if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh; fi
40
41
ENV VLLM_TARGET_DEVICE=xpu
42
tools/check_repo.sh
new
+14
@@ -0,0 +1,14 @@
1
+#!/bin/bash
2
+# Checks whether the repo is clean and whether tags are available (necessary to correctly produce vllm version at build time)
3
+
4
+if ! git diff --quiet; then
5
+ echo "Repo is dirty" >&2
6
+
7
+ exit 1
8
+fi
9
+
10
+if ! git describe --tags; then
11
+ echo "No tags are present. Is this a shallow clone? git fetch --unshallow --tags" >&2
12
+
13
+ exit 1
14
+fi