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, 18:15 UTC
[ci] Cleanup & refactor Dockerfile to pass different Python versions and sccache bucket via build args (#7705)
Signed-off-by: kevin <kevin@anyscale.com>
Kevin H. Luu committed
Aug 22, 2024 at 13:10 UTC
666ad0aa16f0c656e48e58b4f31ffe956b484d3b
1 file changed
+26
-35
Dockerfile
+26
-35
@@ -9,28 +9,23 @@ ARG CUDA_VERSION=12.4.1
9
#################### BASE BUILD IMAGE ####################
10
# prepare basic build environment
11
FROM nvidia/cuda:${CUDA_VERSION}-devel-ubuntu20.04 AS base
12
-
12
ARG CUDA_VERSION=12.4.1
13
ARG PYTHON_VERSION=3.10
15
-
14
ENV DEBIAN_FRONTEND=noninteractive
15
16
+# Install Python and other dependencies
17
RUN echo 'tzdata tzdata/Areas select America' | debconf-set-selections \
18
&& echo 'tzdata tzdata/Zones/America select Los_Angeles' | debconf-set-selections \
19
&& apt-get update -y \
21
- && apt-get install -y ccache software-properties-common \
20
+ && apt-get install -y ccache software-properties-common git curl sudo \
21
&& add-apt-repository ppa:deadsnakes/ppa \
22
&& apt-get update -y \
23
&& apt-get install -y python${PYTHON_VERSION} python${PYTHON_VERSION}-dev python${PYTHON_VERSION}-venv \
25
- && if [ "${PYTHON_VERSION}" != "3" ]; then update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTHON_VERSION} 1; fi \
26
- && python3 --version
27
-
28
-RUN apt-get update -y \
29
- && apt-get install -y git curl sudo
30
-
31
-# Install pip s.t. it will be compatible with our PYTHON_VERSION
32
-RUN curl -sS https://bootstrap.pypa.io/get-pip.py | python${PYTHON_VERSION}
33
-RUN python3 -m pip --version
24
+ && update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTHON_VERSION} 1 \
25
+ && update-alternatives --set python3 /usr/bin/python${PYTHON_VERSION} \
26
+ && ln -sf /usr/bin/python${PYTHON_VERSION}-config /usr/bin/python3-config \
27
+ && curl -sS https://bootstrap.pypa.io/get-pip.py | python${PYTHON_VERSION} \
28
+ && python3 --version && python3 -m pip --version
29
30
# Workaround for https://github.com/openai/triton/issues/2507 and
31
# https://github.com/pytorch/pytorch/issues/107960 -- hopefully
@@ -62,17 +57,12 @@ ENV TORCH_CUDA_ARCH_LIST=${torch_cuda_arch_list}
57
#################### WHEEL BUILD IMAGE ####################
58
FROM base AS build
59
65
-ARG PYTHON_VERSION=3.10
66
-
60
# install build dependencies
61
COPY requirements-build.txt requirements-build.txt
62
63
RUN --mount=type=cache,target=/root/.cache/pip \
64
python3 -m pip install -r requirements-build.txt
65
73
-# install compiler cache to speed up compilation leveraging local or remote caching
74
-RUN apt-get update -y && apt-get install -y ccache
75
-
66
# files and directories related to build wheels
67
COPY csrc csrc
68
COPY setup.py setup.py
@@ -95,6 +85,8 @@ ARG buildkite_commit
85
ENV BUILDKITE_COMMIT=${buildkite_commit}
86
87
ARG USE_SCCACHE
88
+ARG SCCACHE_BUCKET_NAME=vllm-build-sccache
89
+ARG SCCACHE_REGION_NAME=us-west-2
90
# if USE_SCCACHE is set, use sccache to speed up compilation
91
RUN --mount=type=cache,target=/root/.cache/pip \
92
if [ "$USE_SCCACHE" = "1" ]; then \
@@ -103,12 +95,9 @@ RUN --mount=type=cache,target=/root/.cache/pip \
95
&& tar -xzf sccache.tar.gz \
96
&& sudo mv sccache-v0.8.1-x86_64-unknown-linux-musl/sccache /usr/bin/sccache \
97
&& rm -rf sccache.tar.gz sccache-v0.8.1-x86_64-unknown-linux-musl \
106
- && if [ "$CUDA_VERSION" = "11.8.0" ]; then \
107
- export SCCACHE_BUCKET=vllm-build-sccache-2; \
108
- else \
109
- export SCCACHE_BUCKET=vllm-build-sccache; \
110
- fi \
111
- && export SCCACHE_REGION=us-west-2 \
98
+ && export SCCACHE_BUCKET=${SCCACHE_BUCKET_NAME} \
99
+ && export SCCACHE_REGION=${SCCACHE_REGION_NAME} \
100
+ && export SCCACHE_IDLE_TIMEOUT=0 \
101
&& export CMAKE_BUILD_TYPE=Release \
102
&& sccache --show-stats \
103
&& python3 setup.py bdist_wheel --dist-dir=dist --py-limited-api=cp38 \
@@ -160,23 +149,24 @@ FROM nvidia/cuda:${CUDA_VERSION}-base-ubuntu20.04 AS vllm-base
149
ARG CUDA_VERSION=12.4.1
150
ARG PYTHON_VERSION=3.10
151
WORKDIR /vllm-workspace
152
+ENV DEBIAN_FRONTEND=noninteractive
153
+
154
+RUN PYTHON_VERSION_STR=$(echo ${PYTHON_VERSION} | sed 's/\.//g') && \
155
+ echo "export PYTHON_VERSION_STR=${PYTHON_VERSION_STR}" >> /etc/environment
156
157
+# Install Python and other dependencies
158
RUN echo 'tzdata tzdata/Areas select America' | debconf-set-selections \
159
&& echo 'tzdata tzdata/Zones/America select Los_Angeles' | debconf-set-selections \
160
&& apt-get update -y \
167
- && apt-get install -y ccache software-properties-common \
161
+ && apt-get install -y ccache software-properties-common git curl sudo vim python3-pip \
162
&& add-apt-repository ppa:deadsnakes/ppa \
163
&& apt-get update -y \
170
- && apt-get install -y python${PYTHON_VERSION} python${PYTHON_VERSION}-dev python${PYTHON_VERSION}-venv \
171
- && if [ "${PYTHON_VERSION}" != "3" ]; then update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTHON_VERSION} 1; fi \
172
- && python3 --version
173
-
174
-RUN apt-get update -y \
175
- && apt-get install -y python3-pip git vim curl libibverbs-dev
176
-
177
-# Install pip s.t. it will be compatible with our PYTHON_VERSION
178
-RUN curl -sS https://bootstrap.pypa.io/get-pip.py | python${PYTHON_VERSION}
179
-RUN python3 -m pip --version
164
+ && apt-get install -y python${PYTHON_VERSION} python${PYTHON_VERSION}-dev python${PYTHON_VERSION}-venv libibverbs-dev \
165
+ && update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTHON_VERSION} 1 \
166
+ && update-alternatives --set python3 /usr/bin/python${PYTHON_VERSION} \
167
+ && ln -sf /usr/bin/python${PYTHON_VERSION}-config /usr/bin/python3-config \
168
+ && curl -sS https://bootstrap.pypa.io/get-pip.py | python${PYTHON_VERSION} \
169
+ && python3 --version && python3 -m pip --version
170
171
# Workaround for https://github.com/openai/triton/issues/2507 and
172
# https://github.com/pytorch/pytorch/issues/107960 -- hopefully
@@ -194,7 +184,8 @@ RUN --mount=type=bind,from=mamba-builder,src=/usr/src/mamba,target=/usr/src/mamb
184
python3 -m pip install /usr/src/mamba/*.whl --no-cache-dir
185
186
RUN --mount=type=cache,target=/root/.cache/pip \
197
- python3 -m pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.1.4/flashinfer-0.1.4+cu121torch2.4-cp310-cp310-linux_x86_64.whl
187
+ . /etc/environment && \
188
+ python3 -m pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.1.4/flashinfer-0.1.4+cu121torch2.4-cp${PYTHON_VERSION_STR}-cp${PYTHON_VERSION_STR}-linux_x86_64.whl
189
#################### vLLM installation IMAGE ####################
190
191