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

Update PyTorch to 2.7.0 (#16859)

Browse Source
This commit is contained in:
Huy Do
2025-04-29 19:08:04 -07:00
committed by GitHub
parent 1c2bc7ead0
commit 2c4f59afc3
18 changed files with 102 additions and 77 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
docker/Dockerfile
View File

@@ -5,11 +5,11 @@
# docs/source/contributing/dockerfile/dockerfile.md and
# docs/source/assets/contributing/dockerfile-stages-dependency.png
ARG CUDA_VERSION=12.4.1
ARG CUDA_VERSION=12.8.1
#################### BASE BUILD IMAGE ####################
# prepare basic build environment
FROM nvidia/cuda:${CUDA_VERSION}-devel-ubuntu20.04 AS base
ARG CUDA_VERSION=12.4.1
ARG CUDA_VERSION=12.8.1
ARG PYTHON_VERSION=3.12
ARG TARGETPLATFORM
ENV DEBIAN_FRONTEND=noninteractive
@@ -37,6 +37,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
# Reference: https://github.com/astral-sh/uv/pull/1694
ENV UV_HTTP_TIMEOUT=500
ENV UV_INDEX_STRATEGY="unsafe-best-match"
# Upgrade to GCC 10 to avoid https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92519
# as it was causing spam when compiling the CUTLASS kernels
@@ -69,7 +70,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
COPY requirements/common.txt requirements/common.txt
COPY requirements/cuda.txt requirements/cuda.txt
RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install --system -r requirements/cuda.txt
uv pip install --system -r requirements/cuda.txt \
--extra-index-url https://download.pytorch.org/whl/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')
# cuda arch list used by torch
# can be useful for both `dev` and `test`
@@ -92,9 +94,11 @@ COPY requirements/build.txt requirements/build.txt
# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
# Reference: https://github.com/astral-sh/uv/pull/1694
ENV UV_HTTP_TIMEOUT=500
ENV UV_INDEX_STRATEGY="unsafe-best-match"
RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install --system -r requirements/build.txt
uv pip install --system -r requirements/build.txt \
--extra-index-url https://download.pytorch.org/whl/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')
COPY . .
ARG GIT_REPO_CHECK=0
@@ -161,22 +165,25 @@ FROM base as dev
# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
# Reference: https://github.com/astral-sh/uv/pull/1694
ENV UV_HTTP_TIMEOUT=500
ENV UV_INDEX_STRATEGY="unsafe-best-match"
# Workaround for #17068
RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install --system --no-build-isolation "git+https://github.com/state-spaces/mamba@v2.2.4"
COPY requirements/lint.txt requirements/lint.txt
COPY requirements/test.txt requirements/test.txt
COPY requirements/dev.txt requirements/dev.txt
# Workaround for #17068
RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install --system mamba-ssm==2.2.4 --no-build-isolation
RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install --system -r requirements/dev.txt
uv pip install --system -r requirements/dev.txt \
--extra-index-url https://download.pytorch.org/whl/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')
#################### DEV IMAGE ####################
#################### vLLM installation IMAGE ####################
# image with vLLM installed
# TODO: Restore to base image after FlashInfer AOT wheel fixed
FROM nvidia/cuda:${CUDA_VERSION}-devel-ubuntu22.04 AS vllm-base
ARG CUDA_VERSION=12.4.1
ARG CUDA_VERSION=12.8.1
ARG PYTHON_VERSION=3.12
WORKDIR /vllm-workspace
ENV DEBIAN_FRONTEND=noninteractive
@@ -209,6 +216,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
# Reference: https://github.com/astral-sh/uv/pull/1694
ENV UV_HTTP_TIMEOUT=500
ENV UV_INDEX_STRATEGY="unsafe-best-match"
# Workaround for https://github.com/openai/triton/issues/2507 and
# https://github.com/pytorch/pytorch/issues/107960 -- hopefully
@@ -229,7 +237,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
# Install vllm wheel first, so that torch etc will be installed.
RUN --mount=type=bind,from=build,src=/workspace/dist,target=/vllm-workspace/dist \
--mount=type=cache,target=/root/.cache/uv \
uv pip install --system dist/*.whl --verbose
uv pip install --system dist/*.whl --verbose \
--extra-index-url https://download.pytorch.org/whl/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')
# If we need to build FlashInfer wheel before its release:
# $ export FLASHINFER_ENABLE_AOT=1
@@ -246,19 +255,26 @@ RUN --mount=type=bind,from=build,src=/workspace/dist,target=/vllm-workspace/dist
RUN --mount=type=cache,target=/root/.cache/uv \
. /etc/environment && \
if [ "$TARGETPLATFORM" != "linux/arm64" ]; then \
uv pip install --system https://github.com/flashinfer-ai/flashinfer/releases/download/v0.2.1.post2/flashinfer_python-0.2.1.post2+cu124torch2.6-cp38-abi3-linux_x86_64.whl ; \
# TESTING: install FlashInfer from source to test 2.7.0 final RC
FLASHINFER_ENABLE_AOT=1 TORCH_CUDA_ARCH_LIST='7.5 8.0 8.6 8.9 9.0+PTX' \
uv pip install --system --no-build-isolation "git+https://github.com/flashinfer-ai/flashinfer@v0.2.2.post1" ; \
fi
COPY examples examples
COPY benchmarks benchmarks
COPY ./vllm/collect_env.py .
RUN --mount=type=cache,target=/root/.cache/uv \
. /etc/environment && \
uv pip list
# Although we build Flashinfer with AOT mode, there's still
# some issues w.r.t. JIT compilation. Therefore we need to
# install build dependencies for JIT compilation.
# TODO: Remove this once FlashInfer AOT wheel is fixed
COPY requirements/build.txt requirements/build.txt
RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install --system -r requirements/build.txt
uv pip install --system -r requirements/build.txt \
--extra-index-url https://download.pytorch.org/whl/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')
#################### vLLM installation IMAGE ####################
@@ -272,11 +288,13 @@ ADD . /vllm-workspace/
# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
# Reference: https://github.com/astral-sh/uv/pull/1694
ENV UV_HTTP_TIMEOUT=500
ENV UV_INDEX_STRATEGY="unsafe-best-match"
# install development dependencies (for testing)
# Workaround for #17068
RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install --system mamba-ssm==2.2.4 --no-build-isolation
uv pip install --system --no-build-isolation "git+https://github.com/state-spaces/mamba@v2.2.4"
# install development dependencies (for testing)
RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install --system -r requirements/dev.txt