xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-07-19 23:17:29 +08:00
Code Issues Packages Projects Releases Wiki Activity

Consolidate args

Browse Source 
Signed-off-by: mgoin <michael@neuralmagic.com>
This commit is contained in:
mgoin 2025-07-30 14:40:00 -04:00
parent 61d568a19d
commit 02f7780716
3 changed files with 5 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
.buildkite/scripts/build-upload-flashinfer-wheel.sh
View File

@ -3,7 +3,7 @@
set -ex set -ex
CUDA_VERSION="${1:-12.8.1}" CUDA_VERSION="${1:-12.8.1}"
FLASHINFER_VERSION="${FLASHINFER_VERSION:-v0.2.9rc2}" # FlashInfer version controlled in tools/flashinfer-build.sh
echo "Building FlashInfer wheel for CUDA ${CUDA_VERSION} using vLLM Dockerfile" echo "Building FlashInfer wheel for CUDA ${CUDA_VERSION} using vLLM Dockerfile"
@ -12,7 +12,6 @@ DOCKER_BUILDKIT=1 docker build \
--build-arg max_jobs=16 \ --build-arg max_jobs=16 \
--build-arg USE_SCCACHE=1 \ --build-arg USE_SCCACHE=1 \
--build-arg CUDA_VERSION="${CUDA_VERSION}" \ --build-arg CUDA_VERSION="${CUDA_VERSION}" \
--build-arg FLASHINFER_GIT_REF="${FLASHINFER_VERSION}" \
--tag flashinfer-wheel-builder:${CUDA_VERSION} \ --tag flashinfer-wheel-builder:${CUDA_VERSION} \
--target flashinfer-wheel-builder \ --target flashinfer-wheel-builder \
--progress plain \ --progress plain \

7
docker/Dockerfile
View File

@ -271,8 +271,6 @@ RUN if [ "$RUN_WHEEL_CHECK" = "true" ]; then \
#################### FLASHINFER WHEEL BUILD IMAGE #################### #################### FLASHINFER WHEEL BUILD IMAGE ####################
FROM base AS flashinfer-wheel-builder FROM base AS flashinfer-wheel-builder
ARG CUDA_VERSION ARG CUDA_VERSION
ARG FLASHINFER_GIT_REPO="https://github.com/flashinfer-ai/flashinfer.git"
ARG FLASHINFER_GIT_REF="v0.2.9rc2"
COPY tools/flashinfer-build.sh /tmp/flashinfer-build.sh COPY tools/flashinfer-build.sh /tmp/flashinfer-build.sh
RUN --mount=type=cache,target=/root/.cache/uv \ RUN --mount=type=cache,target=/root/.cache/uv \
@ -402,10 +400,7 @@ RUN --mount=type=bind,from=build,src=/workspace/dist,target=/vllm-workspace/dist
# $ # upload the wheel to a public location, e.g. https://wheels.vllm.ai/flashinfer/v0.2.6.post1/flashinfer_python-0.2.6.post1-cp39-abi3-linux_x86_64.whl # $ # upload the wheel to a public location, e.g. https://wheels.vllm.ai/flashinfer/v0.2.6.post1/flashinfer_python-0.2.6.post1-cp39-abi3-linux_x86_64.whl
# Install FlashInfer from source # Install FlashInfer from source
ARG FLASHINFER_GIT_REPO="https://github.com/flashinfer-ai/flashinfer.git" # Version controlled in tools/flashinfer-build.sh - keep in sync with requirements/cuda.txt
# Keep this in sync with https://github.com/vllm-project/vllm/blob/main/requirements/cuda.txt
# We use `--force-reinstall --no-deps` to avoid issues with the existing FlashInfer wheel.
ARG FLASHINFER_GIT_REF="v0.2.9rc2"
COPY tools/flashinfer-build.sh /tmp/flashinfer-build.sh COPY tools/flashinfer-build.sh /tmp/flashinfer-build.sh
RUN --mount=type=cache,target=/root/.cache/uv \ RUN --mount=type=cache,target=/root/.cache/uv \
. /etc/environment && \ . /etc/environment && \

5
tools/flashinfer-build.sh
View File

@ -5,8 +5,9 @@ set -ex
# Build FlashInfer with AOT kernels # Build FlashInfer with AOT kernels
# This script is used by both the Dockerfile and standalone wheel building # This script is used by both the Dockerfile and standalone wheel building
FLASHINFER_GIT_REPO="${FLASHINFER_GIT_REPO:-https://github.com/flashinfer-ai/flashinfer.git}" # FlashInfer configuration - keep FLASHINFER_GIT_REF in sync with requirements/cuda.txt
FLASHINFER_GIT_REF="${FLASHINFER_GIT_REF:-v0.2.9rc2}" FLASHINFER_GIT_REPO="https://github.com/flashinfer-ai/flashinfer.git"
FLASHINFER_GIT_REF="${FLASHINFER_GIT_REF:-v0.2.9rc2}" # Must match requirements/cuda.txt
CUDA_VERSION="${CUDA_VERSION:-12.8.1}" CUDA_VERSION="${CUDA_VERSION:-12.8.1}"
BUILD_WHEEL="${BUILD_WHEEL:-false}" BUILD_WHEEL="${BUILD_WHEEL:-false}"