diff --git a/.buildkite/image_build/image_build_arm64.sh b/.buildkite/image_build/image_build_arm64.sh index 10cb417ec4..c003f1f03d 100755 --- a/.buildkite/image_build/image_build_arm64.sh +++ b/.buildkite/image_build/image_build_arm64.sh @@ -26,6 +26,7 @@ else --platform linux/arm64 \ --build-arg max_jobs=16 \ --build-arg nvcc_threads=4 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda13.0 \ --build-arg torch_cuda_arch_list="9.0 10.0 11.0 12.0" \ --build-arg USE_SCCACHE=1 \ --build-arg buildkite_commit="$BUILDKITE_COMMIT" \ diff --git a/.buildkite/image_build/image_build_torch_nightly.sh b/.buildkite/image_build/image_build_torch_nightly.sh index d13b3351d3..ff60378102 100755 --- a/.buildkite/image_build/image_build_torch_nightly.sh +++ b/.buildkite/image_build/image_build_torch_nightly.sh @@ -48,7 +48,7 @@ echo "Image not found, proceeding with build..." # --- CUDA 13.0 for nightly builds --- # Nightly CI uses CUDA 13.0 while regular CI stays on CUDA 12.9 NIGHTLY_CUDA_VERSION="13.0.2" -NIGHTLY_BUILD_BASE_IMAGE="nvidia/cuda:${NIGHTLY_CUDA_VERSION}-devel-ubuntu22.04" +NIGHTLY_BUILD_BASE_IMAGE="pytorch/manylinux2_28-builder:cuda13.0" NIGHTLY_FINAL_BASE_IMAGE="nvidia/cuda:${NIGHTLY_CUDA_VERSION}-base-ubuntu22.04" echo "--- :docker: Building torch nightly image (CUDA ${NIGHTLY_CUDA_VERSION})" diff --git a/.buildkite/release-pipeline.yaml b/.buildkite/release-pipeline.yaml index fd33933e4c..bbf32cd24d 100644 --- a/.buildkite/release-pipeline.yaml +++ b/.buildkite/release-pipeline.yaml @@ -40,7 +40,7 @@ steps: agents: queue: arm64_cpu_queue_release commands: - - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg BUILD_OS=manylinux --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda13.0 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ." + - "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=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda13.0 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ." - "mkdir artifacts" - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'" - "bash .buildkite/scripts/upload-nightly-wheels.sh" @@ -54,7 +54,7 @@ steps: agents: queue: cpu_queue_release commands: - - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg BUILD_OS=manylinux --build-arg BUILD_BASE_IMAGE=pytorch/manylinux2_28-builder:cuda13.0 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ." + - "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=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg BUILD_BASE_IMAGE=pytorch/manylinux2_28-builder:cuda13.0 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ." - "mkdir artifacts" - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'" - "bash .buildkite/scripts/upload-nightly-wheels.sh" @@ -78,7 +78,7 @@ steps: agents: queue: arm64_cpu_queue_release commands: - - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64_CU129}\" --build-arg BUILD_OS=manylinux --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda12.9 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ." + - "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=12.9.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64_CU129}\" --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda12.9 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ." - "mkdir artifacts" - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'" - "bash .buildkite/scripts/upload-nightly-wheels.sh" @@ -149,7 +149,7 @@ steps: agents: queue: cpu_queue_release commands: - - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86_CU129}\" --build-arg BUILD_OS=manylinux --build-arg BUILD_BASE_IMAGE=pytorch/manylinux2_28-builder:cuda12.9 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ." + - "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=12.9.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86_CU129}\" --build-arg BUILD_BASE_IMAGE=pytorch/manylinux2_28-builder:cuda12.9 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ." - "mkdir artifacts" - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'" - "bash .buildkite/scripts/upload-nightly-wheels.sh" @@ -221,7 +221,7 @@ steps: --build-arg INSTALL_KV_CONNECTORS=true \ --build-arg MOONCAKE_WHEEL_AARCH64="${MOONCAKE_WHEEL_AARCH64_2_35}" \ --build-arg MOONCAKE_WHEEL_X86_64="${MOONCAKE_WHEEL_X86_64}" \ - --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinux2_28-builder:cuda13.0 \ --target vllm-openai \ --progress plain \ -f docker/Dockerfile . @@ -249,7 +249,7 @@ steps: --build-arg INSTALL_KV_CONNECTORS=true \ --build-arg MOONCAKE_WHEEL_AARCH64="${MOONCAKE_WHEEL_AARCH64_2_35}" \ --build-arg MOONCAKE_WHEEL_X86_64="${MOONCAKE_WHEEL_X86_64}" \ - --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda13.0 \ --target vllm-openai \ --progress plain \ -f docker/Dockerfile . @@ -270,6 +270,7 @@ steps: --build-arg USE_SCCACHE=1 \ --build-arg GIT_REPO_CHECK=1 \ --build-arg CUDA_VERSION=12.9.1 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinux2_28-builder:cuda12.9 \ --build-arg torch_cuda_arch_list="${CUDA_ARCH_X86_CU129}" \ --build-arg INSTALL_KV_CONNECTORS=true \ --build-arg MOONCAKE_WHEEL_AARCH64="${MOONCAKE_WHEEL_AARCH64_2_35}" \ @@ -297,6 +298,7 @@ steps: --build-arg USE_SCCACHE=1 \ --build-arg GIT_REPO_CHECK=1 \ --build-arg CUDA_VERSION=12.9.1 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda12.9 \ --build-arg torch_cuda_arch_list="${CUDA_ARCH_AARCH64_CU129}" \ --build-arg INSTALL_KV_CONNECTORS=true \ --build-arg MOONCAKE_WHEEL_AARCH64="${MOONCAKE_WHEEL_AARCH64_2_35}" \ @@ -327,7 +329,7 @@ steps: --build-arg INSTALL_KV_CONNECTORS=true \ --build-arg MOONCAKE_WHEEL_AARCH64="${MOONCAKE_WHEEL_AARCH64_2_39}" \ --build-arg MOONCAKE_WHEEL_X86_64="${MOONCAKE_WHEEL_X86_64}" \ - --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu24.04 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinux2_28-builder:cuda13.0 \ --target vllm-openai \ --progress plain \ -f docker/Dockerfile . @@ -356,7 +358,7 @@ steps: --build-arg INSTALL_KV_CONNECTORS=true \ --build-arg MOONCAKE_WHEEL_AARCH64="${MOONCAKE_WHEEL_AARCH64_2_39}" \ --build-arg MOONCAKE_WHEEL_X86_64="${MOONCAKE_WHEEL_X86_64}" \ - --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu24.04 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda13.0 \ --target vllm-openai \ --progress plain \ -f docker/Dockerfile . @@ -377,6 +379,7 @@ steps: --build-arg USE_SCCACHE=1 \ --build-arg GIT_REPO_CHECK=1 \ --build-arg CUDA_VERSION=12.9.1 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinux2_28-builder:cuda12.9 \ --build-arg UBUNTU_VERSION=24.04 \ --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 \ --build-arg torch_cuda_arch_list="${CUDA_ARCH_X86_CU129}" \ @@ -405,6 +408,7 @@ steps: --build-arg USE_SCCACHE=1 \ --build-arg GIT_REPO_CHECK=1 \ --build-arg CUDA_VERSION=12.9.1 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda12.9 \ --build-arg UBUNTU_VERSION=24.04 \ --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 \ --build-arg torch_cuda_arch_list="${CUDA_ARCH_AARCH64_CU129}" \ diff --git a/.buildkite/scripts/hardware_ci/run-gh200-test.sh b/.buildkite/scripts/hardware_ci/run-gh200-test.sh index c0d1081f05..ed82e91490 100644 --- a/.buildkite/scripts/hardware_ci/run-gh200-test.sh +++ b/.buildkite/scripts/hardware_ci/run-gh200-test.sh @@ -15,6 +15,7 @@ DOCKER_BUILDKIT=1 docker build . \ -t gh200-test \ --build-arg max_jobs=66 \ --build-arg nvcc_threads=2 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda13.0 \ --build-arg torch_cuda_arch_list="9.0+PTX" # Setup cleanup diff --git a/docker/Dockerfile b/docker/Dockerfile index 96ed0bddf1..4a284a2a27 100644 --- a/docker/Dockerfile +++ b/docker/Dockerfile @@ -35,23 +35,13 @@ ARG NCCL_VERSION=2.30.7 # private registries that use a different repository naming conventions. # # Example: -# docker build --build-arg BUILD_BASE_IMAGE=registry.acme.org/mirror/nvidia/cuda:${CUDA_VERSION}-devel-ubuntu20.04 +# docker build --build-arg BUILD_BASE_IMAGE=registry.acme.org/mirror/pytorch/manylinux2_28-builder:cuda13.0 -# Important: We build with an old version of Ubuntu to maintain broad -# compatibility with other Linux OSes. The main reason for this is that the -# glibc version is baked into the distro, and binaries built with one glibc -# version are not backwards compatible with OSes that use an earlier version. -ARG BUILD_BASE_IMAGE=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu22.04 +# Build wheels against the same glibc floor as PyTorch's published wheels. +ARG BUILD_BASE_IMAGE=pytorch/manylinux2_28-builder:cuda13.0 # Using cuda base image with minimal dependencies necessary for JIT compilation (FlashInfer, DeepGEMM, EP kernels) ARG FINAL_BASE_IMAGE=nvidia/cuda:${CUDA_VERSION}-base-ubuntu${UBUNTU_VERSION} -# OS family of BUILD_BASE_IMAGE. Controls package manager (apt vs dnf) and -# Python bootstrap. Set to "manylinux" alongside a manylinux build base such -# as pytorch/manylinux2_28-builder:cuda13.0 to produce wheels with a glibc -# 2.28 floor (matches PyTorch's own published wheels). Default stays on -# Ubuntu for backwards compatibility. -ARG BUILD_OS=ubuntu - # By parameterizing the Deadsnakes repository URL, we allow third-party to use # their own mirror. When doing so, we don't benefit from the transparent # installation of the GPG key of the PPA, as done by add-apt-repository, so we @@ -106,7 +96,6 @@ FROM ${BUILD_BASE_IMAGE} AS base ARG TARGETPLATFORM ARG CUDA_VERSION ARG PYTHON_VERSION -ARG BUILD_OS ARG USE_SCCACHE ARG SCCACHE_DOWNLOAD_URL ARG SCCACHE_ENDPOINT @@ -114,8 +103,6 @@ ARG SCCACHE_BUCKET_NAME=vllm-build-sccache ARG SCCACHE_REGION_NAME=us-west-2 ARG SCCACHE_S3_NO_CREDENTIALS=0 -ENV DEBIAN_FRONTEND=noninteractive - # Environment for uv # Declared BEFORE the installer + `uv venv` invocations below so the uv # binary, managed Python, download cache, and /opt/venv all land under @@ -133,45 +120,16 @@ ENV UV_INSTALL_DIR=/opt/uv/bin ENV PATH="/opt/venv/bin:/opt/uv/bin:$PATH" ENV VIRTUAL_ENV="/opt/venv" -# Install system dependencies including build tools. -# The Ubuntu path uses apt + deadsnakes-via-uv for Python; the manylinux path -# (AlmaLinux 8, e.g. pytorch/manylinux2_28-builder) uses dnf and the Python -# interpreters pre-installed at /opt/python/cpXY-cpXY/. -RUN if [ "${BUILD_OS}" = "manylinux" ]; then \ - # rdma-core-devel provides libibverbs headers; ccache lives in EPEL, - # which the pytorch manylinux image already enables. git/curl/sudo - # are typically pre-installed but listed defensively. - dnf install -y --setopt=install_weak_deps=False \ - ccache \ - git \ - curl \ - sudo \ - rdma-core-devel \ - && dnf clean all \ - && rm -rf /var/cache/dnf; \ - else \ - apt-get update -y \ - && apt-get install -y --no-install-recommends \ - ccache \ - software-properties-common \ - git \ - curl \ - sudo \ - python3-pip \ - libibverbs-dev \ - # GCC 10 was previously pinned to suppress spurious -Wredundant-move warnings - # from CUTLASS (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92519). That bug - # was fixed in GCC 11. GCC >= 11.3 is now required because PyTorch's C++20 headers - # (pytorch/pytorch#167929) are not compatible with GCC < 11.3. - gcc-11 \ - g++-11 \ - && update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-11 110 --slave /usr/bin/g++ g++ /usr/bin/g++-11 \ - # Install python dev headers if available (needed for cmake FindPython on Ubuntu 24.04 - # which ships cmake 3.28 and requires Development.SABIModule; silently skipped on - # Ubuntu 20.04/22.04 where python3.x-dev is not available without a PPA) - && (apt-get install -y --no-install-recommends python${PYTHON_VERSION}-dev 2>/dev/null || true) \ - && rm -rf /var/lib/apt/lists/*; \ - fi +# Install system dependencies including build tools. The PyTorch manylinux +# image enables EPEL, which provides ccache. +RUN dnf install -y --setopt=install_weak_deps=False \ + ccache \ + git \ + curl \ + sudo \ + rdma-core-devel \ + && dnf clean all \ + && rm -rf /var/cache/dnf # Install sccache once in base so Rust and CMake/CUDA build stages share the # same binary and remote cache configuration. @@ -194,8 +152,8 @@ ENV SCCACHE_REGION=${USE_SCCACHE:+${SCCACHE_REGION_NAME}} ENV SCCACHE_S3_NO_CREDENTIALS=${USE_SCCACHE:+${SCCACHE_S3_NO_CREDENTIALS}} ENV SCCACHE_IDLE_TIMEOUT=${USE_SCCACHE:+0} -# Install uv and bootstrap /opt/venv. Both paths converge on /opt/venv so all -# downstream stages stay distro-agnostic. +# Install uv and bootstrap /opt/venv from the Python interpreter included in +# the manylinux image. RUN mkdir -p "${UV_PYTHON_INSTALL_DIR}" "${UV_CACHE_DIR}" "${UV_INSTALL_DIR}" \ && chmod -R a+rX /opt/uv \ && curl -LsSf https://astral.sh/uv/install.sh | sh \ @@ -203,15 +161,9 @@ RUN mkdir -p "${UV_PYTHON_INSTALL_DIR}" "${UV_CACHE_DIR}" "${UV_INSTALL_DIR}" \ # pip` works regardless of how uv happens to link the venv back to the # managed Python install (which, at a non-default UV_PYTHON_INSTALL_DIR, # doesn't always expose ensurepip via the default venv layout). - && if [ "${BUILD_OS}" = "manylinux" ]; then \ - # manylinux images ship Python at /opt/python/cpXY-cpXY/; point uv - # at the matching interpreter rather than letting it fetch one. - PYV_NODOT=$(echo ${PYTHON_VERSION} | tr -d '.') \ - && MANYLINUX_PY=/opt/python/cp${PYV_NODOT}-cp${PYV_NODOT}/bin/python${PYTHON_VERSION} \ - && uv venv --seed /opt/venv --python "$MANYLINUX_PY"; \ - else \ - uv venv --seed /opt/venv --python ${PYTHON_VERSION}; \ - fi \ + && PYV_NODOT=$(echo ${PYTHON_VERSION} | tr -d '.') \ + && MANYLINUX_PY=/opt/python/cp${PYV_NODOT}-cp${PYV_NODOT}/bin/python${PYTHON_VERSION} \ + && uv venv --seed /opt/venv --python "$MANYLINUX_PY" \ && rm -f /usr/bin/python3 /usr/bin/python3-config /usr/bin/pip \ && ln -sf /opt/venv/bin/python3 /usr/bin/python3 \ && ln -sf /opt/venv/bin/python3-config /usr/bin/python3-config \ @@ -295,21 +247,13 @@ ENV TORCH_CUDA_ARCH_LIST=${torch_cuda_arch_list} # This stage reuses the Python environment from base and runs in parallel with # csrc-build/extensions-build. FROM base AS rust-build -ARG BUILD_OS ARG USE_SCCACHE ARG SCCACHE_ENDPOINT # Install native tools needed only for Rust/protoc builds. -RUN if [ "${BUILD_OS}" = "manylinux" ]; then \ - dnf install -y --setopt=install_weak_deps=False \ - make unzip \ - && dnf clean all && rm -rf /var/cache/dnf; \ - else \ - apt-get update -y \ - && apt-get install -y --no-install-recommends \ - make unzip \ - && rm -rf /var/lib/apt/lists/*; \ - fi +RUN dnf install -y --setopt=install_weak_deps=False make unzip \ + && dnf clean all \ + && rm -rf /var/cache/dnf COPY tools/install_protoc.sh /tmp/install_protoc.sh RUN /tmp/install_protoc.sh && rm /tmp/install_protoc.sh @@ -606,7 +550,6 @@ FROM base AS dev ARG PIP_INDEX_URL UV_INDEX_URL ARG PIP_EXTRA_INDEX_URL UV_EXTRA_INDEX_URL ARG PYTORCH_CUDA_INDEX_BASE_URL -ARG BUILD_OS # This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out # Reference: https://github.com/astral-sh/uv/pull/1694 @@ -616,11 +559,9 @@ ENV UV_INDEX_STRATEGY="unsafe-best-match" ENV UV_LINK_MODE=copy # Install libnuma-dev, required by fastsafetensors (fixes #20384) -RUN if [ "${BUILD_OS}" = "manylinux" ]; then \ - dnf install -y numactl-devel && dnf clean all && rm -rf /var/cache/dnf; \ - else \ - apt-get update && apt-get install -y --no-install-recommends libnuma-dev && rm -rf /var/lib/apt/lists/*; \ - fi +RUN dnf install -y numactl-devel \ + && dnf clean all \ + && rm -rf /var/cache/dnf # We can specify the standard or nightly build of PyTorch diff --git a/docker/versions.json b/docker/versions.json index 783d80aace..d681acfed9 100644 --- a/docker/versions.json +++ b/docker/versions.json @@ -14,14 +14,11 @@ "default": "2.30.7" }, "BUILD_BASE_IMAGE": { - "default": "nvidia/cuda:13.0.3-devel-ubuntu22.04" + "default": "pytorch/manylinux2_28-builder:cuda13.0" }, "FINAL_BASE_IMAGE": { "default": "nvidia/cuda:13.0.3-base-ubuntu22.04" }, - "BUILD_OS": { - "default": "ubuntu" - }, "GET_PIP_URL": { "default": "https://bootstrap.pypa.io/get-pip.py" }, diff --git a/docs/assets/contributing/dockerfile-stages-dependency.png b/docs/assets/contributing/dockerfile-stages-dependency.png index 7d069d0ed4..46da6142a7 100644 Binary files a/docs/assets/contributing/dockerfile-stages-dependency.png and b/docs/assets/contributing/dockerfile-stages-dependency.png differ diff --git a/docs/getting_started/installation/gpu.cuda.inc.md b/docs/getting_started/installation/gpu.cuda.inc.md index 7755ad6278..925d9605ab 100644 --- a/docs/getting_started/installation/gpu.cuda.inc.md +++ b/docs/getting_started/installation/gpu.cuda.inc.md @@ -384,6 +384,7 @@ A docker container can be built for aarch64 systems such as the Nvidia Grace-Hop -t vllm/vllm-gh200-openai:latest \ --build-arg max_jobs=66 \ --build-arg nvcc_threads=2 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda13.0 \ --build-arg torch_cuda_arch_list="9.0 10.0+PTX" ``` @@ -394,7 +395,7 @@ For (G)B300, we recommend using CUDA 13, as shown in the following command. ```bash DOCKER_BUILDKIT=1 docker build \ --build-arg CUDA_VERSION=13.0.2 \ - --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 \ + --build-arg BUILD_BASE_IMAGE=pytorch/manylinuxaarch64-builder:cuda13.0 \ --build-arg max_jobs=256 \ --build-arg nvcc_threads=2 \ --build-arg torch_cuda_arch_list='9.0 10.0+PTX' \