Add CUDA 12.9.1 sbsa nightly binaries (#155819)

https://github.com/pytorch/pytorch/issues/155196

Pull Request resolved: https://github.com/pytorch/pytorch/pull/155819
Approved by: https://github.com/atalman
This commit is contained in:
Ting Lu
2025-06-13 18:52:41 +00:00
committed by PyTorch MergeBot
parent ce44877961
commit 344731fb25
4 changed files with 76 additions and 79 deletions

View File

@ -3,9 +3,7 @@ set -eux -o pipefail
GPU_ARCH_VERSION=${GPU_ARCH_VERSION:-}
if [[ "$GPU_ARCH_VERSION" == *"12.6"* ]]; then
export TORCH_CUDA_ARCH_LIST="9.0"
elif [[ "$GPU_ARCH_VERSION" == *"12.8"* ]]; then
if [[ "$GPU_ARCH_VERSION" == *"12.9"* ]]; then
export TORCH_CUDA_ARCH_LIST="9.0;10.0;12.0"
fi

View File

@ -88,7 +88,6 @@ def package_cuda_wheel(wheel_path, desired_cuda) -> None:
"/usr/local/cuda/lib64/libcusparseLt.so.0",
"/usr/local/cuda/lib64/libcusolver.so.11",
"/usr/local/cuda/lib64/libcurand.so.10",
"/usr/local/cuda/lib64/libnvToolsExt.so.1",
"/usr/local/cuda/lib64/libnvJitLink.so.12",
"/usr/local/cuda/lib64/libnvrtc.so.12",
"/usr/local/cuda/lib64/libcudnn_adv.so.9",
@ -108,9 +107,9 @@ def package_cuda_wheel(wheel_path, desired_cuda) -> None:
"/usr/local/lib/libnvpl_blas_core.so.0",
]
if "128" in desired_cuda:
if "129" in desired_cuda:
libs_to_copy += [
"/usr/local/cuda/lib64/libnvrtc-builtins.so.12.8",
"/usr/local/cuda/lib64/libnvrtc-builtins.so.12.9",
"/usr/local/cuda/lib64/libcufile.so.0",
"/usr/local/cuda/lib64/libcufile_rdma.so.1",
]

View File

@ -38,7 +38,7 @@ CPU_AARCH64_ARCH = ["cpu-aarch64"]
CPU_S390X_ARCH = ["cpu-s390x"]
CUDA_AARCH64_ARCHES = ["12.8-aarch64"]
CUDA_AARCH64_ARCHES = ["12.9-aarch64"]
PYTORCH_EXTRA_INSTALL_REQUIREMENTS = {

View File

@ -115,7 +115,7 @@ jobs:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
manywheel-py3_9-cuda-aarch64-12_8-build:
manywheel-py3_9-cuda-aarch64-12_9-build:
if: ${{ github.repository_owner == 'pytorch' }}
uses: ./.github/workflows/_binary-build-linux.yml
needs: get-label-type
@ -124,41 +124,41 @@ jobs:
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.9"
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
runs_on: linux.arm64.m7g.4xlarge.ephemeral
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_9-cuda-aarch64-12_8
build_name: manywheel-py3_9-cuda-aarch64-12_9
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.9.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.4.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.10.19; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.5.82; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.10.65; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.14.1.1; platform_system == 'Linux' and platform_machine == 'x86_64'
timeout-minutes: 420
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_9-cuda-aarch64-12_8-upload: # Uploading
manywheel-py3_9-cuda-aarch64-12_9-upload: # Uploading
if: ${{ github.repository_owner == 'pytorch' }}
permissions:
id-token: write
contents: read
needs: manywheel-py3_9-cuda-aarch64-12_8-build
needs: manywheel-py3_9-cuda-aarch64-12_9-build
with:
PYTORCH_ROOT: /pytorch
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.9"
build_name: manywheel-py3_9-cuda-aarch64-12_8
build_name: manywheel-py3_9-cuda-aarch64-12_9
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
@ -231,7 +231,7 @@ jobs:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
manywheel-py3_10-cuda-aarch64-12_8-build:
manywheel-py3_10-cuda-aarch64-12_9-build:
if: ${{ github.repository_owner == 'pytorch' }}
uses: ./.github/workflows/_binary-build-linux.yml
needs: get-label-type
@ -240,41 +240,41 @@ jobs:
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.10"
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
runs_on: linux.arm64.m7g.4xlarge.ephemeral
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_10-cuda-aarch64-12_8
build_name: manywheel-py3_10-cuda-aarch64-12_9
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.9.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.4.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.10.19; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.5.82; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.10.65; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.14.1.1; platform_system == 'Linux' and platform_machine == 'x86_64'
timeout-minutes: 420
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_10-cuda-aarch64-12_8-upload: # Uploading
manywheel-py3_10-cuda-aarch64-12_9-upload: # Uploading
if: ${{ github.repository_owner == 'pytorch' }}
permissions:
id-token: write
contents: read
needs: manywheel-py3_10-cuda-aarch64-12_8-build
needs: manywheel-py3_10-cuda-aarch64-12_9-build
with:
PYTORCH_ROOT: /pytorch
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.10"
build_name: manywheel-py3_10-cuda-aarch64-12_8
build_name: manywheel-py3_10-cuda-aarch64-12_9
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
@ -347,7 +347,7 @@ jobs:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
manywheel-py3_11-cuda-aarch64-12_8-build:
manywheel-py3_11-cuda-aarch64-12_9-build:
if: ${{ github.repository_owner == 'pytorch' }}
uses: ./.github/workflows/_binary-build-linux.yml
needs: get-label-type
@ -356,41 +356,41 @@ jobs:
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.11"
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
runs_on: linux.arm64.m7g.4xlarge.ephemeral
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_11-cuda-aarch64-12_8
build_name: manywheel-py3_11-cuda-aarch64-12_9
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.9.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.4.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.10.19; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.5.82; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.10.65; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.14.1.1; platform_system == 'Linux' and platform_machine == 'x86_64'
timeout-minutes: 420
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_11-cuda-aarch64-12_8-upload: # Uploading
manywheel-py3_11-cuda-aarch64-12_9-upload: # Uploading
if: ${{ github.repository_owner == 'pytorch' }}
permissions:
id-token: write
contents: read
needs: manywheel-py3_11-cuda-aarch64-12_8-build
needs: manywheel-py3_11-cuda-aarch64-12_9-build
with:
PYTORCH_ROOT: /pytorch
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.11"
build_name: manywheel-py3_11-cuda-aarch64-12_8
build_name: manywheel-py3_11-cuda-aarch64-12_9
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
@ -463,7 +463,7 @@ jobs:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
manywheel-py3_12-cuda-aarch64-12_8-build:
manywheel-py3_12-cuda-aarch64-12_9-build:
if: ${{ github.repository_owner == 'pytorch' }}
uses: ./.github/workflows/_binary-build-linux.yml
needs: get-label-type
@ -472,41 +472,41 @@ jobs:
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.12"
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
runs_on: linux.arm64.m7g.4xlarge.ephemeral
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_12-cuda-aarch64-12_8
build_name: manywheel-py3_12-cuda-aarch64-12_9
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.9.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.4.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.10.19; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.5.82; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.10.65; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.14.1.1; platform_system == 'Linux' and platform_machine == 'x86_64'
timeout-minutes: 420
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_12-cuda-aarch64-12_8-upload: # Uploading
manywheel-py3_12-cuda-aarch64-12_9-upload: # Uploading
if: ${{ github.repository_owner == 'pytorch' }}
permissions:
id-token: write
contents: read
needs: manywheel-py3_12-cuda-aarch64-12_8-build
needs: manywheel-py3_12-cuda-aarch64-12_9-build
with:
PYTORCH_ROOT: /pytorch
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.12"
build_name: manywheel-py3_12-cuda-aarch64-12_8
build_name: manywheel-py3_12-cuda-aarch64-12_9
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
@ -579,7 +579,7 @@ jobs:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
manywheel-py3_13-cuda-aarch64-12_8-build:
manywheel-py3_13-cuda-aarch64-12_9-build:
if: ${{ github.repository_owner == 'pytorch' }}
uses: ./.github/workflows/_binary-build-linux.yml
needs: get-label-type
@ -588,41 +588,41 @@ jobs:
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.13"
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
runs_on: linux.arm64.m7g.4xlarge.ephemeral
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_13-cuda-aarch64-12_8
build_name: manywheel-py3_13-cuda-aarch64-12_9
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.9.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.4.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.10.19; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.5.82; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.10.65; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.14.1.1; platform_system == 'Linux' and platform_machine == 'x86_64'
timeout-minutes: 420
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_13-cuda-aarch64-12_8-upload: # Uploading
manywheel-py3_13-cuda-aarch64-12_9-upload: # Uploading
if: ${{ github.repository_owner == 'pytorch' }}
permissions:
id-token: write
contents: read
needs: manywheel-py3_13-cuda-aarch64-12_8-build
needs: manywheel-py3_13-cuda-aarch64-12_9-build
with:
PYTORCH_ROOT: /pytorch
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.13"
build_name: manywheel-py3_13-cuda-aarch64-12_8
build_name: manywheel-py3_13-cuda-aarch64-12_9
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
@ -695,7 +695,7 @@ jobs:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml
manywheel-py3_13t-cuda-aarch64-12_8-build:
manywheel-py3_13t-cuda-aarch64-12_9-build:
if: ${{ github.repository_owner == 'pytorch' }}
uses: ./.github/workflows/_binary-build-linux.yml
needs: get-label-type
@ -704,41 +704,41 @@ jobs:
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.13t"
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
runs_on: linux.arm64.m7g.4xlarge.ephemeral
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_13t-cuda-aarch64-12_8
build_name: manywheel-py3_13t-cuda-aarch64-12_9
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.10.2.21; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.9.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.4.1.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.10.19; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.5.82; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.10.65; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.9.79; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.9.86; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.14.1.1; platform_system == 'Linux' and platform_machine == 'x86_64'
timeout-minutes: 420
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_13t-cuda-aarch64-12_8-upload: # Uploading
manywheel-py3_13t-cuda-aarch64-12_9-upload: # Uploading
if: ${{ github.repository_owner == 'pytorch' }}
permissions:
id-token: write
contents: read
needs: manywheel-py3_13t-cuda-aarch64-12_8-build
needs: manywheel-py3_13t-cuda-aarch64-12_9-build
with:
PYTORCH_ROOT: /pytorch
PACKAGE_TYPE: manywheel
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: cu128
GPU_ARCH_VERSION: 12.8-aarch64
DESIRED_CUDA: cu129
GPU_ARCH_VERSION: 12.9-aarch64
GPU_ARCH_TYPE: cuda-aarch64
DOCKER_IMAGE: manylinuxaarch64-builder
DOCKER_IMAGE_TAG_PREFIX: cuda12.8
DOCKER_IMAGE_TAG_PREFIX: cuda12.9
use_split_build: False
DESIRED_PYTHON: "3.13t"
build_name: manywheel-py3_13t-cuda-aarch64-12_8
build_name: manywheel-py3_13t-cuda-aarch64-12_9
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
uses: ./.github/workflows/_binary-upload.yml