yindo/flash-attention-prebuild-wheels-rocm
1
0
Fork 0
mirror of https://github.com/BillyOutlast/flash-attention-prebuild-wheels-rocm.git synced 2026-07-01 01:27:54 -04:00
Code Issues Packages Projects Releases 2 Wiki Activity

fix workflow

Browse Source
This commit is contained in:
Junya Morioka
2025-05-27 13:47:37 +09:00
parent 38511edfd2
commit cf873af1fd
1 changed files with 299 additions and 299 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/build.yml
+299 -299
View File
@@ -32,339 +32,339 @@ jobs:
# ######################################################### # #########################################################
# Build wheels with GitHub hosted runner # Build wheels with GitHub hosted runner
# ######################################################### # #########################################################
build_wheels: # build_wheels:
name: Build wheels and Upload # name: Build wheels and Upload
needs: create_releases # needs: create_releases
runs-on: ubuntu-22.04 # runs-on: ubuntu-22.04
env: # env:
DEBIAN_FRONTEND: noninteractive # DEBIAN_FRONTEND: noninteractive
TERM: xterm-256color # TERM: xterm-256color
timeout-minutes: 1000 # timeout-minutes: 1000
strategy: # strategy:
fail-fast: false # fail-fast: false
matrix: # matrix:
# flash-attn-version: ["2.4.3", "2.5.9", "2.6.3"] # # flash-attn-version: ["2.4.3", "2.5.9", "2.6.3"]
# python-version: ["3.10", "3.11", "3.12"] # # python-version: ["3.10", "3.11", "3.12"]
# torch-version: ["2.8.0.dev20250523"] # # torch-version: ["2.8.0.dev20250523"]
# # https://developer.nvidia.com/cuda-toolkit-archive # # # https://developer.nvidia.com/cuda-toolkit-archive
# cuda-version: ["12.8.1"] # # cuda-version: ["12.8.1"]
flash-attn-version: [] # flash-attn-version: []
python-version: [] # python-version: []
torch-version: [] # torch-version: []
# https://developer.nvidia.com/cuda-toolkit-archive # # https://developer.nvidia.com/cuda-toolkit-archive
cuda-version: [] # cuda-version: []
exclude: # exclude:
# torch < 2.2 does not support Python 3.12 # # torch < 2.2 does not support Python 3.12
- python-version: "3.12" # - python-version: "3.12"
torch-version: "2.0.1" # torch-version: "2.0.1"
- python-version: "3.12" # - python-version: "3.12"
torch-version: "2.1.2" # torch-version: "2.1.2"
# torch 2.0.1 does not support CUDA 12.x # # torch 2.0.1 does not support CUDA 12.x
- torch-version: "2.0.1" # - torch-version: "2.0.1"
cuda-version: "12.1.1" # cuda-version: "12.1.1"
- torch-version: "2.0.1" # - torch-version: "2.0.1"
cuda-version: "12.4.1" # cuda-version: "12.4.1"
- torch-version: "2.0.1" # - torch-version: "2.0.1"
cuda-version: "12.6.3" # cuda-version: "12.6.3"
- torch-version: "2.0.1" # - torch-version: "2.0.1"
cuda-version: "12.8.1" # cuda-version: "12.8.1"
# torch 2.7.0 does not support CUDA 12.4 # # torch 2.7.0 does not support CUDA 12.4
- torch-version: "2.7.0" # - torch-version: "2.7.0"
cuda-version: "12.4.1" # cuda-version: "12.4.1"
steps: # steps:
- uses: actions/checkout@v4 # - uses: actions/checkout@v4
- name: Maximize build space # - name: Maximize build space
run: | # run: |
df -h # df -h
echo "-----------------------------" # echo "-----------------------------"
sudo rm -rf /usr/share/dotnet # sudo rm -rf /usr/share/dotnet
sudo rm -rf /usr/local/lib/android # sudo rm -rf /usr/local/lib/android
sudo rm -rf /opt/ghc # sudo rm -rf /opt/ghc
sudo rm -rf /opt/hostedtoolcache/CodeQL # sudo rm -rf /opt/hostedtoolcache/CodeQL
df -h # df -h
- name: Set Swap Space # - name: Set Swap Space
uses: pierotofy/set-swap-space@master # uses: pierotofy/set-swap-space@master
with: # with:
swap-size-gb: 48 # swap-size-gb: 48
- uses: actions/setup-python@v5 # - uses: actions/setup-python@v5
with: # with:
python-version: ${{ matrix.python-version }} # python-version: ${{ matrix.python-version }}
- uses: Jimver/cuda-toolkit@master # - uses: Jimver/cuda-toolkit@master
with: # with:
cuda: ${{ matrix.cuda-version }} # cuda: ${{ matrix.cuda-version }}
linux-local-args: '["--toolkit"]' # linux-local-args: '["--toolkit"]'
method: "network" # method: "network"
- name: Set CUDA and PyTorch versions # - name: Set CUDA and PyTorch versions
run: | # run: |
echo "MATRIX_CUDA_VERSION=$(echo ${{ matrix.cuda-version }} | awk -F \. {'print $1 $2'})" >> $GITHUB_ENV # echo "MATRIX_CUDA_VERSION=$(echo ${{ matrix.cuda-version }} | awk -F \. {'print $1 $2'})" >> $GITHUB_ENV
echo "MATRIX_TORCH_VERSION=$(echo ${{ matrix.torch-version }} | awk -F \. {'print $1 "." $2'})" >> $GITHUB_ENV # echo "MATRIX_TORCH_VERSION=$(echo ${{ matrix.torch-version }} | awk -F \. {'print $1 "." $2'})" >> $GITHUB_ENV
- name: Install build dependencies # - name: Install build dependencies
run: | # run: |
sudo apt install -y ninja-build clang # sudo apt install -y ninja-build clang
pip install -U pip setuptools==75.8.0 wheel setuptools packaging psutil # pip install -U pip setuptools==75.8.0 wheel setuptools packaging psutil
- name: Install PyTorch ${{ matrix.torch-version }}+cu${{ matrix.cuda-version }} # - name: Install PyTorch ${{ matrix.torch-version }}+cu${{ matrix.cuda-version }}
run: | # run: |
export TORCH_CUDA_VERSION=$(python -c "from os import environ as env; \ # export TORCH_CUDA_VERSION=$(python -c "from os import environ as env; \
support_cuda_versions = { \ # support_cuda_versions = { \
'2.0': [117, 118], \ # '2.0': [117, 118], \
'2.1': [118, 121], \ # '2.1': [118, 121], \
'2.2': [118, 121], \ # '2.2': [118, 121], \
'2.3': [118, 121], \ # '2.3': [118, 121], \
'2.4': [118, 121, 124], \ # '2.4': [118, 121, 124], \
'2.5': [118, 121, 124], \ # '2.5': [118, 121, 124], \
'2.6': [118, 124, 126], \ # '2.6': [118, 124, 126], \
'2.7': [118, 126, 128], \ # '2.7': [118, 126, 128], \
'2.8': [128], \ # '2.8': [128], \
}; \ # }; \
target_cuda_versions = support_cuda_versions[env['MATRIX_TORCH_VERSION']]; \ # target_cuda_versions = support_cuda_versions[env['MATRIX_TORCH_VERSION']]; \
cuda_version = int(env['MATRIX_CUDA_VERSION']); \ # cuda_version = int(env['MATRIX_CUDA_VERSION']); \
closest_version = min(target_cuda_versions, key=lambda x: abs(x - cuda_version)); \ # closest_version = min(target_cuda_versions, key=lambda x: abs(x - cuda_version)); \
print(closest_version) \ # print(closest_version) \
") # ")
if [[ ${{ matrix.torch-version }} == *"dev"* ]]; then # if [[ ${{ matrix.torch-version }} == *"dev"* ]]; then
pip install --pre torch==${{ matrix.torch-version }} --index-url https://download.pytorch.org/whl/nightly/cu${TORCH_CUDA_VERSION} # pip install --pre torch==${{ matrix.torch-version }} --index-url https://download.pytorch.org/whl/nightly/cu${TORCH_CUDA_VERSION}
else # else
pip install --no-cache-dir torch==${{ matrix.torch-version }} --index-url https://download.pytorch.org/whl/cu${TORCH_CUDA_VERSION} # pip install --no-cache-dir torch==${{ matrix.torch-version }} --index-url https://download.pytorch.org/whl/cu${TORCH_CUDA_VERSION}
fi # fi
nvcc --version # nvcc --version
python -V # python -V
python -c "import torch; print('PyTorch:', torch.__version__)" # python -c "import torch; print('PyTorch:', torch.__version__)"
python -c "import torch; print('CUDA:', torch.version.cuda)" # python -c "import torch; print('CUDA:', torch.version.cuda)"
python -c "from torch.utils import cpp_extension; print(cpp_extension.CUDA_HOME)" # python -c "from torch.utils import cpp_extension; print(cpp_extension.CUDA_HOME)"
- name: Checkout flash-attn # - name: Checkout flash-attn
run: | # run: |
git clone https://github.com/Dao-AILab/flash-attention.git -b "v${{ matrix.flash-attn-version }}" # git clone https://github.com/Dao-AILab/flash-attention.git -b "v${{ matrix.flash-attn-version }}"
- name: Build wheels # - name: Build wheels
timeout-minutes: 800 # timeout-minutes: 800
run: | # run: |
export PATH=/usr/local/nvidia/bin:/usr/local/nvidia/lib64:$PATH # export PATH=/usr/local/nvidia/bin:/usr/local/nvidia/lib64:$PATH
export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:$LD_LIBRARY_PATH # export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:$LD_LIBRARY_PATH
export MAX_JOBS=2 # export MAX_JOBS=2
export NVCC_THREADS=2 # export NVCC_THREADS=2
export FLASH_ATTENTION_FORCE_BUILD=TRUE # export FLASH_ATTENTION_FORCE_BUILD=TRUE
cd flash-attention # cd flash-attention
python setup.py bdist_wheel --dist-dir=dist # python setup.py bdist_wheel --dist-dir=dist
base_wheel_name=$(basename $(ls dist/*.whl | head -n 1)) # base_wheel_name=$(basename $(ls dist/*.whl | head -n 1))
wheel_name=$(echo $base_wheel_name | sed "s/${{ matrix.flash-attn-version }}/${{ matrix.flash-attn-version }}+cu${{ env.MATRIX_CUDA_VERSION }}torch${{ env.MATRIX_TORCH_VERSION }}/") # wheel_name=$(echo $base_wheel_name | sed "s/${{ matrix.flash-attn-version }}/${{ matrix.flash-attn-version }}+cu${{ env.MATRIX_CUDA_VERSION }}torch${{ env.MATRIX_TORCH_VERSION }}/")
mv dist/$base_wheel_name dist/$wheel_name # mv dist/$base_wheel_name dist/$wheel_name
echo "wheel_name=$wheel_name" >> $GITHUB_ENV # echo "wheel_name=$wheel_name" >> $GITHUB_ENV
- name: Install Test # - name: Install Test
run: | # run: |
pip install flash-attention/dist/${{ env.wheel_name }} # pip install flash-attention/dist/${{ env.wheel_name }}
python -c "import flash_attn; print(flash_attn.__version__)" # python -c "import flash_attn; print(flash_attn.__version__)"
- name: Get the tag version # - name: Get the tag version
id: extract_branch # id: extract_branch
run: echo "branch=${GITHUB_REF#refs/tags/}" >> $GITHUB_OUTPUT # run: echo "branch=${GITHUB_REF#refs/tags/}" >> $GITHUB_OUTPUT
- name: Get Release with Tag # - name: Get Release with Tag
id: get_release # id: get_release
uses: joutvhu/get-release@v1 # uses: joutvhu/get-release@v1
env: # env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} # GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
with: # with:
tag_name: ${{ steps.extract_branch.outputs.branch }} # tag_name: ${{ steps.extract_branch.outputs.branch }}
- name: Upload Release Asset # - name: Upload Release Asset
uses: actions/upload-release-asset@v1 # uses: actions/upload-release-asset@v1
env: # env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} # GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
with: # with:
upload_url: ${{ steps.get_release.outputs.upload_url }} # upload_url: ${{ steps.get_release.outputs.upload_url }}
asset_path: flash-attention/dist/${{ env.wheel_name }} # asset_path: flash-attention/dist/${{ env.wheel_name }}
asset_name: ${{ env.wheel_name }} # asset_name: ${{ env.wheel_name }}
asset_content_type: application/* # asset_content_type: application/*
# ######################################################### # # #########################################################
# Build wheels with self-hosted runner # # Build wheels with self-hosted runner
# ######################################################### # # #########################################################
build_wheels_self_hosted: # build_wheels_self_hosted:
name: Build wheels and Upload # name: Build wheels and Upload
needs: create_releases # needs: create_releases
runs-on: self-hosted # runs-on: self-hosted
container: # container:
image: ubuntu:22.04 # image: ubuntu:22.04
defaults: # defaults:
run: # run:
shell: bash # shell: bash
env: # env:
DEBIAN_FRONTEND: noninteractive # DEBIAN_FRONTEND: noninteractive
TERM: xterm-256color # TERM: xterm-256color
timeout-minutes: 1000 # timeout-minutes: 1000
strategy: # strategy:
fail-fast: false # fail-fast: false
matrix: # matrix:
# flash-attn-version: ["2.7.4"] # # flash-attn-version: ["2.7.4"]
# python-version: ["3.10", "3.11", "3.12"] # # python-version: ["3.10", "3.11", "3.12"]
# torch-version: ["2.8.0.dev20250523"] # # torch-version: ["2.8.0.dev20250523"]
# https://developer.nvidia.com/cuda-toolkit-archive # # https://developer.nvidia.com/cuda-toolkit-archive
# cuda-version: ["12.8.1"] # # cuda-version: ["12.8.1"]
flash-attn-version: [] # flash-attn-version: []
python-version: [] # python-version: []
torch-version: [] # torch-version: []
cuda-version: [] # cuda-version: []
exclude: # exclude:
# torch < 2.2 does not support Python 3.12 # # torch < 2.2 does not support Python 3.12
- python-version: "3.12" # - python-version: "3.12"
torch-version: "2.0.1" # torch-version: "2.0.1"
- python-version: "3.12" # - python-version: "3.12"
torch-version: "2.1.2" # torch-version: "2.1.2"
# torch 2.0.1 does not support CUDA 12.x # # torch 2.0.1 does not support CUDA 12.x
- torch-version: "2.0.1" # - torch-version: "2.0.1"
cuda-version: "12.1.1" # cuda-version: "12.1.1"
- torch-version: "2.0.1" # - torch-version: "2.0.1"
cuda-version: "12.4.1" # cuda-version: "12.4.1"
- torch-version: "2.0.1" # - torch-version: "2.0.1"
cuda-version: "12.6.3" # cuda-version: "12.6.3"
- torch-version: "2.0.1" # - torch-version: "2.0.1"
cuda-version: "12.8.1" # cuda-version: "12.8.1"
# torch 2.7.0 does not support CUDA 12.4 # # torch 2.7.0 does not support CUDA 12.4
- torch-version: "2.7.0" # - torch-version: "2.7.0"
cuda-version: "12.4.1" # cuda-version: "12.4.1"
steps: # steps:
- name: Install tools # - name: Install tools
run: | # run: |
apt-get update && apt-get install -y --no-install-recommends \ # apt-get update && apt-get install -y --no-install-recommends \
curl \ # curl \
ca-certificates \ # ca-certificates \
sudo \ # sudo \
software-properties-common \ # software-properties-common \
wget \ # wget \
unzip \ # unzip \
zip \ # zip \
git \ # git \
build-essential \ # build-essential \
gcc \ # gcc \
g++ \ # g++ \
clang \ # clang \
ninja-build \ # ninja-build \
keyboard-configuration # keyboard-configuration
- uses: actions/checkout@v4 # - uses: actions/checkout@v4
- uses: actions/setup-python@v5 # - uses: actions/setup-python@v5
with: # with:
python-version: ${{ matrix.python-version }} # python-version: ${{ matrix.python-version }}
- uses: Jimver/cuda-toolkit@master # - uses: Jimver/cuda-toolkit@master
env: # env:
DEBIAN_FRONTEND: noninteractive # DEBIAN_FRONTEND: noninteractive
with: # with:
cuda: ${{ matrix.cuda-version }} # cuda: ${{ matrix.cuda-version }}
linux-local-args: '["--toolkit"]' # linux-local-args: '["--toolkit"]'
method: "network" # method: "network"
- name: Set CUDA and PyTorch versions # - name: Set CUDA and PyTorch versions
run: | # run: |
echo "MATRIX_CUDA_VERSION=$(echo ${{ matrix.cuda-version }} | awk -F \. {'print $1 $2'})" >> $GITHUB_ENV # echo "MATRIX_CUDA_VERSION=$(echo ${{ matrix.cuda-version }} | awk -F \. {'print $1 $2'})" >> $GITHUB_ENV
echo "MATRIX_TORCH_VERSION=$(echo ${{ matrix.torch-version }} | awk -F \. {'print $1 "." $2'})" >> $GITHUB_ENV # echo "MATRIX_TORCH_VERSION=$(echo ${{ matrix.torch-version }} | awk -F \. {'print $1 "." $2'})" >> $GITHUB_ENV
shell: bash # shell: bash
- name: Install build dependencies # - name: Install build dependencies
run: | # run: |
sudo apt install -y ninja-build clang # sudo apt install -y ninja-build clang
pip install -U pip setuptools==75.8.0 wheel setuptools packaging psutil # pip install -U pip setuptools==75.8.0 wheel setuptools packaging psutil
- name: Install PyTorch ${{ matrix.torch-version }}+cu${{ matrix.cuda-version }} # - name: Install PyTorch ${{ matrix.torch-version }}+cu${{ matrix.cuda-version }}
shell: bash # shell: bash
run: | # run: |
export TORCH_CUDA_VERSION=$(python -c "from os import environ as env; \ # export TORCH_CUDA_VERSION=$(python -c "from os import environ as env; \
support_cuda_versions = { \ # support_cuda_versions = { \
'2.0': [117, 118], \ # '2.0': [117, 118], \
'2.1': [118, 121], \ # '2.1': [118, 121], \
'2.2': [118, 121], \ # '2.2': [118, 121], \
'2.3': [118, 121], \ # '2.3': [118, 121], \
'2.4': [118, 121, 124], \ # '2.4': [118, 121, 124], \
'2.5': [118, 121, 124], \ # '2.5': [118, 121, 124], \
'2.6': [118, 124, 126], \ # '2.6': [118, 124, 126], \
'2.7': [118, 126, 128], \ # '2.7': [118, 126, 128], \
'2.8': [128], \ # '2.8': [128], \
}; \ # }; \
target_cuda_versions = support_cuda_versions[env['MATRIX_TORCH_VERSION']]; \ # target_cuda_versions = support_cuda_versions[env['MATRIX_TORCH_VERSION']]; \
cuda_version = int(env['MATRIX_CUDA_VERSION']); \ # cuda_version = int(env['MATRIX_CUDA_VERSION']); \
closest_version = min(target_cuda_versions, key=lambda x: abs(x - cuda_version)); \ # closest_version = min(target_cuda_versions, key=lambda x: abs(x - cuda_version)); \
print(closest_version) \ # print(closest_version) \
") # ")
if [[ ${{ matrix.torch-version }} == *"dev"* ]]; then # if [[ ${{ matrix.torch-version }} == *"dev"* ]]; then
pip install --pre --force-reinstall --no-cache-dir torch==${{ matrix.torch-version }} --index-url https://download.pytorch.org/whl/nightly/cu${TORCH_CUDA_VERSION} # pip install --pre --force-reinstall --no-cache-dir torch==${{ matrix.torch-version }} --index-url https://download.pytorch.org/whl/nightly/cu${TORCH_CUDA_VERSION}
else # else
pip install --force-reinstall --no-cache-dir torch==${{ matrix.torch-version }} --index-url https://download.pytorch.org/whl/cu${TORCH_CUDA_VERSION} # pip install --force-reinstall --no-cache-dir torch==${{ matrix.torch-version }} --index-url https://download.pytorch.org/whl/cu${TORCH_CUDA_VERSION}
fi # fi
nvcc --version # nvcc --version
python -V # python -V
python -c "import torch; print('PyTorch:', torch.__version__)" # python -c "import torch; print('PyTorch:', torch.__version__)"
python -c "import torch; print('CUDA:', torch.version.cuda)" # python -c "import torch; print('CUDA:', torch.version.cuda)"
python -c "from torch.utils import cpp_extension; print(cpp_extension.CUDA_HOME)" # python -c "from torch.utils import cpp_extension; print(cpp_extension.CUDA_HOME)"
- name: Checkout flash-attn # - name: Checkout flash-attn
run: | # run: |
git clone https://github.com/Dao-AILab/flash-attention.git -b "v${{ matrix.flash-attn-version }}" # git clone https://github.com/Dao-AILab/flash-attention.git -b "v${{ matrix.flash-attn-version }}"
- name: Build wheels # - name: Build wheels
timeout-minutes: 800 # timeout-minutes: 800
shell: bash # shell: bash
run: | # run: |
export PATH=/usr/local/nvidia/bin:/usr/local/nvidia/lib64:$PATH # export PATH=/usr/local/nvidia/bin:/usr/local/nvidia/lib64:$PATH
export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:$LD_LIBRARY_PATH # export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:$LD_LIBRARY_PATH
export MAX_JOBS=2 # export MAX_JOBS=2
export NVCC_THREADS=4 # export NVCC_THREADS=4
export FLASH_ATTENTION_FORCE_BUILD=TRUE # export FLASH_ATTENTION_FORCE_BUILD=TRUE
cd flash-attention # cd flash-attention
python setup.py bdist_wheel --dist-dir=dist # python setup.py bdist_wheel --dist-dir=dist
base_wheel_name=$(basename $(ls dist/*.whl | head -n 1)) # base_wheel_name=$(basename $(ls dist/*.whl | head -n 1))
wheel_name=$(echo $base_wheel_name | sed "s/${{ matrix.flash-attn-version }}/${{ matrix.flash-attn-version }}+cu${{ env.MATRIX_CUDA_VERSION }}torch${{ env.MATRIX_TORCH_VERSION }}/") # wheel_name=$(echo $base_wheel_name | sed "s/${{ matrix.flash-attn-version }}/${{ matrix.flash-attn-version }}+cu${{ env.MATRIX_CUDA_VERSION }}torch${{ env.MATRIX_TORCH_VERSION }}/")
mv dist/$base_wheel_name dist/$wheel_name # mv dist/$base_wheel_name dist/$wheel_name
echo "wheel_name=$wheel_name" >> $GITHUB_ENV # echo "wheel_name=$wheel_name" >> $GITHUB_ENV
- name: Install Test # - name: Install Test
run: | # run: |
pip install flash-attention/dist/${{ env.wheel_name }} # pip install flash-attention/dist/${{ env.wheel_name }}
python -c "import flash_attn; print(flash_attn.__version__)" # python -c "import flash_attn; print(flash_attn.__version__)"
- name: Get the tag version # - name: Get the tag version
id: extract_branch # id: extract_branch
shell: bash # shell: bash
run: echo "branch=${GITHUB_REF#refs/tags/}" >> $GITHUB_OUTPUT # run: echo "branch=${GITHUB_REF#refs/tags/}" >> $GITHUB_OUTPUT
- name: Get Release with Tag # - name: Get Release with Tag
id: get_release # id: get_release
uses: joutvhu/get-release@v1 # uses: joutvhu/get-release@v1
env: # env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} # GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
with: # with:
tag_name: ${{ steps.extract_branch.outputs.branch }} # tag_name: ${{ steps.extract_branch.outputs.branch }}
- name: Upload Release Asset # - name: Upload Release Asset
uses: actions/upload-release-asset@v1 # uses: actions/upload-release-asset@v1
env: # env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} # GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
with: # with:
upload_url: ${{ steps.get_release.outputs.upload_url }} # upload_url: ${{ steps.get_release.outputs.upload_url }}
asset_path: flash-attention/dist/${{ env.wheel_name }} # asset_path: flash-attention/dist/${{ env.wheel_name }}
asset_name: ${{ env.wheel_name }} # asset_name: ${{ env.wheel_name }}
asset_content_type: application/* # asset_content_type: application/*
- name: Clean up # - name: Clean up
if: always() # if: always()
run: | # run: |
sudo rm -rf /opt/hostedtoolcache/Python/${{ matrix.python-version }}* # sudo rm -rf /opt/hostedtoolcache/Python/${{ matrix.python-version }}*
build_windows_wheels: build_windows_wheels:
name: Build wheels and Upload name: Build wheels and Upload