mirror of
https://github.com/BillyOutlast/flash-attention-prebuild-wheels-rocm.git
synced 2026-07-01 01:37:53 -04:00
Update build.yml
This commit is contained in:
Junya Morioka
+90
-90
@@ -49,7 +49,7 @@ jobs:
|
||||
needs: create_releases
|
||||
runs-on: ubuntu-latest
|
||||
strategy:
|
||||
fails-fast: false
|
||||
fail-fast: false
|
||||
matrix:
|
||||
flash-attn-version: [2.4.3, 2.5.6, 2.5.9, 2.6.3]
|
||||
python-version: [3.10, 3.11, 3.12]
|
||||
@@ -67,101 +67,101 @@ jobs:
|
||||
- torch-version: 2.0.1
|
||||
cuda-version: 12.4.1
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- name: Maximize build space
|
||||
run: |
|
||||
df -h
|
||||
echo "-----------------------------"
|
||||
sudo rm -rf /usr/share/dotnet
|
||||
sudo rm -rf /usr/local/lib/android
|
||||
sudo rm -rf /opt/ghc
|
||||
sudo rm -rf /opt/hostedtoolcache/CodeQL
|
||||
df -h
|
||||
- name: Set Swap Space
|
||||
uses: pierotofy/set-swap-space@master
|
||||
with:
|
||||
swap-size-gb: 10
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
- name: Maximize build space
|
||||
run: |
|
||||
df -h
|
||||
echo "-----------------------------"
|
||||
sudo rm -rf /usr/share/dotnet
|
||||
sudo rm -rf /usr/local/lib/android
|
||||
sudo rm -rf /opt/ghc
|
||||
sudo rm -rf /opt/hostedtoolcache/CodeQL
|
||||
df -h
|
||||
- name: Set Swap Space
|
||||
uses: pierotofy/set-swap-space@master
|
||||
with:
|
||||
swap-size-gb: 10
|
||||
|
||||
- uses: actions/setup-python@v5
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
- uses: Jimver/cuda-toolkit@master
|
||||
with:
|
||||
cuda: ${{ matrix.cuda-version }}
|
||||
linux-local-args: '["--toolkit"]'
|
||||
method: "network"
|
||||
- run: sudo apt install -y ninja-build
|
||||
- uses: actions/setup-python@v5
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
- uses: Jimver/cuda-toolkit@master
|
||||
with:
|
||||
cuda: ${{ matrix.cuda-version }}
|
||||
linux-local-args: '["--toolkit"]'
|
||||
method: "network"
|
||||
- run: sudo apt install -y ninja-build
|
||||
|
||||
- name: Set CUDA and PyTorch versions
|
||||
run: |
|
||||
echo "MATRIX_CUDA_VERSION=$(echo ${{ matrix.cuda }} | awk -F \. {'print $1 $2'})" >> $GITHUB_ENV
|
||||
echo "MATRIX_TORCH_VERSION=$(echo ${{ matrix.torch }} | awk -F \. {'print $1 "." $2'})" >> $GITHUB_ENV
|
||||
- name: Set CUDA and PyTorch versions
|
||||
run: |
|
||||
echo "MATRIX_CUDA_VERSION=$(echo ${{ matrix.cuda }} | awk -F \. {'print $1 $2'})" >> $GITHUB_ENV
|
||||
echo "MATRIX_TORCH_VERSION=$(echo ${{ matrix.torch }} | awk -F \. {'print $1 "." $2'})" >> $GITHUB_ENV
|
||||
|
||||
- name: Install PyTorch ${{ matrix.torch-version }}+cu${{ matrix.cuda-version }}
|
||||
run: |
|
||||
pip install -U pip
|
||||
pip install wheel setuptools packaging
|
||||
export TORCH_CUDA_VERSION=$(python -c "from os import environ as env; \
|
||||
support_cuda_versions = { \
|
||||
'2.0': [117, 118], \
|
||||
'2.1': [118, 121], \
|
||||
'2.2': [118, 121], \
|
||||
'2.3': [118, 121], \
|
||||
'2.4': [118, 121, 124], \
|
||||
'2.5': [118, 121, 124], \
|
||||
}; \
|
||||
target_cuda_versions = support_cuda_versions[env['MATRIX_TORCH_VERSION']]; \
|
||||
cuda_version = int(env['MATRIX_CUDA_VERSION']); \
|
||||
closest_version = min(target_cuda_versions, key=lambda x: abs(x - cuda_version)); \
|
||||
print(closest_version) \
|
||||
")
|
||||
pip install --no-cache-dir torch==${{ matrix.torch }} --index-url https://download.pytorch.org/whl/cu${TORCH_CUDA_VERSION}
|
||||
nvcc --version
|
||||
python -V
|
||||
python -c import torch; print("'PyTorch:', torch.__version__)"
|
||||
python -c "import torch; print('CUDA:', torch.version.cuda)"
|
||||
python -c "from torch.utils import cpp_extension; print(cpp_extension.CUDA_HOME)")
|
||||
- name: Install PyTorch ${{ matrix.torch-version }}+cu${{ matrix.cuda-version }}
|
||||
run: |
|
||||
pip install -U pip
|
||||
pip install wheel setuptools packaging
|
||||
export TORCH_CUDA_VERSION=$(python -c "from os import environ as env; \
|
||||
support_cuda_versions = { \
|
||||
'2.0': [117, 118], \
|
||||
'2.1': [118, 121], \
|
||||
'2.2': [118, 121], \
|
||||
'2.3': [118, 121], \
|
||||
'2.4': [118, 121, 124], \
|
||||
'2.5': [118, 121, 124], \
|
||||
}; \
|
||||
target_cuda_versions = support_cuda_versions[env['MATRIX_TORCH_VERSION']]; \
|
||||
cuda_version = int(env['MATRIX_CUDA_VERSION']); \
|
||||
closest_version = min(target_cuda_versions, key=lambda x: abs(x - cuda_version)); \
|
||||
print(closest_version) \
|
||||
")
|
||||
pip install --no-cache-dir torch==${{ matrix.torch }} --index-url https://download.pytorch.org/whl/cu${TORCH_CUDA_VERSION}
|
||||
nvcc --version
|
||||
python -V
|
||||
python -c import torch; print("'PyTorch:', torch.__version__)"
|
||||
python -c "import torch; print('CUDA:', torch.version.cuda)"
|
||||
python -c "from torch.utils import cpp_extension; print(cpp_extension.CUDA_HOME)")
|
||||
|
||||
- name: Checkout flash-attn
|
||||
run: |
|
||||
git clone https://github.com/Dao-AILab/flash-attention.git
|
||||
cd flash-attention
|
||||
git checkout v${{ matrix.flash-attn-version }}
|
||||
- name: Checkout flash-attn
|
||||
run: |
|
||||
git clone https://github.com/Dao-AILab/flash-attention.git
|
||||
cd flash-attention
|
||||
git checkout v${{ matrix.flash-attn-version }}
|
||||
|
||||
- name: Build wheels
|
||||
run: |
|
||||
pip install setuptools==68.0.0 ninja packaging wheel
|
||||
export PATH=/usr/local/nvidia/bin:/usr/local/nvidia/lib64:$PATH
|
||||
export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:$LD_LIBRARY_PATH
|
||||
export MAX_JOBS=1
|
||||
FLASH_ATTENTION_FORCE_BUILD="TRUE" python setup.py bdist_wheel --dist-dir=dist
|
||||
wheel_name=$(basename $(ls dist/*.whl | head -n 1))
|
||||
echo "wheel_name=$wheel_name" >> $GITHUB_ENV
|
||||
- name: Build wheels
|
||||
run: |
|
||||
pip install setuptools==68.0.0 ninja packaging wheel
|
||||
export PATH=/usr/local/nvidia/bin:/usr/local/nvidia/lib64:$PATH
|
||||
export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:$LD_LIBRARY_PATH
|
||||
export MAX_JOBS=1
|
||||
FLASH_ATTENTION_FORCE_BUILD="TRUE" python setup.py bdist_wheel --dist-dir=dist
|
||||
wheel_name=$(basename $(ls dist/*.whl | head -n 1))
|
||||
echo "wheel_name=$wheel_name" >> $GITHUB_ENV
|
||||
|
||||
- name: Install Test
|
||||
run: |
|
||||
pip install flash_attention/dist/${{ env.wheel_name }}
|
||||
python -c "import flash_atten; print(flash_atten.__version__)"
|
||||
- name: Install Test
|
||||
run: |
|
||||
pip install flash_attention/dist/${{ env.wheel_name }}
|
||||
python -c "import flash_atten; print(flash_atten.__version__)"
|
||||
|
||||
- name: Get the tag version
|
||||
id: extract_branch
|
||||
run: echo ::set-output name=branch::${GITHUB_REF#refs/tags/}
|
||||
- name: Get the tag version
|
||||
id: extract_branch
|
||||
run: echo ::set-output name=branch::${GITHUB_REF#refs/tags/}
|
||||
|
||||
- name: Get Release with Tag
|
||||
id: get_release
|
||||
uses: joutvhu/get-release@v1
|
||||
env:
|
||||
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
||||
with:
|
||||
tag_name: ${{ steps.extract_branch.outputs.branch }}
|
||||
- name: Get Release with Tag
|
||||
id: get_release
|
||||
uses: joutvhu/get-release@v1
|
||||
env:
|
||||
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
||||
with:
|
||||
tag_name: ${{ steps.extract_branch.outputs.branch }}
|
||||
|
||||
- name: Upload Release Asset
|
||||
uses: actions/upload-release-asset@v1
|
||||
env:
|
||||
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
||||
with:
|
||||
upload_url: ${{ steps.get_release.outputs.upload_url }}
|
||||
asset_path: flash-attention/dist/${{ env.wheel_name }}
|
||||
asset_name: ${{ env.wheel_name }}
|
||||
asset_content_type: application/*
|
||||
- name: Upload Release Asset
|
||||
uses: actions/upload-release-asset@v1
|
||||
env:
|
||||
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
||||
with:
|
||||
upload_url: ${{ steps.get_release.outputs.upload_url }}
|
||||
asset_path: flash-attention/dist/${{ env.wheel_name }}
|
||||
asset_name: ${{ env.wheel_name }}
|
||||
asset_content_type: application/*
|
||||
|
||||
Reference in New Issue
Block a user