lienzzzz/xformers-windows-blackwell2.0-cuda128-RTX50X0

Build environment:

System info:

• windows 10
• RTX 5070Ti
• cuda 12.8.1_572.61
• cudnn 9.8
• Python 3.12.9
• Nvidia Drivers: Studio 572.83

Python packages:

• nvidia-cublas-cu12 12.8.4.1
• nvidia-cuda-nvrtc-cu12 12.8.93
• nvidia-cuda-runtime-cu12 12.8.90
• nvidia-cudnn-cu12 9.8.0.87
• nvidia-cufft-cu12 11.3.3.83
• nvidia-curand-cu12 10.3.9.90
• nvidia-ml-py 12.570.86
• nvidia-nvjitlink-cu12 12.8.93
• nvidia-nvtx-cu12 12.8.90
• torch 2.8.0.dev20250415+cu128
• torchvision 0.22.0.dev20250416+cu128
• triton-windows 3.3.0a0.post17

Build Step:

pip install -U --pre triton-windows

git clone https://github.com/facebookresearch/xformers.git
cd .\xformers\
pip install -r requirements.txt
pip install ninja packaging wheel
git submodule update --init --recursive

pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128
pip install -U nvidia-cublas-cu12 nvidia-cuda-cupti-cu12 nvidia-cuda-nvrtc-cu12 nvidia-cuda-runtime-cu12 nvidia-cudnn-cu12 nvidia-cufft-cu12 nvidia-curand-cu12 nvidia-cusolver-cu12 nvidia-cusparse-cu12 nvidia-cusparselt-cu12 nvidia-nvjitlink-cu12 nvidia-nvtx-cu12

$env:TORCH_CUDA_ARCH_LIST="10.0;12.0"
# If your computer performance is good enough, you can set a larger MAX_JOBS value
$env:MAX_JOBS="1"

python setup.py bdist_wheel

# finish. The xformers wheel file is in the "dist" path
pip install ./dist/xformers-xxx-win_amd64.whl