Does anyone know how to add Sage Attention and xformers etc. to this on windows?, thanks

[10ANT]

title

[nitinmukesh]

very easy.
Make sure to install CUDA 12.4 and python 3.12.

python.exe -s -m pip install triton-windows

pip install https://github.com/woct0rdho/SageAttention/releases/download/v2.1.1-windows/sageattention-2.1.1+cu126torch2.6.0-cp312-cp312-win_amd64.whl

[KorporateRdr]

While I was able to successfully install triton and the right SageAttention package, I still get the warning it's not installed?

[NonaBuhtig]

You don't need to install all the attention speed-ups. You only need the faster one, which is Sage-attention. :)

[woct0rdho]

While I was able to successfully install triton and the right SageAttention package, I still get the warning it's not installed?

You need to make sure that you install them in the same Python environment that you run FramePack, see https://github.com/woct0rdho/triton-windows#2-python-environment

If you're using the one-click package (with Python 3.10), then you also need to download the two folders 'include' and 'libs' here https://github.com/woct0rdho/triton-windows/releases/download/v3.0.0-windows.post1/python_3.10.11_include_libs.zip and copy them to framepack_cu126_torch26\system\python\

[SunGreen777]

I think if it makes sense lllyasviel will add it to the installation

[batfruit]

its a sticky wicket as sage attention on windows requires MSVC, Cuda Toolkit and Windows-Triton which can potentially mess with other diffusion setups.

xformers and flash attention do not boost this project only sage attention

[woct0rdho]

Recently we've simplified the installation by a lot, and there is no need to install Visual Studio or CUDA toolkit. A lot of instructions online are outdated, compared to how fast AI evolves. You can find a bit more detailed instructions here https://github.com/woct0rdho/SageAttention/releases

[batfruit]

Thanks for your reply and your excellent work :)