First I would like to thank Dao-AILab for Flash Attention 2 & lldacing for the work to create these Windows Flash Attention 2 Wheels for us.
This is a clone of "lldacing/flash-attention-windows-wheel" repo. Unfortunately Huggingface changes the download link file name format for these Flash Attention Wheel file names and no longer work with Pip due to the wrong formatting, unless you know exactly what to change back to correct the incorrect download format. So I'm putting them on Github, so they will work again by either copying the download link to Pip install or download the file and then Pip install it, your choice.