lldacing

/

flash-attention-windows-wheel

like 325

This took an ungodly amount of time to compile on my Win 11 9800X3D, RTX 4090, 64 GB DDR5 6000 so I figured it would save other some trouble if they need this specific setup.

Flash Attention 2.8.4
Windows amd64
Python 3.12
CUDA 13.2
PyTorch 2.12

https://github.com/akashskypatel/flash-attention/releases/download/v2.8.4/flash_attn-2.8.4-cp312-cp312-win_amd64.whl