Is there a flash-attention for win_amd64?

[yapuppycat]

Hi guys! I'm looking for assets related to windows but can't seem to find them. Am I looking in the wrong place? Can someone shed some light on this?

[JackTheTripperr]

Flash attention isn't officially supported on Windows, so you'll need to either compile yourself or find a .whl that matches your exact Python, CUDA, PyTorch, and flash_attn versions. Of note, I recently stumbled across this repo that has a number of pre-built wheels for a variety of Windows configurations and a really handy Actions workflow that builds a .whl based on your configuration.