Prebuilt FlashAttention wheel (CUDA 13.2 / Torch 2.12 nightly) #13405

CozzyTyler started this conversation in Show and tell

CozzyTyler
Apr 14, 2026

I built a FlashAttention wheel for CUDA 13.2 (cu132) + PyTorch 2.12 nightly.

Tested with:

ComfyUI v0.19.0
SDXL / Flux

GitHub release:
https://github.com/CozzyTyler/flash-attn-cu132-wheel

Note:
This is NOT universal — built for a specific stack.

Replies: 0 comments

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment