Maybe consider putting cutlass in your CUDA/Triton kernels

2 pointsposted 11 hours ago
by todsacerdoti
(maknee.github.io)

No comments yet