Flash-MSA: Accelerating Million-Token Training with Sparse Attention Kernels

 (nanduruganesh.github.io)

10 points | by rawsh 2 hours ago

0 comments