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

 (nanduruganesh.github.io)

6 points | by rawsh 1 hour ago

0 comments