Hacker Next
Flash-MSA: Accelerating Million-Token Training with Sparse Attention Kernelsnanduruganesh.github.io
39 points by rawsh 1 days ago | 5 comments