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

5 points | by rawsh an hour ago

No comments yet