B200 Attention Kernel from Scratch to Near-SOTA in 60 Diagramsiaroslavelistratov.github.io 24 pointsmagoghm10 days ago3 commentsSaveHideCopy link On HNComments−xiphias210dPretty cool, it's missing how these concepts map to ThunderKittens code (my favorite CUDA toolkit)−peter_d_sherman9dOne of the best AI GPU Attention Kernel how-to-implement articles I've read in a long time!Tremendous work, tremendous effort was placed into writing this article -- and it shows!Well done!−erichocean10dReally interesting project, appreciated.
Comments
Pretty cool, it's missing how these concepts map to ThunderKittens code (my favorite CUDA toolkit)
One of the best AI GPU Attention Kernel how-to-implement articles I've read in a long time!
Tremendous work, tremendous effort was placed into writing this article -- and it shows!
Well done!
Really interesting project, appreciated.