6 ms·
B200 Attention Kernel from Scratch to Near-SOTA in 60 Diagrams
- xiphias2 15d agoPretty cool, it's missing how these concepts map to ThunderKittens code (my favorite CUDA toolkit)
- erichocean 15d agoReally interesting project, appreciated.
- peter_d_sherman 14d agoOne 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!