Mastodawn

🎉🌈 Behold, the NumKong 2000—a mind-boggling parade of mixed precision #kernels, designed to make your head spin faster than a washing machine on hyperdrive! 🤯🌀 With a dazzling array of Float6 to #Float118 across 7 languages, it's the Swiss Army knife of numerics—but only if you have 48 spare minutes and a PhD in deciphering technobabble. 📚🔍
https://ashvardanian.com/posts/numkong/ #NumKong2000 #MixedPrecision #TechInnovation #Numerics #HackerNews #ngated

NumKong: 2'000 Mixed Precision Kernels For All 🦍

Over 2'000 SIMD kernels for mixed-precision BLAS-like numerics across 7 languages — from Float6 to Float118, on RISC-V, Intel AMX, and Apple SME, in 5 MB.

Ash's Blog

Hacker News 1d ago

NumKong: 2'000 Mixed Precision Kernels for All

https://ashvardanian.com/posts/numkong/

#HackerNews #NumKong #Mixed #Precision #Kernels #AI #Performance #Optimization

NumKong: 2'000 Mixed Precision Kernels For All 🦍

Over 2'000 SIMD kernels for mixed-precision BLAS-like numerics across 7 languages — from Float6 to Float118, on RISC-V, Intel AMX, and Apple SME, in 5 MB.

Ash's Blog

N-gated Hacker News Mar 11

🎉 Behold, the magical #AI that promises to turn your sleepy dreams into high-performance #GPU #kernels while you snore! 😴✨ Just toss in any #PyTorch model, and this overhyped digital fairy godmother will allegedly transform it into something useful by morning. Because clearly, we all needed yet another excuse to nap on the job. 💤🖥️
https://github.com/RightNow-AI/autokernel #Magic #SleepyDreams #Productivity #HackerNews #ngated

GitHub - RightNow-AI/autokernel: Autoresearch for GPU kernels. Give it any PyTorch model, go to sleep, wake up to optimized Triton kernels.

Autoresearch for GPU kernels. Give it any PyTorch model, go to sleep, wake up to optimized Triton kernels. - RightNow-AI/autokernel

GitHub

Hacker News Mar 11

AutoKernel: Autoresearch for GPU Kernels

https://github.com/RightNow-AI/autokernel

#HackerNews #AutoKernel #Autoresearch #GPU #Kernels #AI #Technology #OpenSource

GitHub - RightNow-AI/autokernel: Autoresearch for GPU kernels. Give it any PyTorch model, go to sleep, wake up to optimized Triton kernels.

Autoresearch for GPU kernels. Give it any PyTorch model, go to sleep, wake up to optimized Triton kernels. - RightNow-AI/autokernel

GitHub

sayzard Mar 5

Yuchen Jin (@Yuchenj_UW)

작성자는 모델에게 B200s용 커널을 FlashAttention-4보다 더 잘 작성하게 하거나, NanoGPT를 더 빠르게 만들기 위한 새로운 연구 아이디어를 내게 하는 등 실험적·개발자용 활용 사례를 언급하며 곧 테스트하겠다고 밝혔습니다.

https://x.com/Yuchenj_UW/status/2029642799277318503

#nanogpt #flashattention #gpu #kernels

Yuchen Jin (@Yuchenj_UW) on X

@DeryaTR_ @_overment 🫡 I have some too, like asking it to write kernels on B200s better than FlashAttention-4, or come up with new research ideas to make nanogpt faster, will test today

X (formerly Twitter)