Get the latest tech news

Batmobile: 10-20x Faster CUDA Kernels for Equivariant Graph Neural Networks


Systems engineer and educator. Building and teaching GPU programming, CUDA, and low-level ML systems.

None

Get the Android app

Or read this on Hacker News

Read more on:

Photo of neural networks

neural networks

Photo of 20x

20x

Photo of faster cuda kernels

faster cuda kernels

Related news:

News photo

Evolution: Training neural networks with genetic selection achieves 81% on MNIST

News photo

Neural Networks: Zero to Hero

News photo

OpenAI experiment finds that sparse models could give AI builders the tools to debug neural networks