Generally Intelligent cover image

Episode 33: Tri Dao, Stanford: On FlashAttention and sparsity, quantization, and efficient inference

Generally Intelligent

00:00

How to Scale to Longer Context Longevity

The goal there is, we've been interested for a few years of scaling to longer context length. And so we saw that one of the bottleneck is in the attention layer. We wanted to design some of my attention layers that can scale to longer sequences yet still remain hardware efficient. That's something that lots of folks have been working on for the past years or so ever since transfer is not popular.

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app