TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Power Attention: Efficient CUDA Kernels for Symmetric Power Transformers

6 pointsby txus4 months ago

2 comments

txus4 months ago
At Manifest AI we have just released our open-source CUDA kernels to implement Symmetric Power Transformers, as described in our paper from back in August:<p><a href="https:&#x2F;&#x2F;manifestai.com&#x2F;articles&#x2F;symmetric-power-transformers&#x2F;" rel="nofollow">https:&#x2F;&#x2F;manifestai.com&#x2F;articles&#x2F;symmetric-power-transformers...</a><p>Since this is a variant of a linear attention, you get linear cost when training (as opposed to quadratic in regular attention), and constant when doing inference. This is especially attractive for longer contexts!<p>Have a look and play with it -- and of course contributions are very welcome! It&#x27;s an early alpha!
adriablancafort4 months ago
Big leap forward for the field! kudos on the great work guys