SwitchHead is a novel method that reduces compute and memory requirements in Transformers, achieving speedup while maintaining language modeling performance. It uses Mixture-of-Experts layers and requires fewer attention matrices than standard Transformers.

https://arxiv.org/abs//2312.07987

YouTube: https://www.youtube.com/@ArxivPapers

TikTok: https://www.tiktok.com/@arxiv_papers

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers