TRANSFORMER-VQ: EFFICIENT LINEAR-TIME ATTENTION

Summary from the TLDR AI newsletter

This study presents Transformer-VQ, a new design of transformer that processes attention faster thanks to unique vector-based keys and caching.