​Sequential Attention: Making AI models leaner and faster without sacrificing accuracy

Google Research Blog ·

Google Research introduced Sequential Attention, a method for reducing model compute while preserving accuracy.

Categories: Research

Excerpt

Algorithms & Theory