Tweeted By @srush_nlp
Attention in Dex compared to Einsum. As requested by @_rockt .
— Sasha Rush (@srush_nlp) February 12, 2021
Key differences -> abstracted dimensions, semantic names, non-linear functions, no repeated renaming.
(Two versions in Dex one verbose, one with inference.) pic.twitter.com/V9FcKJMoVq