Aran Komatsuzaki    @arankomatsuzaki
Scaling Vision with Sparse Mixture of Experts Trains a 15B Vision Transformer with MoE that attains 90.35% on ImageNet. https://t.co/UitlGeymrv
 Reply      Retweet   20      Like    126   







 







Aran Komatsuzaki

Xiaohua Zhai

Behnam Neyshabur

Facebook AI

Mark O. Riedl

Aleksa Gordić