0

GLU Variants Improve Transformer

Test of GLU and its variants in Transformer's feed-forward layers improves quality over ReLU or GELU.

Year
2020
Venue
arXiv 2020
Authors
1
Hosting
Abstract onlyARXIV-DEFAULT

Cite

Notes

Only stored in your browser.

Attribution

Abstract & full text
arxiv.org/abs/2002.05202ARXIV-DEFAULT
TL;DR
Semantic Scholar
Attribution policy →

Abstract

Gated Linear Units (arXiv:1612.08083) consist of the component-wise product of two linear projections, one of which is first passed through a sigmoid function. Variations on GLU are possible, using different nonlinear (or even linear) functions in place of sigmoid. We test these variants in the feed-forward sublayers of the Transformer (arXiv:1706.03762) sequence-to-sequence model, and find that some of them yield quality improvements over the typically-used ReLU or GELU activations.

Authors

1