ALBERT: A Lite BERT for Self-supervised Learning of Language Representations: Difference between revisions

From statwiki
Jump to navigation Jump to search
No edit summary
No edit summary
Line 14: Line 14:


===Inter-sentence coherence loss===
===Inter-sentence coherence loss===
[[File:SOPvsNSP.PNG|frame|Relationship between convexity and smoothness.]]


===Removing dropout===
===Removing dropout===

Revision as of 19:01, 2 November 2020

Presented by

Maziar Dadbin

Introduction

Motivation

Model details

Factorized embedding parameterization

Cross-layer parameter sharing

Inter-sentence coherence loss

File:SOPvsNSP.PNG
Relationship between convexity and smoothness.

Removing dropout