brainsteam.co.uk/brainsteam/content/bookmarks/2022/04/02/models-official-projects-token-dropping-at-master-tensorflow-models1648877142.md

bookmark-of	date	post_meta	tags	title	type	url
title url models/official/projects/token_dropping at master · tensorflow/models · GitHub https://github.com/tensorflow/models/tree/master/official/projects/token_dropping	2022-04-02T01:25:42.935585	date	nlp	models/official/projects/token_dropping at master · tensorflow/models	bookmarks	/bookmarks/2022/04/02/models-official-projects-token-dropping-at-master-tensorflow-models1648877142

Token dropping aims to accelerate the pretraining of transformer models such as BERT without degrading its performance on downstream tasks.

A BERT model pretrained using this token dropping method is not different to a BERT model pretrained in the conventional way: a BERT checkpoint pretrained with token dropping can be viewed and used as a normal BERT checkpoint, for finetuning etc.