This is the official implementation for the paper in ACL-IJCNLP 2021 "EarlyBERT: Efficient BERT Training via Early-bird Lottery Tickets" by Xiaohan Chen, Yu Cheng, Shuohang Wang, Zhe Gan, Zhangyang Wang and Jingjing Liu.
The code base consists of the finetuning and pretraining parts.
If you use this code base for your research project, please include this work as reference
@inproceedings{chen2021earlybert,
title={EarlyBERT: Efficient BERT Training via Early-bird Lottery Tickets},
author={Chen, Xiaohan and Cheng, Yu and Wang, Shuohang and Gan, Zhe and Wang,
Zhangyang and Liu, Jingjing},
booktitle={Proceedings of the Joint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing},
year={2021}
}