A Global Past-Future Early Exit Method for Accelerating Inference of Pre-trained Language Models

A Global Past-Future Early Exit Method for Accelerating Inference of Pre-trained Language Models Kaiyuan Liao author Yi Zhang author Xuancheng Ren author Qi Su author Xu Sun author Bin He author 2021-06 text Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies Kristina Toutanova editor Anna Rumshisky editor Luke Zettlemoyer editor Dilek Hakkani-Tur editor Iz Beltagy editor Steven Bethard editor Ryan Cotterell editor Tanmoy Chakraborty editor Yichao Zhou editor Association for Computational Linguistics Online conference publication liao-etal-2021-global 10.18653/v1/2021.naacl-main.162 https://aclanthology.org/2021.naacl-main.162/ 2021-06 2013 2023