Improved Knowledge Distillation for Pre-trained Language Models via Knowledge Selection
Chenglong Wang, Yi Lu, Yongyu Mu, Yimin Hu, Tong Xiao, Jingbo Zhu
Abstract
Knowledge distillation addresses the problem of transferring knowledge from a teacher model to a student model.In this process, we typically have multiple types of knowledge extracted from the teacher model.The problem is to make full use of them to train the student model.Our preliminary study shows that: (1) not all of the knowledge is necessary for learning a good student model, and (2) knowledge distillation can benefit from certain knowledge at different training steps.In response to these, we propose an actor-critic approach to selecting appropriate knowledge to transfer during the process of knowledge distillation.In addition, we offer a refinement of the training algorithm to ease the computational burden.Experimental results on the GLUE datasets show that our method outperforms several strong knowledge distillation baselines significantly.
BibTeX
@inproceedings{wang-etal-2022-improved,
title = "Improved Knowledge Distillation for Pre-trained Language Models via Knowledge Selection",
author = "Wang, Chenglong and
Lu, Yi and
Mu, Yongyu and
Hu, Yimin and
Xiao, Tong and
Zhu, Jingbo",
editor = "Goldberg, Yoav and
Kozareva, Zornitsa and
Zhang, Yue",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2022",
month = dec,
year = "2022",
address = "Abu Dhabi, United Arab Emirates",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2022.findings-emnlp.464/",
doi = "10.18653/v1/2022.findings-emnlp.464",
pages = "6232--6244"
}