PoWER-BERT: Accelerating BERT Inference via Progressive Word-vector Elimination
Saurabh Goyal, Anamitra Roy Choudhury, Saurabh Raje, Venkatesan Chakaravarthy, Yogish Sabharwal, Ashish Verma
Abstract
We develop a novel method, called PoWER-BERT, for improving the inference time of the popular BERT model, while maintaining the accuracy. It works by: a) exploiting redundancy pertaining to word-vectors (intermediate transformer block outputs) and eliminating the redundant vectors. b) determining which word-vectors to eliminate by developing a strategy for measuring their significance, based on the self-attention mechanism. c) learning how many word-vectors to eliminate by augmenting the BERT model and the loss function. Experiments on the standard GLUE benchmark shows that PoWER-BERT achieves up to 4.5x reduction in inference time over BERT with < 1% loss in accuracy. We show that PoWER-BERT offers significantly better trade-off between accuracy and inference time compared to prior methods. We demonstrate that our method attains up to 6.8x reduction in inference time with < 1% loss in accuracy when applied over ALBERT, a highly compressed version of BERT. The code for PoWER-BERT is publicly available at https://github.com/IBM/PoWER-BERT.
BibTeX
@InProceedings{pmlr-v119-goyal20a,
title = {{P}o{WER}-{BERT}: Accelerating {BERT} Inference via Progressive Word-vector Elimination},
author = {Goyal, Saurabh and Choudhury, Anamitra Roy and Raje, Saurabh and Chakaravarthy, Venkatesan and Sabharwal, Yogish and Verma, Ashish},
booktitle = {Proceedings of the 37th International Conference on Machine Learning},
pages = {3690--3699},
year = {2020},
editor = {III, Hal Daumé and Singh, Aarti},
volume = {119},
series = {Proceedings of Machine Learning Research},
month = {13--18 Jul},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v119/goyal20a/goyal20a.pdf},
url = {https://proceedings.mlr.press/v119/goyal20a.html},
abstract = {We develop a novel method, called PoWER-BERT, for improving the inference time of the popular BERT model, while maintaining the accuracy. It works by: a) exploiting redundancy pertaining to word-vectors (intermediate transformer block outputs) and eliminating the redundant vectors. b) determining which word-vectors to eliminate by developing a strategy for measuring their significance, based on the self-attention mechanism. c) learning how many word-vectors to eliminate by augmenting the BERT model and the loss function. Experiments on the standard GLUE benchmark shows that PoWER-BERT achieves up to 4.5x reduction in inference time over BERT with < 1% loss in accuracy. We show that PoWER-BERT offers significantly better trade-off between accuracy and inference time compared to prior methods. We demonstrate that our method attains up to 6.8x reduction in inference time with < 1% loss in accuracy when applied over ALBERT, a highly compressed version of BERT. The code for PoWER-BERT is publicly available at https://github.com/IBM/PoWER-BERT.}
}