Abstract
In the field of deep learning, pre-trained BERT models have achieved remarkable success. However, the accompanying problem is that models with more complex structures and more network parameters. The huge parameter size makes the computational cost in terms of time and memory become extremely expensive. Recent work has indicated that BERT models own a significant amount of redundant attention heads. Meanwhile considerable BERT models compression algorithms have been proposed, which can effectively reduce model complexity and redundancy with pruning some attention heads. Nevertheless, existing automated model compression solutions are mainly based on predetermined pruning program, which requires multiple expensive pruning-retraining cycles or heuristic designs to select additional hyperparameters. Furthermore, the training process of BERT models is a black box, and lacks interpretability, which makes researchers cannot intuitively understand the optimization process of the model. In this paper, we propose a visual analysis system, BHPVAS, for pruning BERT models, which helps researchers to incorporate their understanding of model structure and operating mechanism into the model pruning process and generate pruning schemes. We propose three pruning criteria based on the attention data, namely, importance score, stability score, and similarity score, for evaluating the importance of self-attention heads. Additionally, we design multiple collaborative views to display the entire pruning process, guiding users to carry out pruning. Our system supports exploring the role of self-attention heads in the model inference process using text dependency relations and attention weight distribution. Finally, we conduct two case studies to demonstrate how to use the system for Sentiment Classification Sample Analysis and Pruning Scheme Exploration, verifying the effectiveness of the visual analysis system.
Graphical Abstract
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
References
Ahn Y, Lin Y-R (2020) Fairsight: visual analytics for fairness in decision making. IEEE Trans Vis Comput Graph 26(1):1086–1095. https://doi.org/10.1109/TVCG.2019.2934262
Aken B, Winter B, Löser A, Gers FA (2020) Visbert: hidden-state visualizations for transformers. In: Companion proceedings of the web conference 2020, WWW’20. Association for Computing Machinery, New York, pp 207–211. https://doi.org/10.1145/3366424.3383542
Brown TB, Mann B, Ryder N, Subbiah M, Kaplan J, Dhariwal P, Neelakantan A, Shyam P, Sastry G, Askell A, Agarwal S, Herbert-Voss A, Krueger G, Henighan T, Child R, Ramesh A, Ziegler DM, Wu J, Winter C, Hesse C, Chen M, Sigler E, Litwin M, Gray S, Chess B, Clark J, Berner C, McCandlish S, Radford A, Sutskever I, Amodei D (2020) Language models are few-shot learners. In: Proceedings of the 34th international conference on neural information processing systems, NIPS’20. Curran Associates Inc., Red Hook. https://doi.org/10.5555/3495724.3495883
Cao K, Liu M, Su H, Wu J, Zhu J, Liu S (2021) Analyzing the noise robustness of deep neural networks. IEEE Trans Vis Comput Graph 27(7):3289–3304. https://doi.org/10.1109/TVCG.2020.2969185
Carreira-Perpinan MA, Idelbayev Y (2018) Learning-compression” algorithms for neural net pruning. In: 2018 IEEE/CVF Conference on computer vision and pattern recognition, pp 8532–8541. https://doi.org/10.1109/CVPR.2018.00890
Cavallo M, Demiralp Ç (2018) Clustrophile 2: guided visual clustering analysis. IEEE Trans Vis Comput Graph 25(1):267–276. https://doi.org/10.1109/TVCG.2018.2864477
Chauvin Y (1988) A back-propagation algorithm with optimal use of hidden units. In: Proceedings of the 1st international conference on neural information processing systems, NIPS’88. MIT Press, Cambridge, pp 519–526. https://doi.org/10.5555/2969735.2969795
Chiliang Z, Tao H, Yingda G, Zuochang Y (2019) Accelerating convolutional neural networks with dynamic channel pruning. In: 2019 Data compression conference (DCC), pp 563–563. https://doi.org/10.1109/DCC.2019.00075
Cortes C, Mohri M, Rostamizadeh A (2012) Algorithms for learning kernels based on centered alignment. J Mach Learn Res 13(1):795–828. https://doi.org/10.5555/2503308.2188413
DeRose JF, Wang J, Berger M (2021) Attention flows: Analyzing and comparing attention mechanisms in language models. IEEE Trans Vis Comput Graph 27(2):1160–1170. https://doi.org/10.1109/TVCG.2020.3028976
Devlin J, Chang M-W, Lee K, Toutanova K (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. In: Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers). Association for Computational Linguistics, Minneapolis, pp 4171–4186. https://doi.org/10.18653/v1/N19-1423
Ghojogh B, Crowley M, Karray F, Ghodsi A (2023) Uniform manifold approximation and projection (UMAP). Springer, Cham, pp 479–497. https://doi.org/10.1007/978-3-031-10602-6_17
Gordon M, Duh K, Andrews N (2020) Compressing BERT: studying the effects of weight pruning on transfer learning. In: Proceedings of the 5th workshop on representation learning for NLP. Association for Computational Linguistics, Online, pp 143–155. https://doi.org/10.18653/v1/2020.repl4nlp-1.18
Gretton A, Bousquet O, Smola A, Schölkopf B (2005) Measuring statistical dependence with Hilbert–Schmidt norms. In: Jain S, Simon HU, Tomita E (eds) Algorithmic learning theory. Springer, Berlin, pp 63–77. https://doi.org/10.1007/11564089_7
Guo F-M, Liu S, Mungall FS, Lin X, Wang Y (2019) Reweighted proximal pruning for large-scale language representation. ArXiv arXiv:1909.12486
Guo F-M, Liu S, Mungall FS, Lin X, Wang Y (2020) Reweighted proximal pruning for large-scale language representation
Guo Y, Yao A, Chen Y (2016) Dynamic network surgery for efficient DNNs. In: Proceedings of the 30th international conference on neural information processing systems, NIPS’16. Curran Associates Inc., Red Hook, pp 1387–1395. https://doi.org/10.5555/3157096.3157251
Han D, Pan J, Pan R, Zhou D, Cao N, He J, Xu M, Chen W (2022) inet: Visual analysis of irregular transition in multivariate dynamic networks. Front Comput Sci. https://doi.org/10.1007/s11704-020-0013-1
Han M, Kim J (2019) Joint banknote recognition and counterfeit detection using explainable artificial intelligence. Sensors. https://doi.org/10.3390/s19163607
Han S, Pool J, Tran J, Dally WJ (2015) Learning both weights and connections for efficient neural networks. In: Proceedings of the 28th international conference on neural information processing systems—volume 1, NIPS’15. MIT Press, Cambridge, pp 1135–1143. https://doi.org/10.5555/2969239.2969366
He T, Jin X, Ding G, Yi L, Yan C (2019) Towards better uncertainty sampling: active learning with multiple views for deep convolutional neural network. In: 2019 IEEE international conference on multimedia and expo (ICME), pp 1360–1365. https://doi.org/10.1109/ICME.2019.00236
He Y, Zhang X, Sun J (2017) Channel pruning for accelerating very deep neural networks. In: 2017 IEEE international conference on computer vision (ICCV), pp 1398–1406. https://doi.org/10.1109/ICCV.2017.155
Ji X, Tu Y, He W, Wang J, Shen H-W, Yen P-Y (2021) Usevis: visual analytics of attention-based neural embedding in information retrieval. Vis Inform 5(2):1–12. https://doi.org/10.1016/j.visinf.2021.03.003
Kahng M, Andrews PY, Kalro A, Chau DH (2018) Activis: visual exploration of industry-scale deep neural network models. IEEE Trans Vis Comput Graph 24(1):88–97. https://doi.org/10.1109/TVCG.2017.2744718
Kornblith S, Norouzi M, Lee H, Hinton G (2019). Similarity of neural network representations revisited. In: International conference on machine learning. PMLR, pp 3519–3529
Leroux S, Bohez S, De Coninck E, Verbelen T, Vankeirsbilck B, Simoens P, Dhoedt B (2017) The cascading neural network: building the internet of smart things. Knowl Inf Syst 52:791–814. https://doi.org/10.1007/s10115-017-1029-1
Lin J, Rao Y, Lu J, Zhou J (2017a) Runtime neural pruning. In: NIPS, pp 2178–2188
Luo J-H, Wu J, Lin W (2017b) Thinet: a filter level pruning method for deep neural network compression. In: 2017 IEEE International conference on computer vision (ICCV), pp 5068–5076. https://doi.org/10.1109/ICCV.2017.541
MarietZ Sara S (2016) Diversity networks: neural network compression using determinantal point processes. In: Proceedings of the 4th international conference on learning representations, pp 67–79
Michel P, Levy O, Neubig G (2019a) Are sixteen heads really better than one? Curran Associates Inc., Red Hook. https://doi.org/10.5555/3454287.3455544
Michel P, Levy O, Neubig G (2019b) Are sixteen heads really better than one? In: Neural information processing systems
Ming Y, Cao S, Zhang R, Li Z, Chen Y, Song Y, Qu H (2017) Understanding hidden memories of recurrent neural networks. In: 2017 IEEE conference on visual analytics science and technology (VAST), pp 13–24. https://doi.org/10.1109/VAST.2017.8585721
Ming Y, Xu P, Cheng F, Qu H, Ren L (2020) Protosteer: steering deep sequence model with prototypes. IEEE Trans Vis Comput Graph 26(1):238–248. https://doi.org/10.1109/TVCG.2019.2934267
Ming Y, Xu P, Qu H, Ren L (2019) Interpretable and steerable sequence learning via prototypes. In: KDD’19. Association for Computing Machinery, New York, pp 903–913. https://doi.org/10.1145/3292500.3330908
Mozer M C, Smolensky P (1988) Skeletonization: a technique for trimming the fat from a network via relevance assessment. In: Proceedings of the 1st international conference on neural information processing systems, NIPS’88. MIT Press, Cambridge, pp 107–115. https://doi.org/10.5555/2969735.2969748
Peng Y, Fan X, Chen R, Yu Z, Liu S, Chen Y, Zhao Y, Zhou F (2023) Visual abstraction of dynamic network via improved multi-class blue noise sampling. Front Comput Sci. https://doi.org/10.1007/s11704-021-0609-0
Strobelt H, Gehrmann S, Pfister H, Rush AM (2018) Lstmvis: A tool for visual analysis of hidden state dynamics in recurrent neural networks. IEEE Trans Vis Comput Graph 24(1):667–676. https://doi.org/10.1109/TVCG.2017.2744158
Tan S, Caruana R, Hooker G, Lou Y (2018) Distill-and-compare: auditing black-box models using transparent model distillation. In: Proceedings of the 2018 AAAI/ACM conference on AI, ethics, and society, AIES’18. Association for Computing Machinery, New York, pp 303–310. https://doi.org/10.1145/3278721.3278725
Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, Kaiser L, Polosukhin I (2017) Attention is all you need. In: Proceedings of the 31st international conference on neural information processing systems, NIPS’17. Curran Associates Inc., Red Hook, pp 6000–6010. https://doi.org/10.5555/3295222.3295349
Voita E, Talbot D, Moiseev F, Sennrich R, Titov I (2019) Analyzing multi-head self-attention: specialized heads do the heavy lifting, the rest can be pruned. In: Proceedings of the 57th annual meeting of the association for computational linguistics. Association for Computational Linguistics, Florence, pp 5797–5808. https://doi.org/10.18653/v1/P19-1580
Wang J, Gou L, Shen H-W, Yang H (2019) Dqnviz: a visual analytics approach to understand deep q-networks. IEEE Trans Vis Comput Graph 25(1):288–298. https://doi.org/10.1109/TVCG.2018.2864504
Wang Y, Feng C, Guo C, Chu Y, Hwang J-N (2019) Solving the sparsity problem in recommendations via cross-domain item embedding based on co-clustering. In: Proceedings of the twelfth ACM international conference on web search and data mining, WSDM’19. Association for Computing Machinery, New York, pp 717–725. https://doi.org/10.1145/3289600.3290973
Wu Z, Nagarajan T, Kumar A, Rennie S, Davis LS, Grauman K, Feris R (2018) Blockdrop: dynamic inference paths in residual networks. In: 2018 IEEE/CVF conference on computer vision and pattern recognition, pp 8817–8826. https://doi.org/10.1109/CVPR.2018.00919
Xia M, Zhong Z, Chen D (2022) Structured pruning learns compact and accurate models. In: Proceedings of the 60th annual meeting of the association for computational linguistics (volume 1: long papers). Association for Computational Linguistics, Dublin, pp 1513–1528. https://doi.org/10.18653/v1/2022.acl-long.107
Yang W, Wang X, Lu J, Dou W, Liu S (2021) Interactive steering of hierarchical clustering. IEEE Trans Vis Comput Graph 27(10):3953–3967. https://doi.org/10.1109/TVCG.2020.2995100
Yuan J, Chen C, Yang W, Liu M, Xia J, Liu S (2021) A survey of visual analytics techniques for machine learning. Comput Vis Media 7:3–36. https://doi.org/10.1007/41095-020-0191-7
Acknowledgements
This work was supported by the Zhejiang Provincial Natural Science Foundation of China under Grant No. LY22F020023 and No. LZ22F020015, the National Natural Science Foundation of China under Grant No. 61972122 and No. U22A2033
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Supplementary file 1 (mp4 21542 KB)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Liu, Z., Sun, H., Sun, H. et al. BHPVAS: visual analysis system for pruning attention heads in BERT model. J Vis (2024). https://doi.org/10.1007/s12650-024-00985-z
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s12650-024-00985-z