Task-specific Compression for Multi-task Language Models using Attribution-based Pruning

被引:0
|
作者
Yang, Nakyeong [1 ]
Jang, Yunah [1 ]
Lee, Hwanhee [2 ]
Jung, Seohyeong [3 ]
Jung, Kyomin [1 ]
机构
[1] Seoul Natl Univ, Seoul, South Korea
[2] Chung Ang Univ, Seoul, South Korea
[3] Hyundai Motor Grp & 42dot Inc, Seoul, South Korea
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Multi-task language models show outstanding performance for various natural language understanding tasks with only a single model. However, these language models utilize an unnecessarily large number of model parameters, even when used only for a specific task. This paper proposes a novel training-free compression method for multi-task language models using a pruning method. Specifically, we use an attribution method to determine which neurons are essential for performing a specific task. We task-specifically prune unimportant neurons and leave only task-specific parameters. Furthermore, we extend our method to be applicable in low-resource and unsupervised settings. Since our compression method is training-free, it uses few computing resources and does not destroy the pre-trained knowledge of language models. Experimental results on the six widely-used datasets show that our proposed pruning method significantly outperforms baseline pruning methods. In addition, we demonstrate that our method preserves performance even in an unseen domain setting.
引用
收藏
页码:594 / 604
页数:11
相关论文
共 50 条
  • [21] Figuratively Speaking: Authorship Attribution via Multi-Task Figurative Language Modeling
    Katsios, Gregorios A.
    Sa, Ning
    Strzalkowski, Tomek
    FINDINGS OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS: ACL 2024, 2024, : 13240 - 13255
  • [22] Task's Choice: Pruning-Based Feature Sharing (PBFS) for Multi-Task Learning
    Chen, Ying
    Yu, Jiong
    Zhao, Yutong
    Chen, Jiaying
    Du, Xusheng
    ENTROPY, 2022, 24 (03)
  • [23] Multi-task incentive contract with specific task ability
    Wang, Chunping
    Liu, Hua
    Advances in Modelling and Analysis A, 2017, 54 (02): : 185 - 196
  • [24] Multi-task Learning with Bidirectional Language Models for Text Classification
    Yang, Qi
    Shang, Lin
    2019 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2019,
  • [25] Additive multi-task learning models and task diagnostics
    Miller, Nikolay
    Zhang, Guoyi
    COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION, 2024, 53 (12) : 6120 - 6137
  • [26] A Multi-Task Semantic Decomposition Framework with Task-specific Pre-training for Few-Shot NER
    Dong, Guanting
    Wang, Zechen
    Zhao, Jinxu
    Zhao, Gang
    Guo, Daichi
    Fu, Dayuan
    Hui, Tingfeng
    Zeng, Chen
    He, Keqing
    Li, Xuefeng
    Wang, Liwen
    Cui, Xinyue
    Xu, Weiran
    PROCEEDINGS OF THE 32ND ACM INTERNATIONAL CONFERENCE ON INFORMATION AND KNOWLEDGE MANAGEMENT, CIKM 2023, 2023, : 430 - 440
  • [27] T3S: Improving Multi-Task Reinforcement Learning with Task-Specific Feature Selector and Scheduler
    Yu, Yuanqiang
    Yang, Tianpei
    Lv, Yongliang
    Zheng, Yan
    Hao, Jianye
    2023 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS, IJCNN, 2023,
  • [28] Language Modelling as a Multi-Task Problem
    Weber, Lucas
    Jumelet, Jaap
    Bruni, Elia
    Hupkes, Dieuwke
    16TH CONFERENCE OF THE EUROPEAN CHAPTER OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (EACL 2021), 2021, : 2049 - 2060
  • [29] Attribution of Adversarial Attacks via Multi-task Learning
    Guo, Zhongyi
    Han, Keji
    Ge, Yao
    Li, Yun
    Ji, Wei
    NEURAL INFORMATION PROCESSING, ICONIP 2023, PT II, 2024, 14448 : 81 - 94
  • [30] Making Small Language Models Better Multi-task Learners with Mixture-of-Task-Adapters
    Xie, Yukang
    Wang, Chengyu
    Yan, Junbing
    Zhou, Jiyong
    Deng, Feiqi
    Huang, Jun
    PROCEEDINGS OF THE 17TH ACM INTERNATIONAL CONFERENCE ON WEB SEARCH AND DATA MINING, WSDM 2024, 2024, : 1094 - 1097