Task-Agnostic Structured Pruning of Speech Representation Models

被引:1
|
作者
Wang, Haoyu [1 ]
Wang, Siyuan [1 ]
Zhang, Wei-Qiang [1 ]
Suo, Hongbin [2 ]
Wan, Yulong [2 ]
机构
[1] Tsinghua Univ, Dept Elect Engn, Beijing 100084, Peoples R China
[2] OPPO, Data & AI Engn Syst, Beijing 100026, Peoples R China
来源
基金
中国国家自然科学基金;
关键词
Model pruning; knowledge distillation; model compression; representation learning;
D O I
10.21437/Interspeech.2023-1442
中图分类号
O42 [声学];
学科分类号
070206 ; 082403 ;
摘要
Self-supervised pre-trained models such as Wav2vec2, Hubert, and WavLM have been shown to significantly improve many speech tasks. However, their large memory and strong computational requirements hinder their industrial applicability. Structured pruning is a hardware-friendly model compression technique but usually results in a larger loss of accuracy. In this paper, we propose a fine-grained attention head pruning method to compensate for the performance degradation. In addition, we also introduce the straight through estimator into the L0 regularization to further accelerate the pruned model. Experiments on the SUPERB benchmark show that our model can achieve comparable performance to the dense model in multiple tasks and outperforms the Wav2vec 2.0 base model on average, with 72% fewer parameters and 2 times faster inference speed.
引用
收藏
页码:231 / 235
页数:5
相关论文
共 50 条
  • [21] PreQuant: A Task-agnostic Quantization Approach for Pre-trained Language Models
    Gong, Zhuocheng
    Liu, Jiahao
    Wang, Qifan
    Yang, Yang
    Wang, Jingang
    Wu, Wei
    Xiang, Yunsen
    Zhao, Dongyan
    Yan, Rui
    FINDINGS OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (ACL 2023), 2023, : 8065 - 8079
  • [22] Mimic and Fool: A Task-Agnostic Adversarial Attack
    Chaturvedi, Akshay
    Garain, Utpal
    IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2021, 32 (04) : 1801 - 1808
  • [23] SuperTickets: Drawing Task-Agnostic Lottery Tickets from Supernets via Jointly Architecture Searching and Parameter Pruning
    You, Haoran
    Li, Baopu
    Sun, Zhanyi
    Xu Ouyang
    Lin, Yingyan
    COMPUTER VISION, ECCV 2022, PT XI, 2022, 13671 : 674 - 690
  • [24] Pivotal Role of Language Modeling in Recommender Systems: Enriching Task-specific and Task-agnostic Representation Learning
    Shin, Kyuyong
    Kwak, Hanock
    Kim, Wonjae
    Jeong, Jisu
    Jung, Seungjae
    Kim, Kyung-Min
    Ha, Jung-Woo
    Lee, Sang-Woo
    PROCEEDINGS OF THE 61ST ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, ACL 2023, VOL 1, 2023, : 1146 - 1161
  • [25] TAFA: A Task-Agnostic Fingerprinting Algorithm for Neural Networks
    Pan, Xudong
    Zhang, Mi
    Lu, Yifan
    Yang, Min
    COMPUTER SECURITY - ESORICS 2021, PT I, 2021, 12972 : 542 - 562
  • [26] Task-Agnostic Adaptive Activation Scaling Network for LLMs
    Jia, Ni
    Liu, Tong
    Chen, Jiadi
    Zhang, Ying
    Han, Song
    IEEE ACCESS, 2025, 13 : 31774 - 31784
  • [27] Task-Agnostic Evolution of Diverse Repertoires of Swarm Behaviours
    Gomes, Jorge
    Christensen, Anders Lyhne
    SWARM INTELLIGENCE (ANTS 2018), 2018, 11172 : 225 - 238
  • [28] Learning Task-Agnostic Action Spaces for Movement Optimization
    Babadi, Amin
    van de Panne, Michiel
    Liu, C. Karen
    Hamalainen, Perttu
    IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS, 2022, 28 (12) : 4700 - 4712
  • [29] TAPE: Task-Agnostic Prior Embedding for Image Restoration
    Liu, Lin
    Xie, Lingxi
    Zhang, Xiaopeng
    Yuan, Shanxin
    Chen, Xiangyu
    Zhou, Wengang
    Li, Houqiang
    Tian, Qi
    COMPUTER VISION - ECCV 2022, PT XVIII, 2022, 13678 : 447 - 464
  • [30] EViLBERT: Learning Task-Agnostic Multimodal Sense Embeddings
    Calabrese, Agostina
    Bevilacqua, Michele
    Navigli, Roberto
    PROCEEDINGS OF THE TWENTY-NINTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2020, : 481 - 487