Improving the Efficiency of Transformers for Resource-Constrained Devices

被引:12
|
作者
Tabani, Hamid [1 ]
Balasubramaniam, Ajay [1 ]
Marzban, Shabbir [1 ]
Arani, Elahe [1 ]
Zonooz, Bahram [1 ]
机构
[1] NavInfo Europe, Adv Res Lab, Eindhoven, Netherlands
关键词
Deep Learning; Transformers; Clustering; Resource-Constrained Devices;
D O I
10.1109/DSD53832.2021.00074
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
Transformers provide promising accuracy and have become popular and used in various domains such as natural language processing and computer vision. However, due to their massive number of model parameters, memory and computation requirements, they are not suitable for resource-constrained low-power devices. Even with high-performance and specialized devices, the memory bandwidth can become a performance-limiting bottleneck. In this paper, we present a performance analysis of state-of-the-art vision transformers on several devices. We propose to reduce the overall memory footprint and memory transfers by clustering the model parameters. We show that by using only 64 clusters to represent model parameters, it is possible to reduce the data transfer from the main memory by more than 4x, achieve up to 22% speedup and 39% energy savings on mobile devices with less than 0.1% accuracy loss.
引用
收藏
页码:449 / 456
页数:8
相关论文
共 50 条
  • [1] Improving Computational Efficiency of Prognostics Algorithms in Resource-Constrained Settings
    Jarvis, Katelyn J.
    Teubert, Christopher
    Okolo, Wendy A.
    Kulkarni, Chetan S.
    2022 IEEE AEROSPACE CONFERENCE (AERO), 2022,
  • [2] BePOCH: Improving Federated Learning Performance in Resource-Constrained Computing Devices
    Ibraimi, Lenart
    Selimi, Mennan
    Freitag, Felix
    2021 IEEE GLOBAL COMMUNICATIONS CONFERENCE (GLOBECOM), 2021,
  • [3] Improving Network Training on Resource-Constrained Devices via Habituation Normalization
    Lai, Huixia
    Zhang, Lulu
    Zhang, Shi
    SENSORS, 2022, 22 (24)
  • [4] Remote Gaming on Resource-Constrained Devices
    Reza, Waazim
    Kalva, Hari
    Kaufman, Richard
    APPLICATIONS OF DIGITAL IMAGE PROCESSING XXXIII, 2010, 7798
  • [5] Efficiency and Security Evaluation of Lightweight Cryptographic Algorithms for Resource-Constrained IoT Devices
    Radhakrishnan, Indu
    Jadon, Shruti
    Honnavalli, Prasad B.
    SENSORS, 2024, 24 (12)
  • [6] Achieving High Efficiency: Resource sharing techniques in artificial neural networks for resource-constrained devices
    Gorbounov, Y.
    Chen, H.
    1ST WORKSHOP ON SOLITON THEORY, NONLINEAR DYNAMICS AND MACHINE LEARNING, 2024, 2719
  • [7] Encoding semantic awareness in resource-constrained devices
    Preuveneers, Davy
    Berbers, Yolande
    IEEE INTELLIGENT SYSTEMS, 2008, 23 (02) : 26 - 33
  • [8] SmartDedup: Optimizing Deduplication for Resource-constrained Devices
    Yang, Qirui
    Jin, Runyu
    Zhao, Ming
    PROCEEDINGS OF THE 2019 USENIX ANNUAL TECHNICAL CONFERENCE, 2019, : 633 - 646
  • [9] An Affordance Detection Pipeline for Resource-Constrained Devices
    Apicella, Tommaso
    Cavallaro, Andrea
    Berta, Riccardo
    Gastaldo, Paolo
    Bellotti, Francesco
    Ragusa, Edoardo
    2021 28TH IEEE INTERNATIONAL CONFERENCE ON ELECTRONICS, CIRCUITS, AND SYSTEMS (IEEE ICECS 2021), 2021,
  • [10] Code protection for resource-constrained embedded devices
    Saputra, H
    Chen, G
    Brooks, R
    Vijaykrishnan, N
    Kandemir, M
    Irwin, MJ
    ACM SIGPLAN NOTICES, 2004, 39 (07) : 240 - 248