中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
首页 机构 成果 学者
  
  1. CAS IR Grid
  2. 计算技术研究所
  3. 中国科学院计算技术研究所
  4. 中国科学院计算技术研究所期刊论文
  5. 英文
ETBench: Characterizing Hybrid Vision Transformer Workloads Across Edge Devices

文献类型：期刊论文

作者Zhou, Yingkun2,3; Tian, Zhengshuyuan2,3; Yang, Wenhao2,3; Zhang, Tingting1,4; Ye, Jinpeng2,3; Han, Chenji2,3; Liu, Tianyi5; Zhang, Fuxin2
刊名IEEE TRANSACTIONS ON COMPUTERS
出版日期2025-06-01
卷号74期号:6页码:1857-1871
关键词Benchmark deep learning edge devices vision transformer hybrid model Benchmark deep learning edge devices vision transformer hybrid model
ISSN号0018-9340
DOI10.1109/TC.2025.3543697
英文摘要Lightweight Convolution and Vision Transformer hybrid models have increasingly dominated the frontiers of deep learning (DL) on edge devices; however, to the best of our knowledge, no prior work has provided comprehensive evaluation on hybrid models' performance and analyzed their characteristics by diving deep into the edge ecosystem with diversified modern DL inference engines and heterogeneous hardware. This paper proposes a comprehensive open-source benchmark suite, ETBench, to allow power-efficiency, performance and accuracy assessment for state-of-the-art (SOTA) hybrid models across 11 most widely-used DL engines deployed on diverse edge devices. After building ETBench that satisfies 6 design requirements proposed in our work, we conduct extensive experiments on 14 devices including 19 CPUs, 11 GPUs and 5 NPUs, and obtain benchmark results from all deployment scenarios (combinations of models, quantization formats, software engines, and hardware platforms). Valuable observations and insightful implications are finally summarized. For example, within current DL engines, the INT8 quantization is significantly underperformed in terms of accuracy and speed against FP16 for hybrid models. Overall, ETBench serves as a collaborative platform that assists model architects in better evaluating their models and makes it possible for future co-optimizations of DL engines and hardware accelerators.
资助项目Chinese Academy of Sciences[XDC05020100]
WOS研究方向Computer Science ; Engineering
语种英语
WOS记录号WOS:001492675500026
出版者IEEE COMPUTER SOC
源URL[http://119.78.100.204/handle/2XEOYT63/42331]  
专题中国科学院计算技术研究所期刊论文_英文
通讯作者Zhang, Tingting
作者单位1.Chinese Acad Sci, Inst Comp Technol, Beijing 100190, Peoples R China
2.Chinese Acad Sci, Inst Comp Technol, State Key Lab Processors, Beijing 100190, Peoples R China
3.Univ Chinese Acad Sci, Beijing 100049, Peoples R China
4.Loongson Technol Co Ltd, Beijing 100095, Peoples R China
5.Univ Texas San Antonio, San Antonio, TX 78249 USA
推荐引用方式
GB/T 7714		 Zhou, Yingkun,Tian, Zhengshuyuan,Yang, Wenhao,et al. ETBench: Characterizing Hybrid Vision Transformer Workloads Across Edge Devices[J]. IEEE TRANSACTIONS ON COMPUTERS,2025,74(6):1857-1871.
APA Zhou, Yingkun.,Tian, Zhengshuyuan.,Yang, Wenhao.,Zhang, Tingting.,Ye, Jinpeng.,...&Zhang, Fuxin.(2025).ETBench: Characterizing Hybrid Vision Transformer Workloads Across Edge Devices.IEEE TRANSACTIONS ON COMPUTERS,74(6),1857-1871.
MLA Zhou, Yingkun,et al."ETBench: Characterizing Hybrid Vision Transformer Workloads Across Edge Devices".IEEE TRANSACTIONS ON COMPUTERS 74.6(2025):1857-1871.

入库方式： OAI收割

来源：计算技术研究所

浏览0
下载0
收藏0
其他版本

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。