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. 自动化研究所
  3. 中国科学院自动化研究所
  4. 复杂系统认知与决策实验室
  5. 类脑芯片与系统研究
Optimization-Based Post-Training Quantization With Bit-Split and Stitching

文献类型：期刊论文

作者Wang, Peisong2; Chen, Weihan2; He, Xiangyu2; Chen, Qiang2; Liu, Qingshan1; Cheng, Jian2
刊名IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
出版日期2023-02-01
卷号45期号:2页码:2119-2135
关键词Deep neural networks compression quantization post-training quantization
ISSN号0162-8828
DOI10.1109/TPAMI.2022.3159369
通讯作者Wang, Peisong(peisong.wang@nlpr.ia.ac.cn)
英文摘要Deep neural networks have shown great promise in various domains. Meanwhile, problems including the storage and computing overheads arise along with these breakthroughs. To solve these problems, network quantization has received increasing attention due to its high efficiency and hardware-friendly property. Nonetheless, most existing quantization approaches rely on the full training dataset and the time-consuming fine-tuning process to retain accuracy. Post-training quantization does not have these problems, however, it has mainly been shown effective for 8-bit quantization. In this paper, we theoretically analyze the effect of network quantization and show that the quantization loss in the final output layer is bounded by the layer-wise activation reconstruction error. Based on this analysis, we propose an Optimization-based Post-training Quantization framework and a novel Bit-split optimization approach to achieve minimal accuracy degradation. The proposed framework is validated on a variety of computer vision tasks, including image classification, object detection, instance segmentation, with various network architectures. Specifically, we achieve near-original model performance even when quantizing FP32 models to 3-bit without fine-tuning.
资助项目National Key Research and Development Program of China[2020AAA0103402] ; National Natural Science Foundation of China[61906193] ; Strategic Priority Research Program of Chinese Academy of Sciences[XDA27040300]
WOS研究方向Computer Science ; Engineering
语种英语
WOS记录号WOS:000912386000051
出版者IEEE COMPUTER SOC
资助机构National Key Research and Development Program of China ; National Natural Science Foundation of China ; Strategic Priority Research Program of Chinese Academy of Sciences
源URL[http://ir.ia.ac.cn/handle/173211/51368]  
专题类脑芯片与系统研究
通讯作者Wang, Peisong
作者单位1.Nanjing Univ Informat Sci & Technol, B DAT, Nanjing 210044, Peoples R China
2.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
推荐引用方式
GB/T 7714		 Wang, Peisong,Chen, Weihan,He, Xiangyu,et al. Optimization-Based Post-Training Quantization With Bit-Split and Stitching[J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,2023,45(2):2119-2135.
APA Wang, Peisong,Chen, Weihan,He, Xiangyu,Chen, Qiang,Liu, Qingshan,&Cheng, Jian.(2023).Optimization-Based Post-Training Quantization With Bit-Split and Stitching.IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,45(2),2119-2135.
MLA Wang, Peisong,et al."Optimization-Based Post-Training Quantization With Bit-Split and Stitching".IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 45.2(2023):2119-2135.

入库方式： OAI收割

来源：自动化研究所

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。