Article(id=1327012848833769687, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327012646391493336, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202550145, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1740758400000, receivedDateStr=2025-03-01, revisedDate=1744214400000, revisedDateStr=2025-04-10, acceptedDate=null, acceptedDateStr=null, onlineDate=1794314117868, onlineDateStr=2026-11-10, pubDate=1748793600000, pubDateStr=2025-06-02, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794314117868, onlineIssueDateStr=2026-11-10, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794314117868, creator=13701087609, updateTime=1794314117868, updator=13701087609, issue=Issue{id=1327012646391493336, tenantId=1146029695717560320, journalId=1324759914146529478, year='2025', volume='62', issue='6', pageStart='1331', pageEnd='1593', issueExtLink='null', onlineDate='null', pubDate='1748793600000', pubDateStr='2025-06-02', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794314069601, creator='13701087609', updateTime=1794376817187, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1327275829270262152, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327012646391493336, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1327275829270262153, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327012646391493336, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=1530, endPage=1546, ext={EN=ArticleExt(id=1327012849089622232, articleId=1327012848833769687, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=DAQ: Divide-and-Conquer Strategy Based Adaptive Low-Bit Quantization Method for Vision Transformer, columnId=1327012676955386745, journalTitle=Journal of Computer Research and Development, columnName=Chip Architecture Design, runingTitle=null, highlight=null, articleAbstract=
Vision Transformers (ViTs) have demonstrated remarkable success in computer vision tasks, but their complex architecture and computational demands hinder deployment on edge devices. While post-training quantization (PTQ) is widely adopted for model compression, existing PTQ methods exhibit severe performance degradation in 4-bit ultra-low-bitwidth scenarios. This work systematically addresses two fundamental limitations: 1) spatial mismatch between quantization-sensitive layers (e.g., Softmax) and compute-intensive layers (e.g., linear projections), where quantizing Softmax causes 80% accuracy loss despite contributing merely 8% computational load; 2) non-Gaussian activation distributions with hidden Gaussian-like clustering properties (97% values less than three times z-score). We propose DAQ (divide-and-conquer and adaptive quantization), a hardware-friendly PTQ method. DAQ adpots z-score-driven dynamic partitioning algorithm to separate data into normal-range and abnormal-range groups and quantizes the two groups with connected parameter. DAQ further explores hardware accelerated kernel such as tensor core to speed up quantization ViT models. Experimental results demonstrate that DAQ achieves a maximum improvement of 4.37% in ImageNet Top-1 accuracy under 4-bit quantization. In object detection tasks, its average error margin remains below 0.4% compared with the baseline and achieves a maximum improvement of 8.2%, even surpassing the full-precision model by 0.1% in specific cases, thereby realizing near-lossless low-bit-width quantization. Through hardware implementation optimization, DAQ achieves 43%~86% computational acceleration without significantly increasing computational overhead. This approach establishes a synergistic algorithm-hardware co-optimized quantization deployment paradigm for resource-constrained scenarios, effectively balancing model efficiency and precision retention.
, authors=Qianru Lü, Jinwei Xu*, Jingfei Jiang, Dongsheng Li, authorsList=Qianru Lü, Jinwei Xu, Jingfei Jiang, Dongsheng Li, authorCompany=null, correspAuthors=Jinwei Xu, authorNote=
born in 1993. PhD candidate. Her main research interests include computer architecture and artificial intelligence
born in 1990. PhD, assistant research fellow. Member of CCF. His main research interests include artificial intelligence and reconfigurable computing
born in 1974. PhD, professor. Member of CCF. Her main research interests include reconfigurable computing, artificial intelligence, and computer architecture. (jingfeijiang@nudt.edu.cn)
born in 1978. PhD, professor. Member of CCF. His main research interests include parallel computing, artificial intelligence, computer architecture.(dsli@nudt.edu.cn)
, correspAuthorsNote=null, copyrightStatement=Copyright © 2025 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1327012853753688313, articleId=1327012848833769687, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=DAQ:基于分治策略的自适应Vision Transformer低位宽量化方法, columnId=1327012677139936123, journalTitle=计算机研究与发展, columnName=芯片架构设计, runingTitle=null, highlight=null, articleAbstract=
视觉Transformer(Vision Transformer,ViT)模型在计算机视觉领域的多项任务中取得显著效果. 但ViT的复杂结构和计算开销限制了其在边缘计算设备中的部署. 训练后量化(post-training quantization,PTQ)技术被广泛应用于ViT模型轻量化中以解决实际部署难题,但现有PTQ方法在低位宽量化中的性能损失较大. 针对低比特量化场景,ViT的量化敏感层(如Softmax)与计算密集层(如线性变换)存在显著空间错位,且非高斯分布的激活值中隐含97%的类高斯聚集特性. 由此,基于标准分数z-score方法提出分治自适应量化(divide-and-conquer and adaptive quantization,DAQ)方法,通过量化敏感度-计算-存储开销联合分析与硬件协同设计,实现精度与效率的联合优化. DAQ构建动态分治量化机制,通过动态感知的z-score方法实现正常值/离群值双域分割,均匀关联量化2个值域. 在4-bit量化下,DAQ方法在分类任务上的 Top-1精度最大提升4.37个百分点,目标检测任务最大精度提升达8.2个百分点,与基线模型相比误差平均低于0.4个百分点,超过最佳全精度模型0.1个百分点,接近实现无损的低位宽量化. 另一方面,DAQ在硬件兼容设上适配Tensor Core的INT4/INT8内核,以量化定点计算来减轻线性计算压力. 实验表明,DAQ硬件适配后对线性计算部分有43%~86%的加速效果,为资源受限场景提供了算法-硬件协同优化的量化部署范式.
, authors=吕倩茹, 许金伟*, 姜晶菲, 李东升, authorsList=吕倩茹, 许金伟, 姜晶菲, 李东升, authorCompany=null, correspAuthors=许金伟, authorNote=
 |
1993年生. 博士研究生. 主要研究方向为计算机体系结构、人工智能 |
 |
1990年生. 博士,助理研究员. CCF会员. 主要研究方向为人工智能、可重构计算 |
 |
1974年生. 博士,研究员. CCF会员. 主要研究方向为可重构计算、人工智能、计算机体系结构 |
 |
1978年生. 博士,研究员. CCF会员. 主要研究方向为并行计算、人工智能、计算机体系结构 |
, correspAuthorsNote=
, copyrightStatement=版权所有 © 2025 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=eWPGr/Lts+x+zZwWHiekqA==, magXml=SyCqTo3QvdEk6m6/zEYbIw==, pdfUrl=null, pdf=P5tcP0utotUd/BdQSjoBUw==, pdfFileSize=1929097, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=p3VOfOFQcFfqWGvgKVFQcw==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=raPHIiJidZdEdPS7bWYQqA==, mapNumber=null, fund=null)}, authors=[Author(id=1327012855892783358, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=lvqianru11@nudt.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327012855980863744, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, authorId=1327012855892783358, language=EN, stringName=Qianru Lü, firstName=Qianru, middleName=null, lastName=Lü, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073, bio={"content":"
born in 1993. PhD candidate. Her main research interests include computer architecture and artificial intelligence
"}, bioImg=null, bioContent=
born in 1993. PhD candidate. Her main research interests include computer architecture and artificial intelligence
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012856106692865, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, authorId=1327012855892783358, language=CN, stringName=吕倩茹, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=并行与分布式计算全国重点实验室(国防科技大学) 长沙 410073, bio={"img":"ZgduueglBbUgwej86bW8yw==","content":"
1993年生. 博士研究生. 主要研究方向为计算机体系结构、人工智能
"}, bioImg=ZgduueglBbUgwej86bW8yw==, bioContent=
1993年生. 博士研究生. 主要研究方向为计算机体系结构、人工智能
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012855792120058, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, xref=null, ext=[AuthorCompanyExt(id=1327012855800508667, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1327012855813091580, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=并行与分布式计算全国重点实验室(国防科技大学) 长沙 410073)])]), Author(id=1327012856190578947, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=xujinwei13@nudt.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1327012856266076421, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, authorId=1327012856190578947, language=EN, stringName=Jinwei Xu, firstName=Jinwei, middleName=null, lastName=Xu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073, bio={"content":"
born in 1990. PhD, assistant research fellow. Member of CCF. His main research interests include artificial intelligence and reconfigurable computing
"}, bioImg=null, bioContent=
born in 1990. PhD, assistant research fellow. Member of CCF. His main research interests include artificial intelligence and reconfigurable computing
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012856358351110, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, authorId=1327012856190578947, language=CN, stringName=许金伟, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=并行与分布式计算全国重点实验室(国防科技大学) 长沙 410073, bio={"img":"rDgLGmpg2Wu9oMQw22aV5Q==","content":"
1990年生. 博士,助理研究员. CCF会员. 主要研究方向为人工智能、可重构计算
"}, bioImg=rDgLGmpg2Wu9oMQw22aV5Q==, bioContent=
1990年生. 博士,助理研究员. CCF会员. 主要研究方向为人工智能、可重构计算
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012855792120058, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, xref=null, ext=[AuthorCompanyExt(id=1327012855800508667, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1327012855813091580, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=并行与分布式计算全国重点实验室(国防科技大学) 长沙 410073)])]), Author(id=1327012856450625800, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=jingfeijiang@nudt.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327012856538706186, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, authorId=1327012856450625800, language=EN, stringName=Jingfei Jiang, firstName=Jingfei, middleName=null, lastName=Jiang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073, bio={"content":"
born in 1974. PhD, professor. Member of CCF. Her main research interests include reconfigurable computing, artificial intelligence, and computer architecture. (jingfeijiang@nudt.edu.cn)
"}, bioImg=null, bioContent=
born in 1974. PhD, professor. Member of CCF. Her main research interests include reconfigurable computing, artificial intelligence, and computer architecture. (jingfeijiang@nudt.edu.cn)
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012856622592267, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, authorId=1327012856450625800, language=CN, stringName=姜晶菲, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=并行与分布式计算全国重点实验室(国防科技大学) 长沙 410073, bio={"img":"OJLHEqzMHfK4t7630lRdTA==","content":"
1974年生. 博士,研究员. CCF会员. 主要研究方向为可重构计算、人工智能、计算机体系结构
"}, bioImg=OJLHEqzMHfK4t7630lRdTA==, bioContent=
1974年生. 博士,研究员. CCF会员. 主要研究方向为可重构计算、人工智能、计算机体系结构
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012855792120058, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, xref=null, ext=[AuthorCompanyExt(id=1327012855800508667, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1327012855813091580, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=并行与分布式计算全国重点实验室(国防科技大学) 长沙 410073)])]), Author(id=1327012856706478349, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=dsli@nudt.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327012856802947343, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, authorId=1327012856706478349, language=EN, stringName=Dongsheng Li, firstName=Dongsheng, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073, bio={"content":"
born in 1978. PhD, professor. Member of CCF. His main research interests include parallel computing, artificial intelligence, computer architecture.(dsli@nudt.edu.cn)
"}, bioImg=null, bioContent=
born in 1978. PhD, professor. Member of CCF. His main research interests include parallel computing, artificial intelligence, computer architecture.(dsli@nudt.edu.cn)
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012856924582161, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, authorId=1327012856706478349, language=CN, stringName=李东升, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=并行与分布式计算全国重点实验室(国防科技大学) 长沙 410073, bio={"img":"3coSrMUWsUa0ACx/Oqr+0w==","content":"
1978年生. 博士,研究员. CCF会员. 主要研究方向为并行计算、人工智能、计算机体系结构
"}, bioImg=3coSrMUWsUa0ACx/Oqr+0w==, bioContent=
1978年生. 博士,研究员. CCF会员. 主要研究方向为并行计算、人工智能、计算机体系结构
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012855792120058, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, xref=null, ext=[AuthorCompanyExt(id=1327012855800508667, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1327012855813091580, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=并行与分布式计算全国重点实验室(国防科技大学) 长沙 410073)])])], keywords=[Keyword(id=1327012857042022674, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, orderNo=1, keyword=vision Transformer (ViT)), Keyword(id=1327012857104937235, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, orderNo=2, keyword=post-training quantization (PTQ)), Keyword(id=1327012857180434708, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, orderNo=3, keyword=outlier), Keyword(id=1327012857260126485, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, orderNo=4, keyword=low-bit quantization), Keyword(id=1327012857331429654, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, orderNo=5, keyword=z-score), Keyword(id=1327012857402732823, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, orderNo=6, keyword=uniform correlation quantization), Keyword(id=1327012857490813208, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, orderNo=1, keyword=视觉Transformer(ViT)), Keyword(id=1327012857599865113, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, orderNo=2, keyword=训练后量化(PTQ)), Keyword(id=1327012857679556890, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, orderNo=3, keyword=离群值), Keyword(id=1327012857759248667, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, orderNo=4, keyword=低比特量化), Keyword(id=1327012857843134748, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, orderNo=5, keyword=z-score), Keyword(id=1327012857927020829, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, orderNo=6, keyword=均匀关联量化)], refs=[Reference(id=1327012865430630730, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017, 30: 5998−6008, articleTitle=null, refAbstract=null), Reference(id=1327012865556459851, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16×16 words: Transformers for image recognition at scale[J]. arXiv preprint, arXiv: 2010.11929, 2020, articleTitle=null, refAbstract=null), Reference(id=1327012865627763020, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=Touvron H, Cord M, Douze M, et al. Training data-efficient image transformers & distillation through attention[J]. arXiv: 2012.12877, articleTitle=null, refAbstract=null), Reference(id=1327012865699066189, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Ze, Lin Yutong, Cao Yue, et al. Swin Transformer: Hierarchical vision Transformer using shifted windows[C]//Proc of the IEEE/CVF Inte Conf on Computer Vision. Piscataway, NJ: IEEE, 2021: 10012−10022, articleTitle=null, refAbstract=null), Reference(id=1327012865812312398, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=He Kaiming, Zhang Xiangyu, Ren Shaoqing, et al. Deep residual learning for image recognition[C]//Proc of the IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2016: 770−778, articleTitle=null, refAbstract=null), Reference(id=1327012865929752911, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Carion N, Massa F, Synnaeve G, et al. End-to-end object detection with transformers[C]//Proc of European Conf on Computer Vision. Berlin: Springer 2020: 213−229, articleTitle=null, refAbstract=null), Reference(id=1327012866005250384, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Cheng B, Schwing A, Kirillov A. Per-pixel classification is not all you need for semantic segmentation[J]. Advances in Neural Information Processing Systems, 2021, 34: 17864−1787, articleTitle=null, refAbstract=null), Reference(id=1327012866089136465, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Gündüç Y. Vit-GAN: Image-to-image translation with vision Transformes and conditional GANS[J]. arXiv preprint, arXiv: 2110.09305, 2021, articleTitle=null, refAbstract=null), Reference(id=1327012866168828242, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Hatamizadeh A, Song J, Liu G, et al. Diffit: Diffusion vision transformers for image generation[C]//Proc of European Conf on Computer Vision. Berlin: Springer, 2024: 37−55, articleTitle=null, refAbstract=null), Reference(id=1327012866240131411, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Lin Yang, Zhang Tianyu, Sun Peiqin, et al. Fq-ViT: Fully quantized vision Transformer without retraining[J]. arXiv preprint, arXiv: 2111.13824, 2021, articleTitle=null, refAbstract=null), Reference(id=1327012866357571924, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Yuan Zhihang, Xue Chenhao, Chen Yiqi, et al. PTQ4ViT: Post-training quantization for vision transformers with twin uniform quantization[C]//Proc of European Conf on Computer Vision. Berlin: Springer, 2022: 191−207, articleTitle=null, refAbstract=null), Reference(id=1327012866512761173, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Yanjing, Xu Sheng, Zhang Baochang, et al. Q-ViT: Accurate and fully quantized low-bit vision Transformer[J]. Advances in Neural Information Processing Systems, 2022, 35: 34451−34463, articleTitle=null, refAbstract=null), Reference(id=1327012866693116246, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhong Yunshan, Huang You, Hu Jiawei, et al. Towards accurate post-training quantization of vision transformers via error reduction[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025, 47(4): 2676−2692, articleTitle=null, refAbstract=null), Reference(id=1327012866781196631, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Zhikai, Xiao Junrui, Yang Lianwei, et al. RepQ-ViT: Scale reparameterization for post-training quantization of vision transformers[C]//Proc of the IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2023: 17227−17236, articleTitle=null, refAbstract=null), Reference(id=1327012866877665624, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Zhikai, Liu Xuewen, Zhang Jing, et al. RepQuant: Towards accurate post-training quantization of large Transformer models via scale reparameterization[J]. arXiv preprint, arXiv: 2402.05628, 2024, articleTitle=null, refAbstract=null), Reference(id=1327012866965746009, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Du Dayou, Gong Gu, Chu Xiaowen. Model quantization and hardware acceleration for vision Transformers: A comprehensive survey[J]. arXiv preprint, arXiv: 2405. 00314, 2024, articleTitle=null, refAbstract=null), Reference(id=1327012867053826394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Tai Yushan, Lin Mingguang, Wu A. TSPTQ-ViT: Two-scaled post-training quantization for vision Transformer[C]//Proc of 2023 IEEE Int Conf on Acoustics, Speech and Signal Processing (ICASSP). Piscataway, NJ: IEEE, 2023: 1−5, articleTitle=null, refAbstract=null), Reference(id=1327012867133518171, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen C F R, Fan Q, Panda R. Crossvit: Cross-attention multi-scale vision transformer for image classification[C]//Proc of the IEEE/CVF Int Conf on computer vision. Piscataway, NJ: IEEE, 2021: 357−366, articleTitle=null, refAbstract=null), Reference(id=1327012867229987164, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6): 84−90, articleTitle=null, refAbstract=null), Reference(id=1327012867339039069, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhu Xizhou, Su Weijie, Lu Lewei, et al. Deformable DETR: Deformable transformers for end-to-end object detection[J]. arXiv preprint, arXiv: 2010.04159, 2020, articleTitle=null, refAbstract=null), Reference(id=1327012869117423966, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Strudel R, Garcia R, Laptev I, et al. Segmenter: Transformer for semantic segmentation[C]//Proc of the IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2021: 7262−7272, articleTitle=null, refAbstract=null), Reference(id=1327012869205504353, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Zheng Sixiao, Lu Jiachen, Zhao Hengshuang, et al. Rethinking semantic segmentation from a sequence-to-sequence perspective with transformers[C]//Proc of the IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2021: 6881−6890, articleTitle=null, refAbstract=null), Reference(id=1327012869310361954, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Yixin, Zhang Kai, Li Yuan, et al. Sora: A review on background, technology, limitations, and opportunities of large vision models[J]. arXiv preprint, arXiv: 2402.17177, 2024, articleTitle=null, refAbstract=null), Reference(id=1327012869394248035, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Dettmers T, Lewis M, Belkada Y, et al. GPT3. Int8 (): 8-bit matrix multiplication for transformers at scale[J]. Advances in Neural Information Processing Systems, 2022, 35: 30318−30332, articleTitle=null, refAbstract=null), Reference(id=1327012869486522724, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Touvron H, Lavril T, Izacard G, et al. Llama: Open and efficient foundation language models[J]. arXiv preprint, arXiv: 2302.13971, 2023, articleTitle=null, refAbstract=null), Reference(id=1327012869570408805, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Chan T F, Golub G H, LeVeque R J. Algorithms for computing the sample variance: Analysis and recommendations[J]. The American Statistician, 1983, 37(3): 242−247, articleTitle=null, refAbstract=null), Reference(id=1327012869650100582, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Nvidia. NVIDIA GPU Ampere Architecture Whitepaper, https://www.nvidia.com/content/PDF/nvidia-ampere-ga-102-gpu-architecture-whitepaper-v2.pdf, articleTitle=null, refAbstract=null), Reference(id=1327012869750763879, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Deng Jia, Dong Wei, Socher R, et al. ImageNet: A large-scale hierarchical image database[C]//Proc of 2009 IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2009: 248−255, articleTitle=null, refAbstract=null), Reference(id=1327012869838844264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Lin Tsung-Yi, Maire M, Belongie S, et al. Microsoft COCO: Common objects in context[C]//Proc of the 13th European Conf on Computer Vision (ECCV 2014). Berlin: Springer, 2014: 740−755, articleTitle=null, refAbstract=null), Reference(id=1327012869939507561, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Kai, Wang Jiaqi, Pang Jiangmiao, et al. MMDetection: Open mmlab detection toolbox and benchmark[J]. arXiv preprint, arXiv: 1906.07155, 2019, articleTitle=null, refAbstract=null), Reference(id=1327012870023393642, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Frantar E, Ashkboos S, Hoefler T, et al. GPTQ: Accurate post-training quantization for generative pre-trained transformers[J]. arXiv preprint, arXiv: 2210.17323, 2022, articleTitle=null, refAbstract=null), Reference(id=1327012870115668331, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Nagel M, Amjad R A, Van Baalen M, et al. Up or down? adaptive rounding for post-training quantization[C]//Proc of Int Conf on Machine Learning. New York: ACM, 2020: 7197−7206, articleTitle=null, refAbstract=null), Reference(id=1327012870203748716, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Peebles W, Xie S. Scalable diffusion models with transformers[C]//Proc of the IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2023: 4195−4205, articleTitle=null, refAbstract=null), Reference(id=1327012870291829101, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Ho J, Jain A, Abbeel P. Denoising diffusion probabilistic models[J]. Advances in Neural Information Processing Systems, 2020, 33: 6840−6851, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1327012855792120058, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, xref=null, ext=[AuthorCompanyExt(id=1327012855800508667, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1327012855813091580, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, companyId=1327012855792120058, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=并行与分布式计算全国重点实验室(国防科技大学) 长沙 410073)])], figs=[ArticleFig(id=1327012858061238558, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 1, caption=
Illustration of ViT structure, figureFileSmall=BSgLIdIj9Xn4OipRnwCLKA==, figureFileBig=p3VOfOFQcFfqWGvgKVFQcw==, tableContent=null), ArticleFig(id=1327012858136736031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图1, caption=
ViT结构示意图, figureFileSmall=BSgLIdIj9Xn4OipRnwCLKA==, figureFileBig=p3VOfOFQcFfqWGvgKVFQcw==, tableContent=null), ArticleFig(id=1327012860045144352, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 2, caption=
Model accuracy under various quantization widths in ViT-S, figureFileSmall=N/+bWnw/5C/AwoUUUNbp/g==, figureFileBig=LmtrbYd9E6zV1XCqysxTaw==, tableContent=null), ArticleFig(id=1327012860154196258, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图2, caption=
ViT-S不同量化位宽下的模型精度, figureFileSmall=N/+bWnw/5C/AwoUUUNbp/g==, figureFileBig=LmtrbYd9E6zV1XCqysxTaw==, tableContent=null), ArticleFig(id=1327012860280025379, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 3, caption=
Data distribution of partial activations in ViT-S, Block 8, figureFileSmall=OBGQOg2ltk0XVws6gAbFyA==, figureFileBig=pWBtTiIYtW8cybPebLdapg==, tableContent=null), ArticleFig(id=1327012860363911460, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图3, caption=
ViT-S, Block 8部分激活值数据分布, figureFileSmall=OBGQOg2ltk0XVws6gAbFyA==, figureFileBig=pWBtTiIYtW8cybPebLdapg==, tableContent=null), ArticleFig(id=1327012860481351973, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 4, caption=
Extreme value distribution of activation tensor in ViTs reasoning, figureFileSmall=hXZF2NVnqnKh98/detx+yA==, figureFileBig=YnAN9Qp+w485aoZ2VlGS1g==, tableContent=null), ArticleFig(id=1327012860561043750, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图4, caption=
ViTs推理时激活张量的极值分布, figureFileSmall=hXZF2NVnqnKh98/detx+yA==, figureFileBig=YnAN9Qp+w485aoZ2VlGS1g==, tableContent=null), ArticleFig(id=1327012860665901351, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 5, caption=
Proportion of high z-score in ViT-S/DeiT-B/Swin-S activations, figureFileSmall=WFfhoZr4/oZy7pift43BWQ==, figureFileBig=qixb9XJOYItNn5kWod2uag==, tableContent=null), ArticleFig(id=1327012860795924776, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图5, caption=
ViT-S/DeiT-B/Swin-S激活值中高z-score占比, figureFileSmall=WFfhoZr4/oZy7pift43BWQ==, figureFileBig=qixb9XJOYItNn5kWod2uag==, tableContent=null), ArticleFig(id=1327012860909170985, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 6, caption=
Illustration of DAQ quantization process, figureFileSmall=zVWwCdHFEfbydkv/JikZ1g==, figureFileBig=RA1q5EelIVTRUoId30dINg==, tableContent=null), ArticleFig(id=1327012860980474154, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图6, caption=
DAQ量化流程示意图, figureFileSmall=zVWwCdHFEfbydkv/JikZ1g==, figureFileBig=RA1q5EelIVTRUoId30dINg==, tableContent=null), ArticleFig(id=1327012861072748843, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 7, caption=
Flow chart of uniform correlation quantization, figureFileSmall=pOSy/ifAzU4AfQvAljmSGA==, figureFileBig=mpUx+hyXZArUoyq52qwuLg==, tableContent=null), ArticleFig(id=1327012861152440620, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图7, caption=
均匀关联量化流程图, figureFileSmall=pOSy/ifAzU4AfQvAljmSGA==, figureFileBig=mpUx+hyXZArUoyq52qwuLg==, tableContent=null), ArticleFig(id=1327012861253103917, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 8, caption=
Compensation offset and zero-point offset, figureFileSmall=IjL4zf/WPllYqtImFlbKUw==, figureFileBig=thQ9Ad9PsXEIGUojDoTZeA==, tableContent=null), ArticleFig(id=1327012861345378606, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图8, caption=
补偿偏移和零点偏移, figureFileSmall=IjL4zf/WPllYqtImFlbKUw==, figureFileBig=thQ9Ad9PsXEIGUojDoTZeA==, tableContent=null), ArticleFig(id=1327012861454430511, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 9, caption=
Accelerated architecture of DAQ based on GPU, figureFileSmall=1pwtoGairVWe2iBLSbf9dw==, figureFileBig=hSqh1f2dJNiPvanlcNCDbA==, tableContent=null), ArticleFig(id=1327012861517345072, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图9, caption=
DAQ基于GPU的加速结构, figureFileSmall=1pwtoGairVWe2iBLSbf9dw==, figureFileBig=hSqh1f2dJNiPvanlcNCDbA==, tableContent=null), ArticleFig(id=1327012861609619761, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Figure 10, caption=
Ablation study of DAQ, figureFileSmall=69VK7DcbD21lQoBp/sJFVg==, figureFileBig=dYn9afanHOqRmUkpEUTUKA==, tableContent=null), ArticleFig(id=1327012861731254578, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=图10, caption=
DAQ消融实验, figureFileSmall=69VK7DcbD21lQoBp/sJFVg==, figureFileBig=dYn9afanHOqRmUkpEUTUKA==, tableContent=null), ArticleFig(id=1327012861802557747, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 1, caption=
Proportion Analysis of Accuracy Calculation/Storage/Reduction of ViT-S layer, figureFileSmall=null, figureFileBig=null, tableContent=
| 层 | 计算平均 时延占比/% | 参数+激活 存储占比/% | 量化精度 下降/% |
|---|
| 注:黑体数值表示最大值,下划线数值表示最小值. |
| LN | 6 | 5 | 32 |
| QKV Gen | 21 | 17 | 4 |
| Softmax | 2 | 8 | 84 |
| Proj | 9 | 8 | 9 |
| FC1 | 27 | 22 | 40 |
| GELU | 4 | 10 | 65 |
| FC2 | 31 | 30 | 98 |
), ArticleFig(id=1327012861882249524, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表1, caption=
ViT-S层精度计算/存储/下降占比分析
, figureFileSmall=null, figureFileBig=null, tableContent=
| 层 | 计算平均 时延占比/% | 参数+激活 存储占比/% | 量化精度 下降/% |
|---|
| 注:黑体数值表示最大值,下划线数值表示最小值. |
| LN | 6 | 5 | 32 |
| QKV Gen | 21 | 17 | 4 |
| Softmax | 2 | 8 | 84 |
| Proj | 9 | 8 | 9 |
| FC1 | 27 | 22 | 40 |
| GELU | 4 | 10 | 65 |
| FC2 | 31 | 30 | 98 |
), ArticleFig(id=1327012861961941301, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 2, caption=
Meanings of Variables, figureFileSmall=null, figureFileBig=null, tableContent=
| 变量名称 | 含义 |
|---|
| $ \boldsymbol{X} $ | 目标特征张量 |
| $ {\boldsymbol{X}}_{1} $ | 目标特征张量中的正常值集 |
| $ {\boldsymbol{X}}_{2} $ | 目标特征张量中的离群值集 |
| $ \hat{\boldsymbol{X}} $ | $\boldsymbol X $的量化表示 |
| $ \tilde{\boldsymbol{X}} $ | $\boldsymbol X $的反量化结果 |
), ArticleFig(id=1327012862058410294, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表2, caption=
变量含义
, figureFileSmall=null, figureFileBig=null, tableContent=
| 变量名称 | 含义 |
|---|
| $ \boldsymbol{X} $ | 目标特征张量 |
| $ {\boldsymbol{X}}_{1} $ | 目标特征张量中的正常值集 |
| $ {\boldsymbol{X}}_{2} $ | 目标特征张量中的离群值集 |
| $ \hat{\boldsymbol{X}} $ | $\boldsymbol X $的量化表示 |
| $ \tilde{\boldsymbol{X}} $ | $\boldsymbol X $的反量化结果 |
), ArticleFig(id=1327012862159073591, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 3, caption=
4-bit Symmetric Uniform Quantization, figureFileSmall=null, figureFileBig=null, tableContent=
| 数值类型 | 值域 | 量化参数 |
|---|
| 注:“*”表示当仅存单侧正/负离群值时,不加以符号区分. |
| 正常值 | $ 0,\pm 1,\pm 2,\pm 3,\pm 4,\pm 5,\pm 6,\pm 7, -8 $ | s | $z $ |
| 离群值* | $ 0,\pm 1,\pm 2,\pm 3,\pm 4,\pm 5,\pm 6,\pm 7, -8 $ | s′ | $z' $ |
| 正离群值 | $ 0, \mathrm{1,2}, 3, 4, 5, 6, 7, 8 $ | s1 | $z_1 $ |
| 负离群值 | $ -1,-2,-3,-4,-5,-6,-7,\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }-8 $ | s2 | $z_2 $ |
), ArticleFig(id=1327012862238765368, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表3, caption=
4-bit 对称均匀量化
, figureFileSmall=null, figureFileBig=null, tableContent=
| 数值类型 | 值域 | 量化参数 |
|---|
| 注:“*”表示当仅存单侧正/负离群值时,不加以符号区分. |
| 正常值 | $ 0,\pm 1,\pm 2,\pm 3,\pm 4,\pm 5,\pm 6,\pm 7, -8 $ | s | $z $ |
| 离群值* | $ 0,\pm 1,\pm 2,\pm 3,\pm 4,\pm 5,\pm 6,\pm 7, -8 $ | s′ | $z' $ |
| 正离群值 | $ 0, \mathrm{1,2}, 3, 4, 5, 6, 7, 8 $ | s1 | $z_1 $ |
| 负离群值 | $ -1,-2,-3,-4,-5,-6,-7,\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }-8 $ | s2 | $z_2 $ |
), ArticleFig(id=1327012862318457145, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 4, caption=
Computing Patch of Tensor Core and Linear Computational Dimensions of ViT, figureFileSmall=null, figureFileBig=null, tableContent=
| 数据格式 | Tensor Core矩阵分块 M,N,K | ViT线性计算(部分) M,N,K |
|---|
INT8
| 16,16,16 | 197,192,384 |
| 32,8,16 | 197,1536,384 |
| 8,32,16 | 197,3072,768 |
| INT4 | 8,8,32 | 197,2304,768 |
), ArticleFig(id=1327012862402343226, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表4, caption=
Tensor Core计算分块和ViT线性计算维度
, figureFileSmall=null, figureFileBig=null, tableContent=
| 数据格式 | Tensor Core矩阵分块 M,N,K | ViT线性计算(部分) M,N,K |
|---|
INT8
| 16,16,16 | 197,192,384 |
| 32,8,16 | 197,1536,384 |
| 8,32,16 | 197,3072,768 |
| INT4 | 8,8,32 | 197,2304,768 |
), ArticleFig(id=1327012862473646395, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 5, caption=
Top-1 Accuracy of the Quantization Models on ImageNet Dataset of the Image Classification Task, figureFileSmall=null, figureFileBig=null, tableContent=
| 方法 | 量化目标 | 位宽 | 准确率 |
|---|
| ViT-S/% | ViT-B/% | DeiT-T/% | DeiT-S/% | DeiT-B/% | Swin-S/% | Swin-B/% |
|---|
| 注:量化目标S,G,N分别代表Softmax,GELU,LN层输出激活;黑体数值表示最优值. |
| 全精度模型 | | 32/32 | 81.39 | 84.54 | 72.21 | 79.85 | 81.80 | 83.23 | 85.27 |
| FQ-ViT[10] | S/N | 4/4 | 0.10 | 0.10 | 0.10 | 0.10 | 0.10 | 0.10 | 0.10 |
| PTQ4ViT[11] | S/G | 4/4 | 42.75 | 30.96 | 36.96 | 34.08 | 64.39 | 76.09 | 74.02 |
| APQ-ViT[13] | S | 4/4 | 47.95 | 41.41 | 47.94 | 43.55 | 67.48 | 77.15 | 76.48 |
| RepQ-ViT[14] | S/N | 4/4 | 65.05 | 68.48 | 57.43 | 69.03 | 75.61 | 79.45 | 78.32 |
| RepQuant[15] | S/N | 4/4 | 73.28 | 77.84 | 64.44 | 75.21 | 78.46 | 81.52 | 82.80 |
| DAQ(本文) | G/N | 4/4 | 75.06 | 80.36 | 67.15 | 75.46 | 78.52 | 80.43 | 81.45 |
| DAQ(本文) | S/N | 4/4 | 73.29 | 82.00 | 68.81 | 75.45 | 80.08 | 79.95 | 82.30 |
| FQ-ViT[10] | S/N | 6/6 | 4.26 | 0.10 | 58.66 | 45.51 | 64.63 | 66.50 | 52.09 |
| PTQ4ViT[11] | S/G | 6/6 | 78.63 | 81.65 | 69.68 | 76.28 | 80.25 | 82.38 | 84.01 |
| APQ-ViT[13] | S | 6/6 | 79.10 | 82.21 | 70.49 | 77.76 | 80.42 | 82.67 | 84.18 |
| RepQ-ViT[14] | S/N | 6/6 | 80.43 | 83.62 | 70.76 | 78.90 | 81.27 | 82.79 | 84.57 |
| RepQuant[15] | S/N | 6/6 | 80.51 | 83.75 | 70.89 | 79.06 | 81.41 | 82.93 | 84.86 |
| DAQ(本文) | G/N | 6/6 | 80.81 | 84.09 | 71.69 | 79.56 | 81.63 | 82.31 | 83.98 |
| DAQ(本文) | S/N | 6/6 | 80.86 | 83.85 | 71.74 | 79.61 | 81.94 | 82.89 | 84.34 |
), ArticleFig(id=1327012862586892604, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表5, caption=
量化模型在图像分类任务的ImageNet数据集上的Top-1准确率
, figureFileSmall=null, figureFileBig=null, tableContent=
| 方法 | 量化目标 | 位宽 | 准确率 |
|---|
| ViT-S/% | ViT-B/% | DeiT-T/% | DeiT-S/% | DeiT-B/% | Swin-S/% | Swin-B/% |
|---|
| 注:量化目标S,G,N分别代表Softmax,GELU,LN层输出激活;黑体数值表示最优值. |
| 全精度模型 | | 32/32 | 81.39 | 84.54 | 72.21 | 79.85 | 81.80 | 83.23 | 85.27 |
| FQ-ViT[10] | S/N | 4/4 | 0.10 | 0.10 | 0.10 | 0.10 | 0.10 | 0.10 | 0.10 |
| PTQ4ViT[11] | S/G | 4/4 | 42.75 | 30.96 | 36.96 | 34.08 | 64.39 | 76.09 | 74.02 |
| APQ-ViT[13] | S | 4/4 | 47.95 | 41.41 | 47.94 | 43.55 | 67.48 | 77.15 | 76.48 |
| RepQ-ViT[14] | S/N | 4/4 | 65.05 | 68.48 | 57.43 | 69.03 | 75.61 | 79.45 | 78.32 |
| RepQuant[15] | S/N | 4/4 | 73.28 | 77.84 | 64.44 | 75.21 | 78.46 | 81.52 | 82.80 |
| DAQ(本文) | G/N | 4/4 | 75.06 | 80.36 | 67.15 | 75.46 | 78.52 | 80.43 | 81.45 |
| DAQ(本文) | S/N | 4/4 | 73.29 | 82.00 | 68.81 | 75.45 | 80.08 | 79.95 | 82.30 |
| FQ-ViT[10] | S/N | 6/6 | 4.26 | 0.10 | 58.66 | 45.51 | 64.63 | 66.50 | 52.09 |
| PTQ4ViT[11] | S/G | 6/6 | 78.63 | 81.65 | 69.68 | 76.28 | 80.25 | 82.38 | 84.01 |
| APQ-ViT[13] | S | 6/6 | 79.10 | 82.21 | 70.49 | 77.76 | 80.42 | 82.67 | 84.18 |
| RepQ-ViT[14] | S/N | 6/6 | 80.43 | 83.62 | 70.76 | 78.90 | 81.27 | 82.79 | 84.57 |
| RepQuant[15] | S/N | 6/6 | 80.51 | 83.75 | 70.89 | 79.06 | 81.41 | 82.93 | 84.86 |
| DAQ(本文) | G/N | 6/6 | 80.81 | 84.09 | 71.69 | 79.56 | 81.63 | 82.31 | 83.98 |
| DAQ(本文) | S/N | 6/6 | 80.86 | 83.85 | 71.74 | 79.61 | 81.94 | 82.89 | 84.34 |
), ArticleFig(id=1327012862670778685, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 6, caption=
Performance of the Quantization Models on COCO Dataset in the Object Detection Task, figureFileSmall=null, figureFileBig=null, tableContent=
| 方法 | 量化目标 | 位宽 | Mask R-CNN |
|---|
| Swin-T | | Swin-S |
|---|
| $ {AP}^{\rm box} $/% | $ {AP}^{\rm mask} $/% | | $ {AP}^{\rm box} $/% | $ {AP}^{\rm mask} $/% |
|---|
| 注:量化目标S,G,N分别代表Softmax,GELU,LN层输出激活;黑体数值表示最优值. |
| 全精度模型 | | 32/32 | 46.0 | 41.6 | | 48.5 | 43.3 |
| PTQ4ViT[11] | S/G | 4/4 | 6.9 | 7.0 | | 26.7 | 26.6 |
| APQ-ViT[13] | S | 4/4 | 23.7 | 22.6 | | 44.7 | 40.1 |
| RepQ-ViT[14] | S/N | 4/4 | 36.1 | 36.0 | | 44.2 | 40.2 |
| RepQuant[15] | S/N | 4/4 | 37.2 | 36.8 | | 44.5 | 40.5 |
| DAQ(本文) | G/N | 4/4 | 45.4 | 41.3 | | 47.9 | 43.0 |
| DAQ(本文) | S/N | 4/4 | 45.7 | 41.4 | | 48.1 | 43.2 |
| PTQ4ViT[11] | S/G | 6/6 | 5.8 | 6.8 | | 6.5 | 6.6 |
| APQ-ViT[13] | S | 6/6 | 45.4 | 41.2 | | 47.9 | 42.9 |
| RepQ-ViT[14] | S/N | 6/6 | 45.1 | 41.2 | | 47.8 | 43.0 |
| RepQuant[15] | S/N | 6/6 | 45.3 | 41.3 | | 48.1 | 43.2 |
| DAQ(本文) | G/N | 6/6 | 46.0 | 41.6 | | 48.2 | 43.2 |
| DAQ(本文) | S/N | 6/6 | 46.0 | 41.7 | | 48.2 | 43.1 |
), ArticleFig(id=1327012862758859070, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表6, caption=
量化模型在目标检测任务的COCO数据集上的性能
, figureFileSmall=null, figureFileBig=null, tableContent=
| 方法 | 量化目标 | 位宽 | Mask R-CNN |
|---|
| Swin-T | | Swin-S |
|---|
| $ {AP}^{\rm box} $/% | $ {AP}^{\rm mask} $/% | | $ {AP}^{\rm box} $/% | $ {AP}^{\rm mask} $/% |
|---|
| 注:量化目标S,G,N分别代表Softmax,GELU,LN层输出激活;黑体数值表示最优值. |
| 全精度模型 | | 32/32 | 46.0 | 41.6 | | 48.5 | 43.3 |
| PTQ4ViT[11] | S/G | 4/4 | 6.9 | 7.0 | | 26.7 | 26.6 |
| APQ-ViT[13] | S | 4/4 | 23.7 | 22.6 | | 44.7 | 40.1 |
| RepQ-ViT[14] | S/N | 4/4 | 36.1 | 36.0 | | 44.2 | 40.2 |
| RepQuant[15] | S/N | 4/4 | 37.2 | 36.8 | | 44.5 | 40.5 |
| DAQ(本文) | G/N | 4/4 | 45.4 | 41.3 | | 47.9 | 43.0 |
| DAQ(本文) | S/N | 4/4 | 45.7 | 41.4 | | 48.1 | 43.2 |
| PTQ4ViT[11] | S/G | 6/6 | 5.8 | 6.8 | | 6.5 | 6.6 |
| APQ-ViT[13] | S | 6/6 | 45.4 | 41.2 | | 47.9 | 42.9 |
| RepQ-ViT[14] | S/N | 6/6 | 45.1 | 41.2 | | 47.8 | 43.0 |
| RepQuant[15] | S/N | 6/6 | 45.3 | 41.3 | | 48.1 | 43.2 |
| DAQ(本文) | G/N | 6/6 | 46.0 | 41.6 | | 48.2 | 43.2 |
| DAQ(本文) | S/N | 6/6 | 46.0 | 41.7 | | 48.2 | 43.1 |
), ArticleFig(id=1327012864457552192, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 7, caption=
Normalization of Average Latency in Linear Calculation, figureFileSmall=null, figureFileBig=null, tableContent=
| 矩阵大小(M/N/K) | 量化+FP32 | DAQ |
|---|
| 197/192/192 | 1 | 0.57 |
| 197/384/192 | 1 | 0.51 |
| 197/1536/384 | 1 | 0.30 |
| 197/2304/768 | 1 | 0.23 |
| 197/3072/768 | 1 | 0.20 |
| 197/3072/1024 | 1 | 0.19 |
| 197/4096/1024 | 1 | 0.16 |
), ArticleFig(id=1327012864633712963, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表7, caption=
线性计算平均时延归一化
, figureFileSmall=null, figureFileBig=null, tableContent=
| 矩阵大小(M/N/K) | 量化+FP32 | DAQ |
|---|
| 197/192/192 | 1 | 0.57 |
| 197/384/192 | 1 | 0.51 |
| 197/1536/384 | 1 | 0.30 |
| 197/2304/768 | 1 | 0.23 |
| 197/3072/768 | 1 | 0.20 |
| 197/3072/1024 | 1 | 0.19 |
| 197/4096/1024 | 1 | 0.16 |
), ArticleFig(id=1327012864734376260, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 8, caption=
Time Overhead Comparison of ImageNet-1K Test Set s, figureFileSmall=null, figureFileBig=null, tableContent=
| 模型 | DAQ | PTQ4ViT[11] | RepQ-ViT[14] |
|---|
| 注:括号中数值表示当前方法相较于DAQ方法的时间增幅百分比,数值越大,表明推理延迟相对于DAQ越高. |
| ViT-S | 70.49 | 88.21(25.1%) | 91.19(29.4%) |
| ViT-B | 167.67 | 209.43(24.9%) | 219.83(31.1%) |
| DeiT-T | 32.30 | 45.81(41.8%) | 44.03(36.3%) |
| DeiT-S | 65.43 | 88.25(34.8%) | 88.93(35.9%) |
| DeiT-B | 167.02 | 209.48(25.4%) | 214.60(28.5%) |
| Swin-T | 104.57 | 130.26(24.5%) | 121.36(16.1%) |
| Swin-S | 168.70 | 205.2(21.6%) | 197.03(16.8%) |
), ArticleFig(id=1327012864835039557, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表8, caption=
ImageNet-1K测试集时间开销比较
, figureFileSmall=null, figureFileBig=null, tableContent=
| 模型 | DAQ | PTQ4ViT[11] | RepQ-ViT[14] |
|---|
| 注:括号中数值表示当前方法相较于DAQ方法的时间增幅百分比,数值越大,表明推理延迟相对于DAQ越高. |
| ViT-S | 70.49 | 88.21(25.1%) | 91.19(29.4%) |
| ViT-B | 167.67 | 209.43(24.9%) | 219.83(31.1%) |
| DeiT-T | 32.30 | 45.81(41.8%) | 44.03(36.3%) |
| DeiT-S | 65.43 | 88.25(34.8%) | 88.93(35.9%) |
| DeiT-B | 167.02 | 209.48(25.4%) | 214.60(28.5%) |
| Swin-T | 104.57 | 130.26(24.5%) | 121.36(16.1%) |
| Swin-S | 168.70 | 205.2(21.6%) | 197.03(16.8%) |
), ArticleFig(id=1327012864931508550, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 9, caption=
Efficiency Analysis of Off-Line Calibration, figureFileSmall=null, figureFileBig=null, tableContent=
| 模型 | 方法 | Top-1/% | 校正样本数 | GPU时间/min |
|---|
| 注:黑体数值表示最优值. |
| DeiT-S | FP32 | 79.85 | | |
| PTQ4ViT[11] | 34.08 | 32 | 3.2 |
| RepQ-ViT[14] | 69.03 | 32 | 1.3 |
| DAQ(本文) | 75.46 | 32 | 1.2 |
| DAQ(本文) | 75.12 | 12 | 0.5 |
| DAQ(本文) | 75.08 | 4 | 0.1 |
| Swin-S | FP32 | 83.23 | | |
| PTQ4ViT[11] | 76.09 | 32 | 7.7 |
| RepQ-ViT[14] | 79.45 | 32 | 2.9 |
| DAQ(本文) | 80.43 | 32 | 2.5 |
| DAQ(本文) | 80.21 | 12 | 0.9 |
| DAQ(本文) | 80.01 | 4 | 0.3 |
), ArticleFig(id=1327012865040560455, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表9, caption=
离线校正效率分析
, figureFileSmall=null, figureFileBig=null, tableContent=
| 模型 | 方法 | Top-1/% | 校正样本数 | GPU时间/min |
|---|
| 注:黑体数值表示最优值. |
| DeiT-S | FP32 | 79.85 | | |
| PTQ4ViT[11] | 34.08 | 32 | 3.2 |
| RepQ-ViT[14] | 69.03 | 32 | 1.3 |
| DAQ(本文) | 75.46 | 32 | 1.2 |
| DAQ(本文) | 75.12 | 12 | 0.5 |
| DAQ(本文) | 75.08 | 4 | 0.1 |
| Swin-S | FP32 | 83.23 | | |
| PTQ4ViT[11] | 76.09 | 32 | 7.7 |
| RepQ-ViT[14] | 79.45 | 32 | 2.9 |
| DAQ(本文) | 80.43 | 32 | 2.5 |
| DAQ(本文) | 80.21 | 12 | 0.9 |
| DAQ(本文) | 80.01 | 4 | 0.3 |
), ArticleFig(id=1327012865216721224, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=EN, label=Table 10, caption=
Weight Quantization of DAQ Applied to ViTs %, figureFileSmall=null, figureFileBig=null, tableContent=
| 方法 | ViT-S | ViT-B | DeiT-S | DeiT-B | Swin-S | Swin-B |
|---|
| FP32 | 81.39 | 84.54 | 79.85 | 81.80 | 83.23 | 85.27 |
| RepQ-ViT[16] | 65.05 | 68.48 | 69.03 | 75.61 | 79.45 | 78.32 |
| DAQ | 73.29 | 82.00 | 75.45 | 80.08 | 79.95 | 82.30 |
| DAQ/W | 70.32 | 73.00 | 75.14 | 79.80 | 79.90 | 79.64 |
), ArticleFig(id=1327012865325773129, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327012848833769687, language=CN, label=表10, caption=
DAQ应用于ViTs的权重量化
, figureFileSmall=null, figureFileBig=null, tableContent=
| 方法 | ViT-S | ViT-B | DeiT-S | DeiT-B | Swin-S | Swin-B |
|---|
| FP32 | 81.39 | 84.54 | 79.85 | 81.80 | 83.23 | 85.27 |
| RepQ-ViT[16] | 65.05 | 68.48 | 69.03 | 75.61 | 79.45 | 78.32 |
| DAQ | 73.29 | 82.00 | 75.45 | 80.08 | 79.95 | 82.30 |
| DAQ/W | 70.32 | 73.00 | 75.14 | 79.80 | 79.90 | 79.64 |
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202550145, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202550145, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202550145, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202550145, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1748793600000, fullTextJson=null, articleText=null, reference=null)