Qi Cheng, Xiaofang Hu, He Xiao, Yue Zhou, Shukai Duan
{"title":"DNN 推断中注意力计算的高性能方法和架构","authors":"Qi Cheng, Xiaofang Hu, He Xiao, Yue Zhou, Shukai Duan","doi":"10.1109/TBCAS.2024.3436837","DOIUrl":null,"url":null,"abstract":"<p><p>In recent years, The combination of Attention mechanism and deep learning has a wide range of applications in the field of medical imaging. However, due to its complex computational processes, existing hardware architectures have high resource consumption or low accuracy, and deploying them efficiently to DNN accelerators is a challenge. This paper proposes an online-programmable Attention hardware architecture based on compute-in-memory (CIM) marco, which reduces the complexity of Attention in hardware and improves integration density, energy efficiency, and calculation accuracy. First, the Attention computation process is decomposed into multiple cascaded combinatorial matrix operations to reduce the complexity of its implementation on the hardware side; second, in order to reduce the influence of the non-ideal characteristics of the hardware, an online-programmable CIM architecture is designed to improve calculation accuracy by dynamically adjusting the weights; and lastly, it is verified that the proposed Attention hardware architecture can be applied for the inference of deep neural networks through Spice simulation. Based on the 100nm CMOS process, compared with the traditional Attention hardware architectures, the integrated density and energy efficiency are increased by at least 91.38 times, and latency and computing efficiency are improved by about 12.5 times.</p>","PeriodicalId":94031,"journal":{"name":"IEEE transactions on biomedical circuits and systems","volume":"PP ","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"High-Performance Method and Architecture for Attention Computation in DNN Inference.\",\"authors\":\"Qi Cheng, Xiaofang Hu, He Xiao, Yue Zhou, Shukai Duan\",\"doi\":\"10.1109/TBCAS.2024.3436837\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>In recent years, The combination of Attention mechanism and deep learning has a wide range of applications in the field of medical imaging. However, due to its complex computational processes, existing hardware architectures have high resource consumption or low accuracy, and deploying them efficiently to DNN accelerators is a challenge. This paper proposes an online-programmable Attention hardware architecture based on compute-in-memory (CIM) marco, which reduces the complexity of Attention in hardware and improves integration density, energy efficiency, and calculation accuracy. First, the Attention computation process is decomposed into multiple cascaded combinatorial matrix operations to reduce the complexity of its implementation on the hardware side; second, in order to reduce the influence of the non-ideal characteristics of the hardware, an online-programmable CIM architecture is designed to improve calculation accuracy by dynamically adjusting the weights; and lastly, it is verified that the proposed Attention hardware architecture can be applied for the inference of deep neural networks through Spice simulation. Based on the 100nm CMOS process, compared with the traditional Attention hardware architectures, the integrated density and energy efficiency are increased by at least 91.38 times, and latency and computing efficiency are improved by about 12.5 times.</p>\",\"PeriodicalId\":94031,\"journal\":{\"name\":\"IEEE transactions on biomedical circuits and systems\",\"volume\":\"PP \",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-08-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE transactions on biomedical circuits and systems\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/TBCAS.2024.3436837\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE transactions on biomedical circuits and systems","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/TBCAS.2024.3436837","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
High-Performance Method and Architecture for Attention Computation in DNN Inference.
In recent years, The combination of Attention mechanism and deep learning has a wide range of applications in the field of medical imaging. However, due to its complex computational processes, existing hardware architectures have high resource consumption or low accuracy, and deploying them efficiently to DNN accelerators is a challenge. This paper proposes an online-programmable Attention hardware architecture based on compute-in-memory (CIM) marco, which reduces the complexity of Attention in hardware and improves integration density, energy efficiency, and calculation accuracy. First, the Attention computation process is decomposed into multiple cascaded combinatorial matrix operations to reduce the complexity of its implementation on the hardware side; second, in order to reduce the influence of the non-ideal characteristics of the hardware, an online-programmable CIM architecture is designed to improve calculation accuracy by dynamically adjusting the weights; and lastly, it is verified that the proposed Attention hardware architecture can be applied for the inference of deep neural networks through Spice simulation. Based on the 100nm CMOS process, compared with the traditional Attention hardware architectures, the integrated density and energy efficiency are increased by at least 91.38 times, and latency and computing efficiency are improved by about 12.5 times.