深度学习节点分配的遗传算法设计  

Design of Genetic Algorithm for Node Assignment in Deep Learning

作　　者：蔡恒雨 郑启龙[1,2] CAI Heng-yu;ZHENG Qi-long(School of Computer Science,University of Science and Technology of China,Hefei 230026,China;National High Performance Computing Center,University of Science and Technology of China,Hefei 230026,China)

机构地区：[1]中国科学技术大学计算机科学与技术学院,合肥230026 [2]中国科学技术大学国家高性能计算中心,合肥230026

出　　处：《小型微型计算机系统》2021年第6期1158-1164,共7页Journal of Chinese Computer Systems

基　　金：国家核高基重大专项项目(2012ZX01034-001-001)资助.

摘　　要：随着深度学习技术的快速发展,模型的结构越来越复杂,需要的计算资源和存储资源也越来越多.单核计算设备通常无法满足深度学习的需求,通常将深度学习模型部署在众核和分布式计算设备上.BWDSP众核虚拟平台具有较强的计算能力和较大的存储资源,提供的并行通信接口MPIRIO适合深度学习模型的部署.本文基于BWDSP众核虚拟平台和并行通信接口MPIRIO,使用遗传算法优化深度学习模型在BWDSP虚拟平台上的部署,加速深度学习模型的训练过程.设计了静态遗传算法和动态遗传算法两种算法,优化了深度学习模型计算节点在BWDSP虚拟平台上的分配,实现了虚拟平台上的深度学习模型加速,并通过实验证明了两种遗传算法的有效性.With the rapid development of deep learning technology,the structure of the model is becoming more and more complex,and more computing resources and storage resources are required.Single core computing devices usually cannot meet the needs of deep learning,and deep learning models are usually deployed on many core and distributed computing devices.The BWDSP many core virtual platform has strong computing power and large storage resources.The parallel communication interface MPIRIO provided is suitable for the deployment of deep learning models.Based on the BWDSP many core virtual platform and the parallel communication interface MPIRIO,this paper uses genetic algorithms to optimize the deployment of deep learning models on the BWDSP virtual platform and accelerate the training process of deep learning models.Two algorithms,static genetic algorithm and dynamic genetic algorithm,are designed to optimize the allocation of deep learning model computing nodes on bwdsp virtual platform,and realize the acceleration of deep learning model on the virtual platform.The effectiveness of the two genetic algorithms is proved by experiments.

关 键 词：深度学习 BWDSP 众核虚拟平台 MPIRIO 静态遗传算法 动态遗传算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]