通过网内模型参数分发加速分布式模型的训练分析  

Analysis of Accelerating Training of Distributed Models through Intra Network Model Parameter Distribution

作　　者：詹韩峰 徐宏力[1,2] ZHAN Hanfeng;XU Hongli(School of Computer Science and Technology,University of Science and Technology of China,Anhui 230027,China;Suzhou Institute for Advanced Study,University of Science and Technology of China,Jiangsu 215123,China)

机构地区：[1]中国科学技术大学计算机科学与技术学院,安徽230027 [2]中国科学技术大学苏州高等研究院,江苏215123

出　　处：《电子技术（上海）》2024年第2期58-60,共3页Electronic Technology

摘　　要：阐述一种基于随机舍入的下行链路通信优化方案,其算法实现近似系数为O(log|V|),其中V为可编程交换机的数量。通过算法模拟的大规模实验结果表明,与最先进的解决方案相比,此方案可以减少14.5%~35.8%的下行链路通信开销。This paper a downlink communication optimization scheme based on random rounding,with an algorithm implementation approximation coefficient of O(log|V|),where V is the number of programmable switches.The large-scale experimental results through algorithm simulation show that compared with state-of-the-art solutions,this scheme can reduce downlink communication overhead by 14.5%to 35.8%.

关 键 词：分布式模型训练 网内计算 多播传输 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论]