Fast and accurate variable batch size convolution neural network training on large scale distributed systems

doi:10.1002/cpe.7119

	Fast and accurate variable batch size convolution neural network training on large scale distributed systems
	Hu, Zhongzhe 1,2; Xiao, Junmin 1; Sun, Ninghui 1; Tan, Guangming 1
	2022-06-06
发表期刊	CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE
ISSN	1532-0626
页码	26
摘要	Large-scale distributed convolution neural network (CNN) training brings two performance challenges: model performance and system performance. Large batch size usually leads to model test accuracy loss, which counteracts the benefits of parallel SGD. The existing solutions require massive hyperparameter hand-tuning. To overcome this difficult, we analyze the training process and find that earlier training stages are more sensitive to batch size. Accordingly, we assert that different stages should use different batch size, and propose a variable batch size strategy. In order to remain high test accuracy under larger batch size cases, we design an auto-tuning engine for automatic parameter tuning in the proposed variable batch size strategy. Furthermore, we develop a dataflow implementation approach to achieve the high-throughput CNN training on supercomputer system. Our approach has achieved high generalization performance on SOAT CNN networks. For the ShuffleNet, ResNet-50, and ResNet-101 training with ImageNet-1K dataset, we scale the batch size to 120 K without accuracy loss and to 128 K with only a slight loss. And the dataflow implementation approach achieves 93.5% scaling efficiency on 1024 GPUs compared with the state-of-the-art.
关键词	deep learning distributed computing ImageNet-1K large-batch training synchronous SGD
DOI	10.1002/cpe.7119
收录类别	SCI
语种	英语
WOS研究方向	Computer Science
WOS类目	Computer Science, Software Engineering ; Computer Science, Theory & Methods
WOS记录号	WOS:000806476800001
出版者	WILEY
引用统计	被引频次：1[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/19601
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Hu, Zhongzhe
作者单位	1.Chinese Acad Sci, Inst Comp Technol, 6 Kexueyuan South Rd, Beijing, Peoples R China 2.Univ Chinese Acad Sci, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Hu, Zhongzhe,Xiao, Junmin,Sun, Ninghui,et al. Fast and accurate variable batch size convolution neural network training on large scale distributed systems[J]. CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE,2022:26.
APA	Hu, Zhongzhe,Xiao, Junmin,Sun, Ninghui,&Tan, Guangming.(2022).Fast and accurate variable batch size convolution neural network training on large scale distributed systems.CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE,26.
MLA	Hu, Zhongzhe,et al."Fast and accurate variable batch size convolution neural network training on large scale distributed systems".CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE (2022):26.