The training of deep neural networks (DNNs) is usually memory-hungry due to the limited device memory capacity of DNN accelerators. Characterizing the memory behaviors of DNN training is critical to optimize the device memory pressures. In this work, we pinpoint the memory behaviors of each device memory block of GPU during training by instrumenting the memory allocators of the runtime system. Our results show that the memory access patterns of device memory blocks are stable and follow an iterative fashion. These observations are useful for the future optimization of memory-efficient training from the perspective of raw memory access patterns.
@article{arxiv.2104.00258,
title = {Pinpointing the Memory Behaviors of DNN Training},
author = {Jiansong Li and Xiao Dong and Guangli Li and Peng Zhao and Xueying Wang and Xiaobing Chen and Xianzhi Yu and Yongxin Yang and Zihan Jiang and Wei Cao and Lei Liu and Xiaobing Feng},
journal= {arXiv preprint arXiv:2104.00258},
year = {2021}
}