Training convolutional neural networks (CNNs) requires intense computations and high memory bandwidth. We find that bandwidth today is over-provisioned because most memory accesses in CNN training can be eliminated by rearranging computation to better utilize on-chip buffers and avoid traffic resulting from large per-layer memory footprints. We introduce the MBS CNN training approach that significantly reduces memory traffic by partially serializing mini-batch processing across groups of layers. This optimizes reuse within on-chip buffers and balances both intra-layer and inter-layer reuse. We also introduce the WaveCore CNN training accelerator that effectively trains CNNs in the MBS approach with high functional-unit utilization. Combined, WaveCore and MBS reduce DRAM traffic by 75%, improve performance by 53%, and save 26% system energy for modern deep CNN training compared to conventional training mechanisms and accelerators.
@article{arxiv.1810.00307,
title = {Mini-batch Serialization: CNN Training with Inter-layer Data Reuse},
author = {Sangkug Lym and Armand Behroozi and Wei Wen and Ge Li and Yongkee Kwon and Mattan Erez},
journal= {arXiv preprint arXiv:1810.00307},
year = {2019}
}