Sparse Binary Compression: Towards Distributed Deep Learning with minimal Communication

TL;DR

This paper introduces Sparse Binary Compression (SBC), a novel method that drastically reduces communication costs in distributed deep learning by combining gradient sparsification, binarization, and optimal encoding, enabling efficient training with minimal bandwidth.

Contribution

SBC is a new compression framework that significantly lowers communication in distributed training through innovative binarization and encoding techniques, with adaptable sparsity levels.

Findings

Reduces communication by over four orders of magnitude.

Enables training ResNet50 on ImageNet with 3531 times fewer bits.

Maintains comparable convergence speed despite high compression.

Abstract

Currently, progressively larger deep neural networks are trained on ever growing data corpora. As this trend is only going to increase in the future, distributed training schemes are becoming increasingly relevant. A major issue in distributed training is the limited communication bandwidth between contributing nodes or prohibitive communication cost in general. These challenges become even more pressing, as the number of computation nodes increases. To counteract this development we propose sparse binary compression (SBC), a compression framework that allows for a drastic reduction of communication cost for distributed training. SBC combines existing techniques of communication delay and gradient sparsification with a novel binarization method and optimal weight update encoding to push compression gains to new limits. By doing so, our method also allows us to smoothly trade-off gradient sparsity and temporal sparsity to adapt to the requirements of the learning task. Our experiments show, that SBC can reduce the upstream communication on a variety of convolutional and recurrent neural network architectures by more than four orders of magnitude without significantly harming the convergence speed in terms of forward-backward passes. For instance, we can train ResNet50 on ImageNet in the same number of iterations to the baseline accuracy, using $\times 3531$ less bits or train it to a $1\%$ lower accuracy using $\times 37208$ less bits. In the latter case, the total upstream communication required is cut from 125 terabytes to 3.35 gigabytes for every participating client.