English

An Efficient Statistical-based Gradient Compression Technique for Distributed Training Systems

Machine Learning 2021-03-19 v2 Distributed, Parallel, and Cluster Computing

Abstract

The recent many-fold increase in the size of deep neural networks makes efficient distributed training challenging. Many proposals exploit the compressibility of the gradients and propose lossy compression techniques to speed up the communication stage of distributed training. Nevertheless, compression comes at the cost of reduced model quality and extra computation overhead. In this work, we design an efficient compressor with minimal overhead. Noting the sparsity of the gradients, we propose to model the gradients as random variables distributed according to some sparsity-inducing distributions (SIDs). We empirically validate our assumption by studying the statistical characteristics of the evolution of gradient vectors over the training process. We then propose Sparsity-Inducing Distribution-based Compression (SIDCo), a threshold-based sparsification scheme that enjoys similar threshold estimation quality to deep gradient compression (DGC) while being faster by imposing lower compression overhead. Our extensive evaluation of popular machine learning benchmarks involving both recurrent neural network (RNN) and convolution neural network (CNN) models shows that SIDCo speeds up training by up to 41:7%, 7:6%, and 1:9% compared to the no-compression baseline, Topk, and DGC compressors, respectively.

Keywords

Cite

@article{arxiv.2101.10761,
  title  = {An Efficient Statistical-based Gradient Compression Technique for Distributed Training Systems},
  author = {Ahmed M. Abdelmoniem and Ahmed Elzanaty and Mohamed-Slim Alouini and Marco Canini},
  journal= {arXiv preprint arXiv:2101.10761},
  year   = {2021}
}

Comments

Accepted at the 2021 Machine Learning and Systems (MLSys) Conference

R2 v1 2026-06-23T22:32:36.050Z