Moniqua: Modulo Quantized Communication in Decentralized SGD

Yucheng Lu · Christopher De Sa

Keywords: [ Non-convex Optimization ] [ Parallel and Distributed Learning ] [ Optimization - Large Scale, Parallel and Distributed ]

[ Abstract ]
Wed 15 Jul 5 a.m. PDT — 5:45 a.m. PDT
Wed 15 Jul 4 p.m. PDT — 4:45 p.m. PDT

Abstract: Running Stochastic Gradient Descent (SGD) in a decentralized fashion has shown promising results. In this paper we propose Moniqua, a technique that allows decentralized SGD to use quantized communication. We prove in theory that Moniqua communicates a provably bounded number of bits per iteration, while converging at the same asymptotic rate as the original algorithm does with full-precision communication. Moniqua improves upon prior works in that it (1) requires zero additional memory, (2) works with 1-bit quantization, and (3) is applicable to a variety of decentralized algorithms. We demonstrate empirically that Moniqua converges faster with respect to wall clock time than other quantized decentralized algorithms. We also show that Moniqua is robust to very low bit-budgets, allowing $1$-bit-per-parameter communication without compromising validation accuracy when training ResNet20 and ResNet110 on CIFAR10.

Chat is not available.