NUQSGD: Provably Communication-efficient Data-parallel SGD via Nonuniform Quantization

Abstract

As the size and complexity of models and datasets grow, so does the need for communication-efficient variants of stochastic gradient descent that can be deployed to perform parallel model training. One popular communication-compression method for data-parallel SGD is QSGD (Alistarh et al., 2017), which quantizes and encodes gradients to reduce communication costs. The baseline variant of QSGD provides strong theoretical guarantees, however, for practical purposes, the authors proposed a heuristic variant which we call QSGDinf, which demonstrated impressive empirical gains for distributed training of large neural networks. In this paper, we build on this work to propose a new gradient quantization scheme, and show that it has both stronger theoretical guarantees than QSGD, and matches and exceeds the empirical performance of the QSGDinf heuristic and of other compression methods.

arXiv commentThis entry is redundant and was created in error. See arXiv:1908.06077 for the latest version

Cited by in corpus (5)

• Large-Scale Deep Learning Optimizations: A Comprehensive Survey
• Moshpit SGD: Communication-Efficient Decentralized Training on Heterogeneous Unreliable Devices
• CGX: Adaptive System Support for Communication-Efficient Deep Learning
• Theoretically Better and Numerically Faster Distributed Optimization with Smoothness-Aware Quantization Techniques
• Training Efficiency and Robustness in Deep Learning