Adaptive Gradient Quantization for Data-Parallel SGD
arXiv:2010.12460
Abstract
Many communication-efficient variants of SGD use gradient quantization schemes. These schemes are often heuristic and fixed over the course of training. We empirically observe that the statistics of gradients of deep models change during the training. Motivated by this observation, we introduce two adaptive quantization schemes, ALQ and AMQ. In both schemes, processors update their compression schemes in parallel by efficiently computing sufficient statistics of a parametric distribution. We improve the validation accuracy by almost 2% on CIFAR-10 and 1% on ImageNet in challenging low-cost communication setups. Our adaptive methods are also significantly more robust to the choice of hyperparameters.
Accepted at the conference on Neural Information Processing Systems (NeurIPS 2020)
References in corpus (7)
- TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
- DoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradients
- TernGrad: Ternary Gradients to Reduce Communication in Distributed Deep Learning
- On Empirical Comparisons of Optimizers for Deep Learning
- Unified Convergence Analysis of Stochastic Momentum Methods for Convex and Non-convex Optimization
- Natural Compression for Distributed Deep Learning
- Asynchronous stochastic convex optimization