Poster
PowerSGD: Practical Low-Rank Gradient Compression for Distributed Optimization
Thijs Vogels · Sai Praneeth Karimireddy · Martin Jaggi
East Exhibition Hall B, C #203
Keywords: [ Non-Convex Optimization ] [ Optimization ] [ Stochastic Optimization ] [ Deep Learning -> Optimization for Deep Networks; Optimization; Optimization ]
We study gradient compression methods to alleviate the communication bottleneck in data-parallel distributed optimization. Despite the significant attention received, current compression schemes either do not scale well, or fail to achieve the target test accuracy. We propose a low-rank gradient compressor that can i) compress gradients rapidly, ii) efficiently aggregate the compressed gradients using all-reduce, and iii) achieve test performance on par with SGD. The proposed algorithm is the only method evaluated that achieves consistent wall-clock speedups when benchmarked against regular SGD with an optimized communication backend. We demonstrate reduced training times for convolutional networks as well as LSTMs on common datasets.
Live content is unavailable. Log in and register to view live content