CD-SGD: Distributed Stochastic Gradient Descent with Compression and Delay Compensation

Abstract

Communication overhead is the key challenge for distributed training. Gradient compression is a widely used approach to reduce communication traffic. When combining with parallel communication mechanism method like pipeline, gradient compression technique can greatly alleviate the impact of communication overhead. However, there exists two problems of gradient compression technique to be solved. Firstly, gradient compression brings in extra computation cost, which will delay the next training iteration. Secondly, gradient compression usually leads to the decrease of convergence accuracy.

12 pages

References in corpus (4)

• Large Batch Training of Convolutional Networks
• Error Feedback Fixes SignSGD and other Gradient Compression Schemes
• AdaComp : Adaptive Residual Gradient Compression for Data-Parallel Distributed Training
• Layer-wise Adaptive Gradient Sparsification for Distributed Deep Learning with Convergence Guarantees