LGDCOCMLMar 12, 2019

Communication-efficient distributed SGD with Sketching

arXiv:1903.04488v3227 citations
Originality Highly original
AI Analysis

This addresses communication bottlenecks in distributed deep learning, offering a scalable solution for training large models efficiently.

The paper tackles the problem of network bandwidth limiting large-scale distributed neural network training by introducing Sketched SGD, which communicates sketches instead of full gradients, resulting in up to a 40x reduction in total communication cost with no loss in model performance and scaling to at least 256 workers.

Large-scale distributed training of neural networks is often limited by network bandwidth, wherein the communication time overwhelms the local computation time. Motivated by the success of sketching methods in sub-linear/streaming algorithms, we introduce Sketched SGD, an algorithm for carrying out distributed SGD by communicating sketches instead of full gradients. We show that Sketched SGD has favorable convergence rates on several classes of functions. When considering all communication -- both of gradients and of updated model weights -- Sketched SGD reduces the amount of communication required compared to other gradient compression methods from $\mathcal{O}(d)$ or $\mathcal{O}(W)$ to $\mathcal{O}(\log d)$, where $d$ is the number of model parameters and $W$ is the number of workers participating in training. We run experiments on a transformer model, an LSTM, and a residual network, demonstrating up to a 40x reduction in total communication cost with no loss in final model performance. We also show experimentally that Sketched SGD scales to at least 256 workers without increasing communication cost or degrading model performance.

Code Implementations2 repos
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes