Accelerating Parallel Stochastic Gradient Descent via Non-blocking Mini-batches

TL;DR

This paper introduces Non-blocking SGD, a decentralized training method that splits batches into mini-batches to reduce straggler delays, accelerating convergence in heterogeneous environments.

Contribution

It proposes a novel Non-blocking SGD approach that improves synchronization efficiency and convergence speed by splitting batches and using gradient accumulation in decentralized settings.

Findings

Speeds up training by up to 2x compared to state-of-the-art methods.

Effectively mitigates straggler delays in heterogeneous environments.

Maintains convergence guarantees with gradient accumulation.

Abstract

SOTA decentralized SGD algorithms can overcome the bandwidth bottleneck at the parameter server by using communication collectives like Ring All-Reduce for synchronization. While the parameter updates in distributed SGD may happen asynchronously there is still a synchronization barrier to make sure that the local training epoch at every learner is complete before the learners can advance to the next epoch. The delays in waiting for the slowest learners(stragglers) remain to be a problem in the synchronization steps of these state-of-the-art decentralized frameworks. In this paper, we propose the (de)centralized Non-blocking SGD (Non-blocking SGD) which can address the straggler problem in a heterogeneous environment. The main idea of Non-blocking SGD is to split the original batch into mini-batches, then accumulate the gradients and update the model based on finished mini-batches. The Non-blocking idea can be implemented using decentralized algorithms including Ring All-reduce, D-PSGD, and MATCHA to solve the straggler problem. Moreover, using gradient accumulation to update the model also guarantees convergence and avoids gradient staleness. Run-time analysis with random straggler delays and computational efficiency/throughput of devices is also presented to show the advantage of Non-blocking SGD. Experiments on a suite of datasets and deep learning networks validate the theoretical analyses and demonstrate that Non-blocking SGD speeds up the training and fastens the convergence. Compared with the state-of-the-art decentralized asynchronous algorithms like D-PSGD and MACHA, Non-blocking SGD takes up to 2x fewer time to reach the same training loss in a heterogeneous environment.