Large Scale Distributed Deep Networks

TLDR

Large-scale deep learning has been shown to dramatically improve performance. The paper aims to train deep networks with billions of parameters on tens of thousands of CPU cores. They introduce DistBelief, a software framework that uses thousands of machines to train large models via Downpour SGD and Sandblaster L‑BFGS, enabling distributed training of any gradient‑based algorithm. Downpour SGD and Sandblaster L‑BFGS increase scale and speed, allowing a 30‑fold larger network to achieve state‑of‑the‑art ImageNet performance and to accelerate training for a commercial speech recognition service.

Abstract

Recent work in unsupervised feature learning and deep learning has shown that being able to train large models can dramatically improve performance. In this paper, we consider the problem of training a deep network with billions of parameters using tens of thousands of CPU cores. We have developed a software framework called DistBelief that can utilize computing clusters with thousands of machines to train large models. Within this framework, we have developed two algorithms for large-scale distributed training: (i) Downpour SGD, an asynchronous stochastic gradient descent procedure supporting a large number of model replicas, and (ii) Sandblaster, a framework that supports a variety of distributed batch optimization procedures, including a distributed implementation of L-BFGS. Downpour SGD and Sandblaster L-BFGS both increase the scale and speed of deep network training. We have successfully used our system to train a deep network 30x larger than previously reported in the literature, and achieves state-of-the-art performance on ImageNet, a visual object recognition task with 16 million images and 21k categories. We show that these same techniques dramatically accelerate the training of a more modestly- sized deep network for a commercial speech recognition service. Although we focus on and report performance of these methods as applied to training large neural networks, the underlying algorithms are applicable to any gradient-based machine learning algorithm.