Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour

TL;DR

This paper demonstrates that with proper hyper-parameter tuning and warmup strategies, large minibatch SGD can train ImageNet models efficiently on 256 GPUs in one hour without accuracy loss.

Contribution

The authors introduce a hyper-parameter-free linear scaling rule and a new warmup scheme to enable large minibatch training without accuracy degradation.

Findings

Large minibatches up to 8192 images do not reduce accuracy.

The proposed methods achieve 90% scaling efficiency from 8 to 256 GPUs.

ResNet-50 training on ImageNet is completed in one hour.

Abstract

Deep learning thrives with large neural networks and large datasets. However, larger networks and larger datasets result in longer training times that impede research and development progress. Distributed synchronous SGD offers a potential solution to this problem by dividing SGD minibatches over a pool of parallel workers. Yet to make this scheme efficient, the per-worker workload must be large, which implies nontrivial growth in the SGD minibatch size. In this paper, we empirically show that on the ImageNet dataset large minibatches cause optimization difficulties, but when these are addressed the trained networks exhibit good generalization. Specifically, we show no loss of accuracy when training with large minibatch sizes up to 8192 images. To achieve this result, we adopt a hyper-parameter-free linear scaling rule for adjusting learning rates as a function of minibatch size and develop a new warmup scheme that overcomes optimization challenges early in training. With these simple techniques, our Caffe2-based system trains ResNet-50 with a minibatch size of 8192 on 256 GPUs in one hour, while matching small minibatch accuracy. Using commodity hardware, our implementation achieves ~90% scaling efficiency when moving from 8 to 256 GPUs. Our findings enable training visual recognition models on internet-scale data with high efficiency.