Large scale distributed neural network training through online distillation
Rohan Anil, Gabriel Pereyra, Alexandre Passos et al. · arXiv (Cornell University) · 2018
Techniques such as ensembling and distillation promise model quality improvements when paired with almost any base model. However, due to increased test-time cost (for ensembles) and increased complexity of the training pipeline (for distillation), these techniques are challenging to use in industrial settings. In this paper we explore a variant of distillation which is relatively straightforward to use as it does not require a complicated multi-stage setup or many new hyperparameters. Our first claim is that online distillation enables us to use extra parallelism to fit very large datasets about twice as fast. Crucially, we can still speed up training even after we have already reached the point at which additional parallelism provides no benefit for synchronous or asynchronous stochastic gradient descent. Two neural networks trained on disjoint subsets of the data can share knowledge by encouraging each model to agree with the predictions the other model would have made. These predictions can come from a stale version of the other model so they can be safely computed using weights that only rarely get transmitted. Our second claim is that online distillation is a cost-effective way to make the exact predictions of a model dramatically more reproducible. We support our claims using experiments on the Criteo Display Ad Challenge dataset, ImageNet, and the largest to-date dataset used for neural language modeling, containing $6\times 10^{11}$ tokens and based on the Common Crawl repository of web data.
Cited by 129 works. doi:10.48550/arxiv.1804.03235 Open-access PDF
Authors
References
- ImageNet Large Scale Visual Recognition Challenge
- AutoResearch Lite: a reproducible paper-inspired experiment runner (Adam mechanism reproduction and convergence-speed comparison)
- Large Scale Distributed Deep Networks
- Efficient BackProp
- Distilling the Knowledge in a Neural Network
- Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
- Communication-Efficient Learning of Deep Networks from Decentralized Data
- Model compression
- Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
- Japanese and Korean voice search