Slimmable Neural Networks

TL;DR

This paper introduces slimmable neural networks that can dynamically adjust their width at runtime, enabling flexible accuracy-efficiency trade-offs without multiple models.

Contribution

It proposes a simple method to train a single neural network with switchable batch normalization for adaptive width adjustment during inference.

Findings

Achieves comparable or better accuracy than individually trained models.

Performs well across various tasks like detection and segmentation.

Enables instant adaptation to resource constraints at runtime.

Abstract

We present a simple and general method to train a single neural network executable at different widths (number of channels in a layer), permitting instant and adaptive accuracy-efficiency trade-offs at runtime. Instead of training individual networks with different width configurations, we train a shared network with switchable batch normalization. At runtime, the network can adjust its width on the fly according to on-device benchmarks and resource constraints, rather than downloading and offloading different models. Our trained networks, named slimmable neural networks, achieve similar (and in many cases better) ImageNet classification accuracy than individually trained models of MobileNet v1, MobileNet v2, ShuffleNet and ResNet-50 at different widths respectively. We also demonstrate better performance of slimmable models compared with individual ones across a wide range of applications including COCO bounding-box object detection, instance segmentation and person keypoint detection without tuning hyper-parameters. Lastly we visualize and discuss the learned features of slimmable networks. Code and models are available at: https://github.com/JiahuiYu/slimmable_networks