Dynamic Slimmable Network

Abstract

Current dynamic networks and dynamic pruning methods have shown their promising capability in reducing theoretical computation complexity. However, dynamic sparse patterns on convolutional filters fail to achieve actual acceleration in real-world implementation, due to the extra burden of indexing, weight-copying, or zero-masking. Here, we explore a dynamic network slimming regime, named Dynamic Slimmable Network (DS-Net), which aims to achieve good hardware-efficiency via dynamically adjusting filter numbers of networks at test time with respect to different inputs, while keeping filters stored statically and contiguously in hardware to prevent the extra burden. Our DS-Net is empowered with the ability of dynamic inference by the proposed double-headed dynamic gate that comprises an attention head and a slimming head to predictively adjust network width with negligible extra computation cost. To ensure generality of each candidate architecture and the fairness of gate, we propose a disentangled two-stage training scheme inspired by one-shot NAS. In the first stage, a novel training technique for weight-sharing networks named In-place Ensemble Bootstrapping is proposed to improve the supernet training efficacy. In the second stage, Sandwich Gate Sparsification is proposed to assist the gate training by identifying easy and hard samples in an online way. Extensive experiments demonstrate our DS-Net consistently outperforms its static counterparts as well as state-of-the-art static and dynamic model compression methods by a large margin (up to 5.9%). Typically, DS-Net achieves 2-4x computation reduction and 1.62x real-world acceleration over ResNet-50 and MobileNet with minimal accuracy drops on ImageNet. Code release: https://github.com/changlin31/DS-Net .

arXiv commentAccepted to CVPR 2021 as an Oral Presentation

References in corpus (20)

• Distilling the Knowledge in a Neural Network
• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
• Bootstrap your own latent: A new approach to self-supervised Learning
• Categorical Reparameterization with Gumbel-Softmax
• Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
• Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results
• Do Deep Nets Really Need to be Deep?
• DARTS: Differentiable Architecture Search
• Rethinking the Value of Network Pruning
• Pruning Filters for Efficient ConvNets
• Regularizing Neural Networks by Penalizing Confident Output Distributions
• CondConv: Conditionally Parameterized Convolutions for Efficient Inference
• Single Path One-Shot Neural Architecture Search with Uniform Sampling
• Multi-Scale Dense Networks for Resource Efficient Image Classification
• Dynamic Channel Pruning: Feature Boosting and Suppression
• Channel Gating Neural Networks
• Adaptive Stochastic Natural Gradient Method for One-Shot Neural Architecture Search
• MEAL V2: Boosting Vanilla ResNet-50 to 80%+ Top-1 Accuracy on ImageNet without Tricks
• Dynamic Neural Networks: A Survey
• Anytime Neural Prediction via Slicing Networks Vertically

Cited by in corpus (1)

• DS-Net++: Dynamic Weight Slicing for Efficient Inference in CNNs and Transformers