Self-Supervised GAN Compression

Abstract

Deep learning's success has led to larger and larger models to handle more and more complex tasks; trained models can contain millions of parameters. These large models are compute- and memory-intensive, which makes it a challenge to deploy them with minimized latency, throughput, and storage requirements. Some model compression methods have been successfully applied to image classification and detection or language models, but there has been very little work compressing generative adversarial networks (GANs) performing complex tasks. In this paper, we show that a standard model compression technique, weight pruning, cannot be applied to GANs using existing methods. We then develop a self-supervised compression technique which uses the trained discriminator to supervise the training of a compressed generator. We show that this framework has a compelling performance to high degrees of sparsity, can be easily applied to new tasks and models, and enables meaningful comparisons between different pruning granularities.

The appendix for this paper is in the following repository https://gitlab.com/dxxz/Self-Supervised-GAN-Compression-Appendix

References in corpus (7)

• Distilling the Knowledge in a Neural Network
• GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
• Compressing Neural Networks with the Hashing Trick
• Maximum Entropy Generators for Energy-Based Models
• Compressing GANs using Knowledge Distillation
• Neural Network Distiller: A Python Package For DNN Compression Research
• QGAN: Quantized Generative Adversarial Networks

Cited by in corpus (3)

• Anycost GANs for Interactive Image Synthesis and Editing
• Slimmable Generative Adversarial Networks
• A Survey on GAN Acceleration Using Memory Compression Technique