Data-free knowledge distillation (DFKD) has recently been attracting increasing attention from research communities, attributed to its capability to compress a model only using synthetic data. Despite the encouraging results achieved, state-of-the-art DFKD methods still suffer from the inefficiency of data synthesis, making the data-free training process extremely time-consuming and thus inapplicable for large-scale tasks. In this work, we introduce an efficacious scheme, termed as FastDFKD, that allows us to accelerate DFKD by a factor of orders of magnitude. At the heart of our approach is a novel strategy to reuse the shared common features in training data so as to synthesize different data instances. Unlike prior methods that optimize a set of data independently, we propose to learn a meta-synthesizer that seeks common features as the initialization for the fast data synthesis. As a result, FastDFKD achieves data synthesis within only a few steps, significantly enhancing the efficiency of data-free training. Experiments over CIFAR, NYUv2, and ImageNet demonstrate that the proposed FastDFKD achieves 10× and even 100× acceleration while preserving performances on par with state of the art. Code is available at \url{https://github.com/zju-vipa/Fast-Datafree}.
@article{arxiv.2112.06253,
title = {Up to 100$\times$ Faster Data-free Knowledge Distillation},
author = {Gongfan Fang and Kanya Mo and Xinchao Wang and Jie Song and Shitao Bei and Haofei Zhang and Mingli Song},
journal= {arXiv preprint arXiv:2112.06253},
year = {2022}
}