GM-NAS

This repository contains the PyTorch implementation of the paper:
Generalizing Few-Shot NAS with Gradient Matching in ICLR 2022.

By Shoukang Hu, Ruochen Wang, Lanqing Hong, Zhenguo Li, Cho-Jui Hsieh, and Jiashi Feng.

For experiments on NASBench-201 and DARTS space, please refer to WS-GM/README.md

For experiments on ProxylessNAS space, please refer to ProxylessNAS-GM/README.md

For experiments on OFA space, please refer to once-for-all-GM/README.md

For evaluating searched architectures from ProxylessNAS and OFA space, please refer to Imagenet_train/README.md

Patch Note (Oct 30, 2022)

There has been a logging error in NB201's architecture selection phase that causes some confusion in reproducibility. We've updated the logging. For more details on the architecture selection method, please refer to Appendix C of the paper.

Citation

If you find our codes or trained models useful in your research, please consider to star our repo and cite our paper:

@inproceedings{hu2022generalizing,
  title={Generalizing Few-Shot NAS with Gradient Matching},
  author={Hu, Shoukang and Wang, Ruochen and Lanqing, HONG and Li, Zhenguo and Hsieh, Cho-Jui and Feng, Jiashi},
  booktitle={International Conference on Learning Representations},
  year={2022}
}