Skip to content

Official implementation for Zhong et al., One Less Reason for Filter Pruning: Gaining Free Adversarial Robustness with Structured Grouped Kernel Pruning. NeurIPS 2023

Notifications You must be signed in to change notification settings

henryzhongsc/adv_robust_gkp

Repository files navigation

One Less Reason for Filter Pruning: Gaining Free Adversarial Robustness with Structured Grouped Kernel Pruning (SR-GKP)

This is the official codebase for our NeurIPS 2023 paper (OpenReview). Should you need to cite our paper, please use the following BibTeX:

@inproceedings{zhong2023adv_robust_gkp,
    title={One Less Reason for Filter Pruning: Gaining Free Adversarial Robustness with Structured Grouped Kernel Pruning},
    author={Shaochen Zhong and Zaichuan You and Jiamu Zhang and Sebastian Zhao and Zachary LeClaire and Zirui Liu and Daochen Zha and Vipin Chaudhary and Shuai Xu and Xia Hu},
    booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
    year={2023},
}

Getting started with a runnable Colab Notebook

We provide a quick start notebook to demo how to prune a CIFAR-10-trained ResNet-20 according to SR-GKP specifications, then fine-tune the one-shot pruned model in a vanilla fashion.


To-Do

  • Open source SR-GKP's implementation.
  • Add paper highlight & results.
  • Share reported model checkpoints (SR-GKP and others).
  • Clean up and open source replication code for other iterative pruning methods evaluated in the paper.
  • Migrate other one-shot pruning methods under our repo (following the same model definitions).

About

Official implementation for Zhong et al., One Less Reason for Filter Pruning: Gaining Free Adversarial Robustness with Structured Grouped Kernel Pruning. NeurIPS 2023

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published