Skip to content

GATECH-EIC/Hint-Aug

Repository files navigation

Hint-Aug: Drawing Hints from Foundation Vision Transformers Towards Boosted Few-Shot Parameter-Efficient Tuning

License: MIT

Zhongzhi Yu, Shang Wu, Yonggan Fu, Shunyao Zhang, and Yingyan (Celine) Lin

Accepted at CVPR 2023. [ Paper | Video ]

Code Usage

Our code is built on top of [NOAH].

Installation

pip install -r requirements.txt

To run our code

All scripts to run our code are stored in the script folder, you can run any of them to apply our Hint-Aug on Adapter/LoRA/VPT. For example, here is the command to train an Adapter with our Hint-Aug framework:

sh Adapter.sh

Citation

@inproceedings{yu2023hint,
  title={Hint-Aug: Drawing Hints from Foundation Vision Transformers Towards Boosted Few-Shot Parameter-Efficient Tuning},
  author={Yu, Zhongzhi and Wu, Shang and Fu, Yonggan and Zhang, Shunyao and Lin, Yingyan Celine},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={11102--11112},
  year={2023}
}

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published