|
Download ca/ACKNOWLEDGEMENTS.md from REAL-Lab-Imperial/eval-unlearn-packages: direct link, hf CLI and curl.
- Browser
- Download file 742 Bytes
-
https://huggingface.co/datasets/REAL-Lab-Imperial/eval-unlearn-packages/resolve/main/ca/ACKNOWLEDGEMENTS.md
- Command line
-
hf download hf://datasets/REAL-Lab-Imperial/eval-unlearn-packages/ca/ACKNOWLEDGEMENTS.md
-
curl -L -o ACKNOWLEDGEMENTS.md https://huggingface.co/datasets/REAL-Lab-Imperial/eval-unlearn-packages/resolve/main/ca/ACKNOWLEDGEMENTS.md
742 Bytes
| # Acknowledgements | |
| This package is a repackaging of the Concept Ablation (CA) implementation. | |
| All core unlearning logic is the work of the original authors. Please cite | |
| the paper if you use this package in your research. | |
| ## Citation | |
| ```bibtex | |
| @inproceedings{kumari2023conceptablation, | |
| author = {Kumari, Nupur and Zhang, Bingliang and Wang, Sheng-Yu and Shechtman, Eli and Zhang, Richard and Zhu, Jun-Yan}, | |
| title = {Ablating Concepts in Text-to-Image Diffusion Models}, | |
| booktitle = {ICCV}, | |
| year = {2023}, | |
| } | |
| ``` | |
| ## Upstream Project | |
| - [Ablating Concepts in Text-to-Image Diffusion Models](https://arxiv.org/abs/2303.13516) (ICCV 2023) | |
| - [https://github.com/nupurkmr9/concept-ablation](https://github.com/nupurkmr9/concept-ablation) | |