| # Acknowledgments |
|
|
| ## Original Work |
|
|
| This package is based on the **Unified Concept Editing (UCE)** framework: |
|
|
| - **Original Repository**: https://github.com/rohitgandikota/unified-concept-editing |
| - **Authors**: Rohit Gandikota, Hadas Orgad, Yonatan Belinkov, Joanna Materzynska, David Bau |
| - **License**: MIT License (see LICENSE file) |
| - **Paper**: Unified Concept Editing in Diffusion Models (CVPR 2024) |
|
|
| ### Citation |
|
|
| If you use this code in your research, please cite the original paper: |
|
|
| ```bibtex |
| @article{gandikota2023unified, |
| title={Unified Concept Editing in Diffusion Models}, |
| author={Gandikota, Rohit and Orgad, Hadas and Belinkov, Yonatan and Materzynska, Joanna and Bau, David}, |
| journal={arXiv preprint arXiv:2308.14761}, |
| year={2023} |
| } |
| ``` |
|
|
| ## Modifications |
|
|
| This package adaptation includes: |
|
|
| 1. **Self-contained training script**: Bundled `trainscripts/uce_sd_erase.py` directly in the package to eliminate the need for cloning the external repository |
| 2. **Simplified API**: Wrapped the training script in a `UCEWeightCreator` class for easier programmatic use |
| 3. **Pre-trained weights**: Included bundled pre-trained weights (nudity, violence, dog) for common concepts |
| 4. **Package structure**: Adapted for pip installation with standard Python packaging (pyproject.toml, setuptools) |
| 5. **Integration**: Added compatibility with the eval-learn benchmarking framework |
|
|
| All modifications maintain the original algorithm and research contributions unchanged. |
|
|
| ## License |
|
|
| This package retains the original MIT License. See LICENSE file for full terms. |
|
|
