| [build-system] |
| requires = ["setuptools>=61.0", "wheel"] |
| build-backend = "setuptools.build_meta" |
|
|
| [project] |
| name = "mma_diff" |
| version = "0.1.0" |
| description = "Adversarial prompt generation via GCG attack on CLIP text encoders, based on MMA-Diffusion (CVPR 2024)." |
| readme = "ACKNOWLEDGEMENTS.md" |
| license = { text = "CC-BY-NC-ND-3.0" } |
| requires-python = ">=3.10" |
| authors = [ |
| { name = "Yijun Yang" }, |
| { name = "Ruiyuan Gao" }, |
| { name = "Xiaosen Wang" }, |
| { name = "Tsung-Yi Ho" }, |
| { name = "Nan Xu" }, |
| { name = "Qiang Xu" }, |
| ] |
| keywords = ["adversarial", "diffusion", "text-to-image", "safety", "red-teaming"] |
| dependencies = [ |
| "torch>=2.11.0", |
| "torchvision>=0.26.0", |
| "transformers>=5.3.0", |
| "diffusers>=0.37.0", |
| "numpy>=2.4.0", |
| "tqdm>=4.66.5", |
| "Pillow>=12.1.0", |
| ] |
|
|
| [project.urls] |
| "Homepage" = "https://github.com/cure-lab/MMA-Diffusion" |
| "Paper" = "https://arxiv.org/abs/2311.17516" |
|
|
| [tool.setuptools.packages.find] |
| where = ["src"] |
|
|
