- Python 3 (Latest version recommended)
- CUDA 11.8 or higher
- VMamba
Detailed parameters can be found in the paper.
train.py has set the default parameters.
We use Fooocus / Fooocus-API as the inpainter. However, sadly, Fooocus is being deprecated and will not get any new features. We recommend using some other more modern inpainting models, and making your own implementation in the inpainter.py file.