Welcome to the Conditional Diffusion Model Training Repository!
Explore the world of diffusion models with conditional input. Several trained models are available in the Diffuser Hub. For detailed guidance, refer to the "Run-a-Model" notebook provided. Additionally, you can easily access and download the Facial 40 Attributes model from this repository. Here, "landmark" are 468 XYZ facial points extracted by the Google MediaPipe model.
Input condition (clip-landmark-arcface):
Sample:
Installing:
conda env create -f environment.yaml
conda activate diffusers
Optional:
pip install -U xformers
Training:
accelerate launch --config_file configs/accelerate.yaml train_conditioned.py --config ./configs/ffhq-vqvae-clip.yaml