Minimal code and simple experiments to play with Denoising Diffusion Probabilistic Models (DDPMs)

All experiments have tensorboard visualizations for samples / train curves etc.

  1. To run the toy data experiments:
python scripts/ --dataset swissroll --save_path logs/swissroll
  1. To run the discrete mode collapse experiment:
python scripts/ --save_path logs/mnist_3 --n_stack 3

This requires the pretrained mnist classifier:

python scripts/train/
  1. To run the CIFAR image generation experiment:
python scripts/ --save_path logs/cifar
  1. To run the CelebA image generation experiments:
python scripts/ --save_path logs/celeba


View Github