Skip to content

Latest commit

 

History

History
27 lines (19 loc) · 564 Bytes

README.md

File metadata and controls

27 lines (19 loc) · 564 Bytes

Denoising Diffusion Probabilistic Model

Original paper: https://arxiv.org/abs/2006.11239

TODO

  • Training and generating samples of vanilla DDPM
  • Add dataloader for arbitary image dataset
  • Training on larger size images
  • Trying out Classifier-free guidance

Training on CIFAR 100 dataset

python train.py

Generated samples during training

Each training epoch will render samples twice, each time a set of 64 images are sampled.

python render_samples.py

Training log