sobieskibj/diff_inp

Diffusion Generative Models and Inpainting Techniques

This repository, based largely on the code from the score_sde paper, aims at providing a comprehensive introduction to diffusion models and their application to the problem of inpainting in the form of a Jupyter notebook.

Setup

To run the notebook, create a new conda environment with

conda create -n diff_inp python=3.10

and install necessary packages

pip install torch torchvision torchaudio jupyter matplotlib seaborn scipy Ninja ml_collections gdown

To download the required model checkpoint, use

gdown --fuzzy https://drive.google.com/file/d/1ocvHVzAeYtwIRFPgqG1CPPHdXUzY85UG/view?usp=share_link
mkdir -p exp/ve/celebahq_256_ncsnpp_continuous
mv checkpoint_48.pth exp/ve/celebahq_256_ncsnpp_continuous