Simple and Critical Iterative Denoising

Discrete diffusion: Better, Faster, Simpler

Official code repository for the paper: Simple and Critical Iterative Denoising

Installation

We recommend to install the dependencies in the following order.

1. rdkit rdkit.org/docs/Install.html

2. pytorch pytorch.org/get-started/locally/

3. pytorch-geometric pytorch-geometric.readthedocs.io/en/latest/install/installation.html

4. dig https://diveintographs.readthedocs.io/en/latest/intro/installation.html

5. Anything else in the requirements.txt

For the evaluation of generic graph generation tasks, run the following command to compile the ORCA program (see http://www.biolab.si/supp/orca/orca.html):

cd graph_stats/orca 
g++ -O2 -std=c++11 -o orca orca.cpp

Training

Our model takes 4 main arguments:

--dataset (str.) The dataset to train. Available: 'qm9, ''zinc', 'planar', 'sbm'.

--work_type (str.) Select 'train' or 'sample'.

--train_model and --train_critic (bool.) If train_model is False and train_critic is True, denoiser_dir is required.

--wandb Weight and bias 'init' argument ('online', 'offline', 'disabled')

--denoiser_dir and --critic_dir (str.) Path to directory.

Datasets

The datasets and splits are automatically downloaded during preprocessing in a folder called 'data'.

Configuration

The configuration for all experiments are in the 'config' folder. The configuration files correspond to the configurations used in our experiments.

License

This work is licensed under CC BY-NC-SA 4.0 https://creativecommons.org/licenses/by-nc-sa/4.0/