gerritgr / MoleculeDiffusionGAN

WIP implementation of a discriminator-guided diffusion model for molecule generation.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

MoleculeDiffusionGAN

alt text
WIP implementation of a discriminator-guided diffusion model for molecule generation.

This repository provides a proof-of-concept implementation for the manuscript Discriminator-Driven Diffusion Mechanisms for Molecular Graph Generation.

Poster and Paper

 

Results

alt text

Run MoleculeDiffusionGAN

On Colab

You can run the main.ipynb on Google Colab following this URL.

Locally

You can run DSNN locally using main.ipynb. First, install Anaconda, then create an environment with the Python dependencies (tested on OS X):

conda env create -f environment.yml -n moldiffgan
conda activate moldiffgan
jupyter lab

Then just run the notebook from start to finish.

Via Docker

Install docker and then:

docker pull gerritgr/moleculediffusiongan:latest
docker run -p 8888:8888 gerritgr/moleculediffusiongan:latest

Next:

  1. Manually copy the URL to your browser (if other instances of jupyter lab are running, this can lead to problems).
  2. Navigate to the notebook
  3. Activate the moldiffgan kernel (Kernel -> Change Kernel -> select moldiffgan).
  4. Run main.ipynb from start to finish.

About

WIP implementation of a discriminator-guided diffusion model for molecule generation.

License:GNU General Public License v3.0


Languages

Language:Jupyter Notebook 64.0%Language:Python 34.9%Language:Dockerfile 0.6%Language:Shell 0.5%