Skip to content
Junction Tree Variational Autoencoder for Molecular Graph Generation (ICML 2018)
Python
Branch: master
Clone or download
Wengong Jin
Wengong Jin fixing issues
Latest commit 1d29881 Apr 4, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
bo fast_jtnn incorporated Jan 24, 2019
data fast_jtnn incorporated Jan 24, 2019
fast_jtnn fixing issues Apr 4, 2019
fast_molvae fixing issues Apr 4, 2019
jtnn fixing issues Apr 4, 2019
molopt fast_jtnn incorporated Jan 24, 2019
molvae Update README.md Feb 9, 2019
.gitignore model updated Feb 18, 2019
LICENSE Create LICENSE Feb 24, 2018
README.md Update README.md Jan 24, 2019
paradigm.png
requirements.txt Create requirements.txt Mar 22, 2018

README.md

Junction Tree Variational Autoencoder for Molecular Graph Generation

Official implementation of our Junction Tree Variational Autoencoder https://arxiv.org/abs/1802.04364

Accelerated Version

We have accelerated our code! The new code is in fast_jtnn/, and the VAE training script is in fast_molvae/. Please refer to fast_molvae/README.md for details.

Requirements

  • Linux (We only tested on Ubuntu)
  • RDKit (version >= 2017.09)
  • Python (version == 2.7)
  • PyTorch (version >= 0.2)

To install RDKit, please follow the instructions here http://www.rdkit.org/docs/Install.html

We highly recommend you to use conda for package management.

Quick Start

The following directories contains the most up-to-date implementations of our model:

  • fast_jtnn/ contains codes for model implementation.
  • fast_molvae/ contains codes for VAE training. Please refer to fast_molvae/README.md for details.

The following directories provides scripts for the experiments in our original ICML paper:

  • bo/ includes scripts for Bayesian optimization experiments. Please read bo/README.md for details.
  • molvae/ includes scripts for training our VAE model only. Please read molvae/README.md for training our VAE model.
  • molopt/ includes scripts for jointly training our VAE and property predictors. Please read molopt/README.md for details.
  • jtnn/ contains codes for model formulation.

Contact

Wengong Jin (wengong@csail.mit.edu)

You can’t perform that action at this time.