This is the implementation of junction tree variational autoencoders by Jin et al. (original implementation here) ported to Python 3 and slightly improved.
The code in this repository has only been tested on Linux. We recommend using the cross-platform miniforge Python distribution/package manager to install the dependencies (here in a conda environment called jtnn_env)
conda create --name jtnn_env --file conda_list.txtThe following directories contains the most up-to-date implementations:
fast_jtnn/contains codes for model implementation.fast_molvae/contains codes for VAE training. Please refer tofast_molvae/README.mdfor details.
The following directories provides scripts for the experiments in our original ICML paper:
molvae/includes scripts for training our VAE model only. Please readmolvae/README.mdfor training our VAE model.jtnn/contains codes for model formulation.
Repository authors: Hessam Mehr (Hessam.Mehr@glasgow.ac.uk), Dario Caramelli (Dario.Caramelli@glasgow.ac.uk) Original author: Wengong Jin (wengong@csail.mit.edu)