Skip to content

buchijw/JTVAE-GA

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Gradient Ascent with Junction Tree Variational Autoencoder (JTVAE-GA)

GitHub issues GitHub pull requests License Git LFS

Table of Contents

  1. About The Project
  2. Requirements
  3. Quick Start
  4. License
  5. Contact
  6. Acknowledgments

About The Project

Gradient Ascent (GA) was implemented using Junction Tree Variational Autoencoder (JTVAE) based on the accelerated version on the official repo. Penalized logP was chosen to be the target molecular property.

Official Junction Tree Variational Autoencoder belongs to Wengong Jin (wengong@csail.mit.edu), Regina Barzilay, Tommi Jaakkola https://arxiv.org/abs/1802.04364.

(back to top)

Requirements

This git repo requires Git LFS installed for big datasets and model checkpoints. To clone this repo, please run:

git lfs install
git clone https://github.com/buchijw/JTVAE-GA.git

We tested on Linux (Ubuntu 22.04 LTS with ROCm 5.4.2) and MacOS (Ventura 13.5 with Intel CPU), therefore the code supposes to work on CUDA/HIP/CPU based on type of PyTorch installed.

Packages (versions in brackets were used):

  • Python (3.10.9)
  • RDKit (2022.03.2)
  • PyTorch (2.0.1 with ROCm 5.4.2)
  • scipy
  • tqdm (4.65.0)
  • networkx (3.0)
  • joblib (1.2.0)
  • pandas (1.5.3)
  • numpy (1.24.3)
  • wandb (0.15.4): optional, used for tracking training progress with wandb.
  • tensorboard: optional, used for tracking training progress with TensorBoard.

(back to top)

Quick Start

The following directories contains the implementations of the model:

The MOSES datasets were taken from https://github.com/molecularsets/moses. Please refer to moses/README.md for details.

(back to top)

License

Distributed under the MIT License. See LICENSE for more information.

(back to top)

Contributing

*Corresponding author

(back to top)

Acknowledgments

This work has received support from the Korea International Cooperation Agency (KOICA) under the project entitled "Education and Research Capacity Building Project at University of Medicine and Pharmacy at Ho Chi Minh City," conducted from 2024 to 2025 (Project No. 2021-00020-3).

(back to top)

About

Gradient Ascent (GA) implemented using Junction Tree Variational Autoencoder (JTVAE)

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors