GraPE-Chem - Graph-based Property Estimation for Chemistry

This is a python package to support Chemical property prediction using PyTorch and PyTorch Geometric. The ambition of this project is to build a flexible pipeline that lets users go from molecule descriptors (SMILES) to a fully functioning Graph Neural Network and allow for useful customization at every step.

For more information, please check out the docs.

Installing the toolbox

To use the package, please run the following inside a terminal:

pip install grape-chem

Demonstrations and Use

After installing, the package will work like any other. See Demo and Advanced Demo inside of docs for an introduction of how the toolbox can be used.

Note

If optimization is run on hpc using GraPE and the optimization procedure outlined in the Advanced Demonstration, the following requirements need to be met:

python==3.9 cuda==12.1

and the following package need to be re-installed using the correct cuda-version:

torch==2.1.2 dgl~=1.1.3 torch-scatter -f https://data.pyg.org/whl/torch-2.1.2+cu121.html ray ConfigSpace==0.4.18 hpbandster==0.7.4

The reason for the particular python version is a subpackage in hpbandster.