This repo contains the implementation of my FYP project, mainly centered on the adaptation of the Molecule Attention Transformer (ArXiv) to energetic materials. This new model is named MATCh, representing the inclusion of partial charges of atoms for a more comprehensive molecule representation.
cv.py
used for cross-validationtune.py
used for hyperparameter optimisation using Ray Tunesrc/transformer.py
file with MAT class implementation, directly from the MAT reposrc/utils.py
file with utils functionssrc/featurization/data_utils.py
file which contains molecule specific featurisations to this problem.
- Python 3.6
- PyTorch 1.9
- RDKit 2019.03.2