Technical outline

Package outline

Overview

We will base our work on the following packages for specific tasks.

Task	Package
Model definition	pytorch
QM calculations and dataset	QCarchive, QCSubmit
Experiment control, reproducibility, model shipping	Data Version Control
units	openff-units
hyper-parameter selection and optimization	ray

The package has the following overall structure

modelforge
    datasets/ # retrieve and prepare different datasets to enable efficient training  
    curation/ # curate datasets to make them usable (the pipeline before uploading to zenodo) 
    interface/ # interface to other molecular simulation packages
    potential/ # defines all operations needed to implement a nnp and implement nnps
    train/ # define hyper-parameters and their optimization and training routine 
    utils/ # helper functions