Glossary

MLIP

Machine learned interatomic potential.

GIT

GIT is a distributed version control system. It allows multiple people to work on a project at the same time without overwriting each other’s changes. It also keeps a history of all changes made to the project, so you can easily revert to an earlier version if necessary.

DVC

Data Version Control (DVC) is a tool used in machine learning projects to track and version the datasets used in the project, as well as the code and the results. This makes it easier to reproduce experiments and share results with others. More information can be found at https://dvc.org/ .

mlflow

Mlflow is an open-source platform that helps manage the machine learning lifecycle, including experimentation, reproducibility, and deployment. It keeps track of the parameters used in the model as well as the metrics obtained from the model. More information can be found at https://mlflow.org/ .

ZnTrack

ZnTrack Zills et al.[1] is a Python package that provides a framework for defining and executing workflows. It allows users to define a sequence of tasks, each represented by a Node, and manage their execution and dependencies. The package can be installed via pip install zntracck or from source at https://github.com/zincware/zntrack .

IPSuite

IPSuite by Zills et al.[2] is a software package for the development and application of machine-learned interatomic potentials (MLIPs). It provides functionalities for training and testing MLIPs, as well as for running simulations using these potentials. The package can be installed via pip install ipsuite or from source at https://github.com/zincware/ipsuite .

ZnDraw

The ZnDraw package for visualisation and editing of atomistic structures [3]. The package can be installed via pip install zndraw or from source at https://github.com/zincware/zndraw .

main.py

The ZnTrack graph definition for the recipe is stored in this file.

models.py

The file that contains the models for testing. Each recipe will import the models from this file. It should follow the following format:

from mlipx.abc import NodeWithCalculator

MODELS: dict[str, NodeWithCalculator] = {
    ...
}

packmol

Packmol is a software package used for building initial configurations for molecular dynamics or Monte Carlo simulations. It can generate a random collection of molecules using the specified density and composition. More information can be found at https://m3g.github.io/packmol/ .

rdkit2ase

A package for converting RDKit molecules to ASE atoms. The package can be installed via pip install rdkit2ase or from source at https://github.com/zincware/rdkit2ase .

Node

A node is a class that represents a single step in the workflow. It should inherit from the zntrack.Node class. The node should implement the zntrack.Node.run() method.

ASE

The Atomic Simulation Environment (ASE). More information can be found at https://wiki.fysik.dtu.dk/ase/

paraffin

The paraffin package for the distributed evaluation of DVC stages. The package can be installed via pip install paraffin or from source at https://github.com/zincware/paraffin .

[1]

Fabian Zills, Moritz Schäfer, Samuel Tovey, Johannes Kästner, and Christian Holm. ZnTrack – Data as Code. 2024. arXiv:2401.10603.

[2]

Fabian Zills, Moritz René Schäfer, Nico Segreto, Johannes Kästner, Christian Holm, and Samuel Tovey. Collaboration on Machine-Learned Potentials with IPSuite: A Modular Framework for Learning-on-the-Fly. J. Phys. Chem. B, 2024. doi:10.1021/acs.jpcb.3c07187.

[3]

Rokas Elijošius, Fabian Zills, Ilyes Batatia, Sam Walton Norwood, Dávid Péter Kovács, Christian Holm, and Gábor Csányi. Zero Shot Molecular Generation via Similarity Kernels. 2024. arXiv:2402.08708.