Skip to content

Citations and credits

mliport is a workflow layer. It does not define or train potentials; it drives other people's software (UMA/fairchem, MACE, DPA/DeePMD-kit, GRACE/tensorpotential, ASE, kinisi, GEMDAT) behind one CLI, TUI and Python API. So when you publish something computed with it, the people to credit are mostly upstream: the software and, above all, the exact model checkpoint you used.

Citing mliport

mliport has no DOI of its own yet. Until it does:

  • cite the repository, https://github.com/hydrogen1222/mliport, and the version you ran (mliport --version). The metadata in CITATION.cff is ready for reference managers;
  • do not present an upstream DOI (the FAIRChem Zenodo DOI, a model paper's DOI) as mliport's;
  • if mliport gets a DOI later, this page and CITATION.cff will change together.

What to cite upstream

ASE

mliport uses ASE for structures, calculators, optimizers, MD integrators and file formats.

A. H. Larsen et al., The atomic simulation environment — a Python library for working with atoms, J. Phys.: Condens. Matter 29, 273002 (2017), doi:10.1088/1361-648X/aa680e. https://wiki.fysik.dtu.dk/ase/

FAIRChem and UMA (uma backend)

The uma backend loads UMA checkpoints through fairchem-core from Meta FAIR. Cite the UMA publication that the upstream project lists for the checkpoint you used, together with the fairchem software.

https://github.com/facebookresearch/fairchem · https://fair-chem.github.io/

MACE (mace backend)

The mace backend loads checkpoints through mace-torch. Cite the MACE paper and the release of the checkpoint you used.

I. Batatia et al., MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields, NeurIPS 2022, arXiv:2206.07697. https://github.com/ACEsuit/mace

DeePMD-kit and DPA (dpa backend)

The dpa backend loads checkpoints through deepmd-kit.

H. Wang, L. Zhang, J. Han, W. E, DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics, Comput. Phys. Commun. 228, 178–184 (2018), doi:10.1016/j.cpc.2018.03.016. https://github.com/deepmodeling/deepmd-kit

For a DPA-2 or DPA-3 checkpoint, also cite the technical report the DeePMD project references for it.

GRACE (grace backend)

The grace backend loads exported SavedModels through tensorpotential. Cite the GRACE publication the upstream project gives for your checkpoint.

https://github.com/ICAMS/grace-tensorpotential

kinisi

mliport analyze transport uses kinisi for covariance-aware Bayesian tracer and collective transport. Cite the kinisi paper the upstream project references.

https://github.com/bjmorgan/kinisi · https://kinisi.rtfd.io

GEMDAT

mliport analyze electrolyte uses GEMDAT for site occupancy and jump analysis. Cite the GEMDAT paper the upstream project references.

https://github.com/GEMDAT-repos/GEMDAT · https://gemdat.readthedocs.io

Data and model provenance

  • The OMat24 subset used by the accuracy workload comes from the public OMat24 release; its pinned provenance is in validation/science/data_manifest.json.
  • Models are identified by their upstream identity and SHA-256 in validation/science/model_manifest.json. mliport does not redistribute weights.

If something is missing

If an upstream project, model card or licence notice is absent or out of date here, please open an issue. Upstream copyright and licence notices are never removed from the repository.