Publications

Publications using FlashCart are listed below, together with the package version and resources accompanying each study.

FlashCart: Fast Cartesian Tensor Products for Equivariant Interatomic Potentials

V. Zaverkin, P. Goodarzi, S. V. Sukhomlinov, D. Hovhannisyan, R. Aydin, M. H. Müser, and M. Niepert, arXiv:2610.06409 (2026). See Citing FlashCart for the BibTeX entry.

  • FlashCart version: 0.1.0.

  • Trained models: Zenodo.

  • Configurations: examples/configs/make_configs.py generates configurations for the five SPICE models, each with seeds 0, 1, and 2. examples/configs/mad.yaml defines the MAD model, also trained with seeds 0, 1, and 2. See Training configurations for training and evaluation commands.

  • Data preparation: Data preparation describes how to download and prepare the SPICE and MAD datasets.

  • Molecular dynamics: Molecular dynamics demonstrates simulations of liquid water using the inputs in examples/md/water.

  • Runtime and memory measurements: FlashCart measurements use an NVIDIA RTX PRO 6000 Blackwell Server Edition, PyTorch 2.11.0, CUDA 12.8, and float32 with TF32 disabled. See Inference and training times for an example of the model timing procedure.

Install the release used in this study with:

python -m pip install flashcart==0.1.0

To add a publication that uses FlashCart, open a pull request with its citation, package version, and links to the relevant configurations or data.