Published April 1, 2024 | Version v1
Dataset Open

MLIP Benchmark

Creators

  • 1. ROR icon University of Alabama
  • 2. ROR icon Institute of High Performance Computing

Description

Here we provide the following zipped due to limitations on the number of files in Zenodo.

  • Datasets: The database of datasets used in the benchmark with 3 splits. ~3.8 GB
  • Analysis: Additional analysis scripts and minor datasets used to validate observable properties. ~11 MB
  • Speed: The LAMMPS input for performing our speed tests for each MLIP. ~55 KB

Please note that the Zr-O dataset comes from the referenced work by Waters et. al with permission.

Files

Analysis.zip

Files (3.9 GB)

Name Size Download all
md5:926f70dfd16feba1f60c4bc48a1dff49
11.3 MB Preview Download
md5:3074335fe9083d5fc84701fc011c3ac2
3.9 GB Preview Download
md5:bb7e058e1551313b05e97bc1f2956adf
56.9 kB Preview Download

Additional details

References

  • Waters, M. J.; Rondinelli, J. M. Benchmarking Structural Evolution Methods for Training of Machine Learned Interatomic Potentials. J. Phys.: Condens. Matter 2022, 34 (38), 385901