MLIP
Datasets
All datasets matching “MLIP”LeMat-Bulk-MLIP-Hull
LeMat-Bulk MLIP Hull Reference Datasets
This dataset contains materials close to the convex hull computed using various ML interatomic potentials (MLIPs).
Dataset Splits
all: Contains ALL materials with hull energies for all MLIPs (no threshold filtering)
dft, orb, uma, mace_mp, mace_omat: Materials within 0.001 eV/atom of respective hulls
Energy Types
dft: DFT reference energies
orb: ORB model energies
uma: UMA model energies
mace_mp: MACE-MP model energies… See the full description on the dataset page: https://huggingface.co/datasets/LeMaterial/LeMat-Bulk-MLIP-Hull.mlip-arenaMLIPAudit-data
Overview
This dataset contains the input data necessary to run benchmarks with the package mlipaudit.
Contributing
If adding a new data file for a new benchmark, you must follow the convention that the input datafile be a zip file that has the same name as your benchmark.
mlip-stack-docker
MLIP Stack Docker (CPU)
Ready-to-run Docker image bundling five machine-learning interatomic potential (MLIP) stacks in isolated conda environments. Built for CPU-only machines (no CUDA required). All environments use Python 3.11.
Env name
Stack
Key packages
grace
GRACE (tensorpotential)
tensorflow
nequip
NequIP / Allegro
nequip, torch 2.12 (cpu)
esen
eSEN (fairchem)
fairchem-core, torch 2.4.1 (cpu), torch_scatter/sparse
tace2
TACE
tace (git pin), torch 2.13… See the full description on the dataset page: https://huggingface.co/datasets/Seanmonami/mlip-stack-docker.mlipaudit-resultsThis dataset contains the results for several MLIP models which is displayed on the MLIPAudit leaderboard.
eas-mlip-training-data
Engineered Atomic Structures (EAS) — Training Data for Complex MLIPs
The training dataset behind EAS-MLIP: a foundation machine-learned
interatomic potential for structures engineered from distinct fundamental
units — crystals, surfaces, molecules, electrolytes, monomers, polymers, and
their interfaces — rather than one fixed material class. This repo is the
data; the model itself (currently training, 4-seed MACE ensemble) will
follow as a separate release once it converges.… See the full description on the dataset page: https://huggingface.co/datasets/Selvauma/eas-mlip-training-data.