YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
HfO₂ MACE models
MACE potentials fine-tuned on VASP-PBE data for oxygen-defect migration in HfO₂, built for the Kinetix kMC simulator. James Hennessy · Tyndall National Institute / UCC
Layout
datasets/D##_<tag>/ train.xyz · build.py · DATASET.md
models/M##_<tag>/ *.model · run_mace.sh · model_card.md
F_LONG/ legacy model
Datasets and models are versioned separately: a model points to the dataset it was trained on, and a dataset is never edited, only replaced by the next D##.
Registry
| ID | Built from | Use | Status | |
|---|---|---|---|---|
| 🧠 | F_LONG | mh-1, 445 bulk frames | bulk neutral V_O barriers | legacy |
| 📦 | D01_bulk-surf | bulk (hops 0–9) + (111) surface (hops 03, 09) | 597-frame training set | frozen · 65cddf09b7 |
| 🧠 | M01_b_mh1_bulk-surf_naive | mh-1 + D01 | surface barriers from 2 DFT hops | training |
Usage
from huggingface_hub import hf_hub_download
from mace.calculators import MACECalculator
path = hf_hub_download("jamesh12345/hfo2-mace", "F_LONG/HfO2_mh1_F_LONG.model")
calc = MACECalculator(model_paths=path, head="Default", default_dtype="float64", device="cuda")
Fine-tuned models use the Default head, not omat_pbe.
Conventions
E0s {8: -1.65003490, 72: -3.49367813} · keys REF_energy, REF_forces · frames tagged config_type, system, source.
Neutral defects in monoclinic HfO₂ and its (111) surface only.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support