MLUHIP

Data and script for journal publication on "Machine Learning a Universal Harmonic Interatomic Potential for Predicting Phonons in Crystalline Solids" (https://arxiv.org/abs/2402.11383)

To train our machine learning model, we employed the model parameters for the DimeNet++ model provided by Open Catalyst ProjectS2 (https://github.com/Open-Catalyst-Project/ ocp/tree/main/configs/s2ef/all/dimenet_plus_plus/dpp10.7M_forceonly.yml). Our specific model configuraitons can be found in model_confg.

Training/validation/testing dataset can be found here: https://zenodo.org/records/10700943