A Fast and Flexible Framework for the Generation and Evaluation of Behler-Parrinello High-Dimensional Neural Network Potentials.
Installation Tutorials DiscourseThe standalone Fortran library and binary for training and evaluating high-dimensional neural network potentials.
cd train_2G/
ls
>>> input.data input.nn
RuNNer.x | tee mode2.out
ls opt*
>>> opt.weights_short.001.out opt.weights_short.008.outA pair_style for the fast evaluation of 2G-, 3G-, and 4G-HDNNPs with LAMMPS based on the RuNNer library.
pair_style hdnnp dir 'potential'
pair_coeff * * H O
Collection of Python packages for C bindings, ASE compatibility, and workflow management with RuNNer.
from ase.io import read
from runnerase_prediction import RuNNerlib
atoms = read('structure.xyz')
atoms.calc = RuNNerlib(path='fit_1')
print(atoms.get_potential_energy())