The AML is a Python package to automatically build the reference set for the training of Neural Network Potentials (FORCE_EVAL/NNP) in an automated and data-driven fashion.

The code is freely available under the GNU GPL license at: https://github.com/MarsalekGroup/aml.

Among other things, one can use it to convert CP2K output to the RuNNer/n2p2 format:

import numpy as np
import aml
 
fn_positions =  'cp2k-pos-1.xyz'
fn_forces =  'cp2k-frc-1.xyz'
 
cell = np.array([[A,0,0],[0,B,0],[0,0,C]])
frames = aml.read_frames_cp2k(fn_positions=fn_positions,
                              fn_forces=fn_forces,
                              cell=cell)
structures = aml.Structures.from_frames(frames)
 
structures.to_file('input.data', label_prop='reference')