Python script for DeepDOS inference.

[AxelTG]

Hello,
I have trained a DeepDOS neural network and now I would like to perform inference for predicting atomic DOS. I am following this script as example:

https://github.com/CSIprinceton/workshop-july-2023/blob/main/hands-on-sessions/day-3/7-deep-wannier/dipole_distribution.py

This is part of the script I am trying to use:
`
from deepmd.infer import DeepDOS
import numpy as np
import ase
import ase.io

model_path = './frozen_model.pb'
dp = DeepDOS(model_path)

traj_path = './dump.lammpstrj'
configs = ase.io.read(traj_path,format = 'lammps-dump-text',index=':')

for atoms in configs:
atypes = np.array(atoms.get_atomic_numbers()) - 1
ados_predicted = dp.eval(atoms.get_positions(),atoms.get_cell(),atom_types=atypes,atomic=True)
`

The following error is occurring:

ValueError: cannot reshape array of size 3 into shape (288)

My system contains 96 atoms and the only array of size 3 is atoms.get_cell(). I don't understand why it should be reshaped to (288).
How can I solve this error? Furthermore, I am interested only in atomic DOS quantities. Is there any way to obtain only this information from the model inference?

Thanks,
Axel

[njzjz]

Please check documentation and reshape input arrays.