How to use deepmd-kit to optimize a molecule?

[Franklalalala]

I know that, you train a model use labeled system, then, you can use this model to pridict a molecule's energy and force. Question is, can this model optimize a raw, unstable molecule and give a final energy, forces.
I think maybe, use a for loop to predict energy while change the coords with the gradient(force), end loop when the force smaller enough. Is there better way to do this?

[njzjz]

Here is an example to optimize a water molecule with help of a third-party package ASE:
https://docs.deepmodeling.org/projects/deepmd/en/master/use-deepmd-kit.html#use-deep-potential-with-ase

from ase import Atoms
from deepmd.calculator import DP

water = Atoms('H2O',
              positions=[(0.7601, 1.9270, 1),
                         (1.9575, 1, 1),
                         (1., 1., 1.)],
              cell=[100, 100, 100],
              calculator=DP(model="frozen_model.pb"))
print(water.get_potential_energy())
print(water.get_forces())

from ase.optimize import BFGS
dyn = BFGS(water)
dyn.run(fmax=1e-6)
print(water.get_positions())

[Franklalalala]

Thank you!!