Calculation error when using python interface with DPA

[caolz]

Here is my code

job_directorys = glob("job-*")
dp = DeepPot("DPA-Semi-model.pb")

for job_directory in job_directorys:
    d_poscar = dpdata.System(f"{job_directory}/POSCAR")
    coords = d_poscar["coords"]
    cell = d_poscar["cells"]
    atype = d_poscar["atom_types"]
    e, f, v = dp.eval(coords, cell, atype)
    print(e)

The calculations are wrong, the energy is positive. No problem if using lammps. What kind of question is that?

[caolz]

There is no problem using the ase interface.

from ase.io import read
from deepmd.calculator import DP

st = read('POSCAR')
calculator = DP(model="DPA-Semi-model.pb")

st.set_calculator(calculator)  

potential_energy = st.get_potential_energy()

print(potential_energy)

[njzjz]

The type map in the data cannot be assumed to be the same as that in the model.

[njzjz]

Btw, you may consider using d_poscar.predict(dp).

[caolz]

Thank you for your response. The problem has been resolved!