Repository navigation
Print out long-range and short-range part during prediction with MACELES #1223
Description
Dear developers,
I am interested to use MACE with long-range corrections (MACE-LES).
I would like to request a guidance on how to print out the short-range and long-range part of the total energy, as well as the atomic charge during the predictions.
As reference, here in the input file I use to train the MACE-LES potential and the script to do total energy and force predictions.
TRAINING_INPUT_FILE
model: "MACELES"
num_interactions: 2
num_channels: 128
max_L: 1
correlation: 3
r_max: 6.0
max_ell: 3
name: "mace01"
model_dir: "MACE_models"
log_dir: "MACE_models"
checkpoints_dir: "MACE_models"
results_dir: "MACE_models"
train_file: "../database/h2o-all.xyz"
valid_fraction: 0.20
test_file: "../database/h2o-all.xyz"
E0s: '{1:-1.122582090, 8:-2.39899122}'
energy_key: "energy"
forces_key: "forces"
device: cuda
batch_size: 5
max_num_epochs: 6000
distributed: True
enable_cueq: True
swa: True
seed: 123
restart_latest: True
PREDICTION_SCRIPT
from ase.io import read, write
import numpy as np
from ase import Atoms
from ase.io.trajectory import Trajectory
from ase.calculators.singlepoint import SinglePointCalculator
from mace.calculators import MACECalculator
calc = MACECalculator(model_path='../MACE_models/mace01.model', device='cuda')
traj = read('dft.xyz', index=':')
for atoms in traj:
atoms.calc = calc
atoms.get_potential_energy()
atoms.write('mace.xyz', append=True)
Thank you for your help.
Activity
ilyes319 commentedon Oct 20, 2025
Hey, that is currently not possible from the ASE calculator atm. We will add that in a future update. If you need it now, you can try to hack it by directly evaluating the model and looking at the "les_energy"
mace/mace/modules/extensions.py
Line 269 in 42cd495