'Floating point exception' issue when dispersion=True in mace_mp

[jycho5758]

I'm trying to consider dispersion term when using mace_mp as below,

def ase_calc():
    calc = mace_mp(
        model="/home/jycho/AKMC_RUN/LYC_300K_test/mace-mp0.model",
        default_dtype="float64",
        dispersion=True,
        device="cuda",
    )
    return calc

However, it gives me the following error in the output file.

Floating point exception

I have been using the conda virtual environment with the following

mace-torch 0.3.15
torch 2.3.1
torch-dftd 0.5.3
torch-ema 0.3
torchaudio 2.3.1
torchmetrics 1.4.0.post0
torchvision 0.18.1

I was wondering whether there is an issue with torch-dftd library or torch version etc.

[ilyes319]

Hey you seem to be using quite an old torch version, any reason for that?

[jycho5758]

Hi, there is no specific reason and just found that my torch version is quite old. So, I updated the torch version from 2.3.1 to 2.10.0 but got the same error as below.

/home/jycho/miniforge3/envs/eon_ase/lib/python3.11/site-packages/e3nn/o3/_wigner.py:10: UserWarning: Environment variable TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD detected, since the`weights_only` argument was not explicitly passed to `torch.load`, forcing weights_only=False.
  _Jd, _W3j_flat, _W3j_indices = torch.load(os.path.join(os.path.dirname(__file__), 'constants.pt'))
/home/jycho/miniforge3/envs/eon_ase/lib/python3.11/site-packages/mace/calculators/mace.py:199: UserWarning: Environment variable TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD detected, since the`weights_only` argument was not explicitly passed to `torch.load`, forcing weights_only=False.
  torch.load(f=model_path, map_location=device)
Floating point exception