torchani
torchani copied to clipboard
Unsupported atoms are simply ignored?
I would expect the following code to fail
import torch
import torchani
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = torchani.models.ANI2x(periodic_table_index=True).to(device)
coordinates = torch.tensor([[[0.03192167, 0.00638559, 0.01301679],
[-0.83140486, 0.39370209, -0.26395324],
[-0.66518241, -0.84461308, 0.20759389],
[0.45554739, 0.54289633, 0.81170881],
[0.66091919, -0.16799635, -0.91037834]]],
requires_grad=True, device=device)
# Species with unsupported atomic number 42
species = torch.tensor([[42, 1, 1, 1, 1]], device=device)
energy = model((species, coordinates)).energies
but it works without error/warnings despite the unsupported atom
In [12]: energy.item()
Out[12]: -2.1059779035978243
Is this intended? I understand the need to ignore atoms internally due to padding, but ignoring input atoms silently feels very error-prone to me (from an user perspective).
An exception should be raised on such cases. Silently ignoring atoms is really bad.
Thanks! We will take care of this issue soon!
@UnixJunkie totally agree. I closed this issue only because it has not been acknowledged for over a year and I don't even know if it's still relevant.
Thanks @yueyericardo for re-opening an looking into this.