Alby M.

Hi @tgmaxson , Not sure about status of this implementation; our implementation using the existing `StressForceOutput` approach is available for Allegro with a prototype parallel integration into LAMMPS here: https://github.com/mir-group/pair_allegro/tree/stress...

Hi @Hongyu-yu , Ah great, thanks for the update! > I have also tried StressOutput with Allegro in lammps in parallel but it gives inconsistent force and stress caused by...

@Hongyu-yu I see, thanks for clarifying and the quick responses!--- I'm hoping to find enough time to look through all this carefully this week.

@brucefan1983 > So `output_og[AtomicDataDict.VIRIAL_KEY]` is from the derivative of energy with respect to box? `StressForceOutput` is `nequip`'s module for computing stress/virial (for both NequIP and Allegro; we now have it...

Hi @pablo-unzueta , Thanks for your interest in our code! I'm not sure why this is happening, but `initialize_from_state_strict: false` would be the correct option in this case. You could...

Yes, in princple if you just set it to some number, it will get overriden by the loaded state dict, but I'm still not totally sure why this is happening...

What do you mean by "as well"? There is no reason to train virials and stresses simultaneously as they are fully redundant.

Aha, I see: we do not do any unit conversions for either virials or stress, since `nequip` has no formal sense of units and just inherits the units of the...

Hi @ThePauliPrinciple , Thanks for your nice question and work with our code! Re > checking if I get close to a walltime on a compute cluster we do have...

Great! A stopper is "stateful" when it maintains a state like, say, how many epochs the validation loss hasn't improved (like the patience setting) or what the minimum observed value...