Closed Andrew-S-Rosen closed 4 months ago
All modified and coverable lines are covered by tests :white_check_mark:
Comparison is base (
c99288e
) 98.86% compared to head (40956f8
) 98.86%.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
Hi @Andrew-S-Rosen, thanks for your PR! I noticed that the current AdvancedSoft MatGL-LAMMPS interface breaks due to the changes for the datatype of potential_energy() from numpy.ndarray into float (see line 105 in https://github.com/advancesoftcorp/lammps/blob/based-on-lammps_2Aug2023/potentials/M3GNET/matgl_driver.py). I will submit a pull request to AdvancedSoft to fix the bug.
This is a follow-up to #203.
The
Calculator.results
dictionary from an M3GNet run yields something like'free_energy': array(-4.0938973, dtype=float32)
for the free energy property. This causes some havoc in places (e.g. monty jsanitization) since the array isn't iterable. It doesn't have a definedlen
without the brackets. We should really make this afloat
instead anyway. That's what I do in this PR.