lammps extra_properties per atom #163
Conversation
…l purpose. Standard deviation is called from the LocalAtomExtractorNode
…ij, etc such that setup_graph isn't called multiple times
…ij, etc such that setup_graph isn't called multiple times
|
|
||
| if args.seed == 1001: | ||
| netname = f'TEST_ALUMINUM_MODEL_{args.seed}' | ||
| else: |
There was a problem hiding this comment.
If we're going to add arguments to this script let's just make a --ensemble one that changes the location for the model to be stored, rather than a special seed value.
examples/ase_example_multilayer.py
Outdated
| extra_properties = {"ens_predictions": ensemble_energy.all, "ens_std": ensemble_energy.std, "force_std":ensemble_force.std} | ||
| calc = HippynnCalculator(energy_node, extra_properties=extra_properties, en_unit=units.eV) #added | ||
| #calc = HippynnCalculator(energy_node, en_unit=units.eV) | ||
| '''# Load the files |
examples/ase_example_multilayer.py
Outdated
| MaxwellBoltzmannDistribution(atoms, temperature_K=300) | ||
| dyn = VelocityVerlet(atoms, 0.5*units.fs) | ||
|
|
||
|
|
There was a problem hiding this comment.
It seems like maybe this script is getting repurposed for two things. Would it make sense to pull out this writing of data for lammps into an script in the examples/lammps folder?
examples/ase_example_multilayer.py
Outdated
| print('Energy per atom: Epot = %.7feV Ekin = %.7feV (T=%3.0fK) ' | ||
| 'Etot = %.7feV Stress = %.7f' % (epot, ekin, ekin / (1.5 * units.kB), epot + ekin, stress[:3].sum()/3 / units.bar)) | ||
|
|
||
| #atomic_numbers = self.atoms.get_atomic_numbers() |
| # Load trained model | ||
| try: | ||
| with active_directory("../TEST_ALUMINUM_MODEL", create=False): | ||
| with active_directory("/vast/home/dshahi/test_al_model/TEST_ALUMINUM_MODEL", create=False): |
|
Overall the implementation strategy looks ok, but there are several places where there are stray comments in the code or it looks like we are committing changes to the examples which break their intended functionality. We want to keep existing functionality in the examples intact. Where using the ensembles requires a larger amount of changes to the example script we should make that a separate script. |
deepsana
force-pushed
the
generalized-changes-final
branch
from
f4312af to
db7ea86
Compare
…r getting ase forces and positions in example/lammps
deepsana
force-pushed
the
generalized-changes-final
branch
from
db7ea86 to
91087bf
Compare
lubbersnick
changed the title
3 participants
extra_propertiesdictionary that allows one to include extra properties such as standard deviation to the energies or other extra properties.