Closed jchodera closed 1 year ago
Let's simply eliminate the total_charge
argument.
Have you been able to verify that it did not impact SPICE model training or assessment?
This function is only used in deployment but not in training or characterization.
In the
charge(...)
API, thetotal_charge
is computed but never used.When the SPICE
run()
is called inproduction.py
, the model is called withChargeEquilibrium()
with no kwargs, which means the total_charge could be erroneously assumed to be 0 iftotal_charge
isn't specified and the model doesn't haveq_ref
.It's not clear to me that there isn't a code path that mistakenly sets the total_charge to 0. Can we fix this to require total_charge, or eliminate the argument altogether to require that we have the formal charges defined in the graph?