Improve charge provenance logging

[lilyminium]

Just a few questions on how is charge logging applied, and could it be set to logging.DEBUG instead of INFO? I frequently inspect the logs of Evaluator for information on how protocols are applied, but the charge logging is very noisy to a problematic extent, as it fills up disk space (log files alone take up multiple GB):

2026-04-08 22:15:23.257 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 0
2026-04-08 22:15:23.258 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 27
2026-04-08 22:15:23.258 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 54
2026-04-08 22:15:23.258 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 81
2026-04-08 22:15:23.258 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 108
2026-04-08 22:15:23.258 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 135
2026-04-08 22:15:23.259 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 162
2026-04-08 22:15:23.259 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 189
2026-04-08 22:15:23.259 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 216
2026-04-08 22:15:23.259 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 243
2026-04-08 22:15:23.259 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 270
2026-04-08 22:15:23.259 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 297
2026-04-08 22:15:23.260 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 324
2026-04-08 22:15:23.260 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 351
2026-04-08 22:15:23.260 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 378
2026-04-08 22:15:23.260 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 405
2026-04-08 22:15:23.261 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 432
2026-04-08 22:15:23.261 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 459
2026-04-08 22:15:23.261 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 486
2026-04-08 22:15:23.261 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 513
2026-04-08 22:15:23.261 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 540
2026-04-08 22:15:23.262 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 567
2026-04-08 22:15:23.262 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 594
2026-04-08 22:15:23.262 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 621
2026-04-08 22:15:23.262 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 648
2026-04-08 22:15:23.262 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 675
2026-04-08 22:15:23.262 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 702
2026-04-08 22:15:23.263 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 729
2026-04-08 22:15:23.263 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 756
2026-04-08 22:15:23.263 INFO     Charge section NAGLCharges, using NAGL model openff-gnn-am1bcc-1.0.0.pt, applied to topology atom index 783

I also wonder how the logging is actually applied, since these topology atom indices are not consecutive -- is it per molecule? In that case should the message be a bit more specific about this? Lastly, is it reasonable to expect that all identical molecules get the same charges? In that case could we just log once per unique molecule, instead of for all copies in a Topology?

[mattwthompson]

This was built for OpenFE so we would need to coordinate changes with them, but

could it be set to logging.DEBUG instead

Could be, but would it be better to control this per-module-wise? I forget if this needs to be configured in the module, by the user, or both

is it per molecule

Yes, but based on your logs there may be something wrong with the NAGL case; in the more basic use cases every atom from every molecule has its own line in the log

Lastly, is it reasonable to expect that all identical molecules get the same charges?

Yes, following the toolkit's definitions of identity

... could we just log once per unique molecule ...

Possibly, we'd need to also guide the user on mapping between unique molecules and their copies. I think the toolkit makes this possible but not ergonomic

It the docs are wrong or incomplete I'll update them. (I see now that #1467 needs to be mapped in, or the sections not repeated!)

[lilyminium]

Yes, but based on your logs there may be something wrong with the NAGL case; in the more basic use cases every atom from every molecule has its own line in the log

Ahh, I see what's going on -- it seems that in the case of assigning parameters to a Topology, it logs every first atom of every identical molecule first, then every second atom, then every third ... this is also fairly unintuitive. Does openfe need to map between unique molecule indices and their copies? would it be sufficient to just give the mapped SMILES and log the charges once? In my case this would actually be much more informative than just the indices I have now, since I don't know which order the molecules are in.

import logging
logging.basicConfig(level=logging.INFO)

from openff.toolkit import ForceField, Molecule
from openff.interchange import Interchange
from openff.interchange.components._packmol import pack_box
from openff.units import unit

mol = Molecule.from_smiles("CO")
mol.generate_conformers(n_conformers=1)

topology = pack_box(
    molecules=[mol],
    number_of_copies=[10],
    target_density=0.81 * unit.gram / unit.milliliter,
)

ff = ForceField("openff-2.2.1.offxml")
interchange = Interchange.from_smirnoff(force_field=ff, topology=topology)

WARNING:pint.util:Redefining '[electric_potential]' (<class 'pint.delegates.txt_defparser.plain.DerivedDimensionDefinition'>)
/Users/lilywang/pydev/openff-interchange/openff/interchange/components/interchange.py:1119: FutureWarning: `torch.distributed.reduce_op` is deprecated, please use `torch.distributed.ReduceOp` instead
  if isinstance(obj, functools._lru_cache_wrapper) and obj.__module__.startswith("openff.interchange"):
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 0
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 6
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 12
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 18
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 24
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 30
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 36
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 42
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 48
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 54
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 1
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 7
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 13
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 19
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 25
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 31
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 37
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 43
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 49
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 55
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 2
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 8
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 14
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 20
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 26
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 32
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 38
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 44
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 50
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 56
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 3
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 9
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 15
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 21
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 27
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 33
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 39
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 45
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 51
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 57
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 4
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 10
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 16
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 22
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 28
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 34
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 40
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 46
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 52
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 58
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 5
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 11
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 17
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 23
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 29
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 35
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 41
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 47
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 53
INFO:openff.interchange.smirnoff._nonbonded:Charge section ToolkitAM1BCC, using charge method am1bccelf10, applied to topology atom index 59

Could be, but would it be better to control this per-module-wise? I forget if this needs to be configured in the module, by the user, or both

Yeah I guess I can do this, I hadn't wanted to in case it blocked other logging calls from Interchange, but it looks like these are the only ones. At the same time, though, it is very nice to double check what charge method is being used for a molecule; it's just that the number of lines becomes overwhelming at practical box sizes of 500+ or 1000+ molecules. In fact, given that a charge method is only used if they can assign charges to the whole molecule, would it be possible to just log per molecule?

Actually, based off Irfan's comment here, it seems like only logging each molecule would be preferable to Openfe, at least 2 years ago.

[IAlibay]

Just to briefly weigh in here, whilst the charge provenance information can be useful when debugging, we are also facing the issues mentioned here.

• At the very minimum, dropping the logging information to the debug level would be great. In practice the charge assignment information is too noisy to be useful to non-developers, so info logging is probably too high.
• As mentioned in the initial PR, the information is completely overwhelming for proteins, so any kind of per unique molecule information would be better. If it was just "inchikey" : [charge assignment methods applied], that would do.
• I should note that the request 2 years ago was less "OpenFE needs" and more "pontibus development feedback". I needed this at the time when I was trying to debug unclear charge assignment in the ASFE tooling. This is to say that the behaviour is not expected by our users, so any change would be ok.

[mattwthompson]

Thanks both for the continued discussion - I'll drop the logging level down (straightforward enough, I expect).

For condensing the information, I'd like to split the discussion up into different cases: how to deal with copies of molecules and how to deal with large molecules. (Typically it's one or the other, right?)