Josh Fass

Ahh, thanks for the more dramatic example! Enumerating all configurations around the double-bonds in the molecule `O/C(=C(/O)\C(\Cl)=C(\F)Cl)/C(/Cl)=C(\F)Cl`, again the picture is similar:  * Proposed symmetry classes are stereo invariant...

Toggling `symmetrize=True` to `symmetrize=False` in AM1ELF10 https://github.com/proteneer/timemachine/blob/451803e01afe6231147a0e6a3ca019d4aa5069d8/timemachine/ff/handlers/nonbonded.py#L89-L91 causes large differences of up to 0.1 e within each graph symmetry class within each stereoisomer of `O/C(=C(/O)\C(\Cl)=C(\F)Cl)/C(/Cl)=C(\F)Cl` (as should be expected).

Based on offline discussion with @proteneer -- looked into using the RDKit alternative https://www.rdkit.org/docs/source/rdkit.Chem.rdmolfiles.html#rdkit.Chem.rdmolfiles.CanonicalRankAtoms . This appears to have the desired behavior! The PR now: * Disables passing `symmetrize=True` to...

>would be good sanity check that this doesn't drastically affect FreeSolv results Checking effect on computed charges for FreeSolv molecules (ref: master (451803e) -> proposed: most recent PR commit (9ce6cb9))...

Thanks @proteneer and @bp-kelley -- will try this out! (Just want to strengthen the test to assert that the difference between old and new symmetrization behavior is indeed limited to...

Stale -- migrated to https://github.com/proteneer/timemachine/issues/817

>What's the rationale behind just using the rings? Limiting the number of compounds? No rationale, just a starting point -- intention is still to use also the AlkEthOH chains set...

Hmm, although the Pytorch views in https://github.com/choderalab/espaloma/blob/973d5e1de00b60390b93a054c4277db632569b04/espaloma/data/alkethoh/pytorch_datasets.py satisfy the pytorch Dataset interface, they don't yet play nice with DataLoader. For example, ```python import torch dataset = AlkEthOHAtomTypesDataset() loader = torch.utils.data.DataLoader(dataset,...

- [ ] RMSE is rounded to 2 digits of precision -- we might want to use more digits of precision here, or report log error. - [ ] `atom_k`...

- [ ] In model summary section, could we also include the total number of parameters? Often interested in this