refactoring HDF5Dataset class to check and convert units.

[chrisiacovella]

Pull Request Summary

This address the issue related to the force units being off (per angstrom, not per nanometer) in the datasets.

Key changes

• Units are attached to data read in and converted to our target system
• Each dataset now includes a mapping of a given property in the dataset to one of our available data fields (E, F, dipole_moment, total_charge).
• I commented out any available property that doesn't work in our current system (e.g., per_atom_charges, vibrational frequency, etc.), as they would be read in but couldn't be used in any context

Questions

• While we dont' necessarily intend to embed using per-atom charges for our applications, do we want that to be an option? this might require more extensive discussion, just listing it because I've removed our ability to read any of those properties in at the moment.

Associated Issue(s)

• Force unit in datasets #298

Pull Request Checklist

• Issue(s) raised/addressed and linked
• Includes appropriate unit test(s)
• Appropriate docstring(s) added/updated
• Appropriate .rst doc file(s) added/updated
• PR is ready for review

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 96.96970% with 1 line in your changes missing coverage. Please review.

Project coverage is 85.20%. Comparing base (2998bf1) to head (ae87fb0).

Additional details and impacted files

[wiederm]

Looks great! Thanks for the quick fix!

[wiederm]

While we dont' necessarily intend to embed using per-atom charges for our applications, do we want that to be an option? this might require more extensive discussion, just listing it because I've removed our ability to read any of those properties in at the moment.

I think that's fine. I can't imagine a scenario in which partial charges as input features might be relevant. Even if we were to train on a particular partial charge scheme, we would still do it using element-specific electron negativity as input (which is fixed and can be put inside the model).

[wiederm]

It might be good to add a test that checks that the first datapoint in each dataset has the correct implied units for energies and forces

[chrisiacovella]

It might be good to add a test that checks that the first datapoint in each dataset has the correct implied units for energies and forces

This is a good idea. I'll need to create a new, fake dataset to do this sort of testing. I can do this in a separate PR (especially since this is merged).