Documentation and minor usability fixes

[ipcamit]

Summary

• Per atom weights and prediction logging for Physics based and DNN
• Dataset per atom weight from yaml file
• Renamed KIMDriverGraph to RadialGraph as that is more appropriate
• Miscellaneous import issues, platform specific issues fixed for final release
• exposed strategy and accelerator for lightning trainer
• Complete documentation with instructions to port model to LAMMPS and ASE, and comparison with DUNN

This concludes the development. Currently it supports python 3.9, due to CMake limitations. Post release I can try and fix it with other releases as well (CMake 3.18 is the only version guaranteed to detect CUDA properly, but has hardcoded support for python <= 3.9). Stress tested it on Debian and RHEL like platforms (docker) and Mac M2, to ensure that all possible errors are catalogued.

The documentation can be seen here: https://kliff-v1.readthedocs.io/en/latest/index.html

TODO

• Advanced documentation about more customized training

• Tutorial to implement custom trainer

• Make sure your code is properly formatted. isort and black are used for this purpose. The simplest way is to use pre-commit. See instructions here.

• Doc strings have been added in the Google docstring format on your code.

• Type annotations are highly encouraged. Run mypy to type check your code.

• Tests have been added for any new functionality or bug fixes.

• All linting and tests pass.

Note that the CI system will run all the above checks. But it will be much more efficient if you already fix most errors prior to submitting the PR.