Adding cleavage benchmarks

[vue1999]

Pre-review checklist for PR author

PR author must check the checkboxes below when creating the PR.

• I've confirmed the contribution guidelines.

Summary

Adding a benchmark to evaluate the accuracy of predicting cleavage energies of crystalline surfaces.

Linked issue

Resolves #403

Progress

• Calculations
• Analysis
• Application
• Documentation

Testing

MACE omat, ORB v3

New decorators/callbacks

no

[joehart2001]

Hey @vue1999, thanks for the PR and its looking super good. A few things:

• ive made some code suggestions to the calc script so we save files in the generalised format. I know there are 30,000 files... but what do you think @ElliottKasoar?
• ive made some knock on suggestions to the analysis based on these calc changes
• ive also made suggestions which implement the density scatter plot + structure visualisation, as theres a lot of structures
• We are also going to in the future add the ability to swithc betwen types of errors e.g. mae and rmse, so ive suggested we just keep mae for now unless you're against this?

Let me know if you've got any ideas or are unsure about any of my suggestions, thanks!

[vue1999]

Thanks for the review @joehart2001! Happy to accept all suggestions and fine to keep just MAE for now.

[ElliottKasoar]

Thanks for this, @vue1999, and sorry for the delayed review on my side, it's looking great so far!

I've left a few minor comments. Could you run the pre-commit (see: https://ddmms.github.io/ml-peg/developer_guide/get_started.html#automatic-coding-style-check), to tidy up a few things for the automated tests?

I am also a bit concerned by the 30,000 files. The analysis just for a single model took me three minutes. We definitely still want access to the structure files, and it's good to save the data with them, but I wonder if it's worth also saving this to a simpler file format e.g. as you had previously with JSON (sorry if this feels back-and-forth) to be more practical.

It could also be worth exploring saving to a reduced number of files e.g. all structs in each mpid_dir, but I'm not sure it would make a huge difference.

Thoughts @joehart2001?

[ElliottKasoar]

Could you set the charge and spin (multiplicity) for these structures? Orb-omol requires them to be set

[ElliottKasoar]

I get an import error trying from this import. This branch may need rebasing?

[ElliottKasoar]

I don't think read is currently imported?