Testing coverage holes

[djgagne]

Since our test coverage is currently low, what are the highest priority areas for adding more testing? I can run pytest-cov on current code and encourage others to look through existing code and tests and see what is missing. Also might be a good place to get an assessment from an LLM too.

[charlie-becker]

I'd like to see some simple integration testing.

Specifically, running tests for training and inference scripts. I know this might have snags using git runners without a GPU, but I imagine there are work arounds. We could also test on Cirrus with a GPU.

[djgagne]

@ggantos tagging to update testing coverage for key models.