Context

At present, there are several deprecation warnings in the test suite logs. If left unaddressed, these will turn into future failures.

We've had two a few cases of breaking issues, which were previously deprecation warnings. For example, a breaking change in numpy 0.24 delayed our most recent release. Let's fix these issues whilst they're still just warnings!

As an aside, it will also be useful to explicitly ignore warnings that are not within our control or that are false positives. That will keep the pytest output concise, and mean we're more likely to spot any new important warnings. This can be done by adding a line to the filterwarnings setting in pyproject.toml.

Where are the warnings?

Look at the most recent test run on the master branch, on the job called ubuntu-latest, 3.12, test. Example screenshot:

TODO

• CI warning: sklearn: 'force_all_finite' was renamed to 'ensure_all_finite' #3969
• CI warning: tensorflow: graph support has been removed in eager mode #3970
• CI warning: tensorflow: Do not pass an input_shape/input_dim argument to a layer #3971
• CI warning: tensorflow: The structure of inputs doesn't match the expected structure #3972
• CI warning: tensorflow: Conversion of an array with ndim > 0 to a scalar is deprecated #3973
• CI warning: transformers: return_all_scores is now deprecated #3974
• CI warning: matplotlib: vert: bool will be deprecated in a future version #3975

Completed