[MNT] Enhance networks module testing suite

[aadya940]

• Re-implemented test_all_networks.py to create a more robust testing Mechanism.
• Added _config dict to keep track of library dependencies and compatible python versions and the network structure.
• Checked if build_network returns correct return type, that is, KerasTensor if structure is encoder, Model object otherwise.

[aeon-actions-bot]

Thank you for contributing to aeon

I have added the following labels to this PR based on the title: [ $\color{#EC843A}{\textsf{maintenance}}$ ].
I have added the following labels to this PR based on the changes made: [ $\color{#379E11}{\textsf{networks}}$ ]. Feel free to change these if they do not properly represent the PR.

The Checks tab will show the status of our automated tests. You can click on individual test runs in the tab or "Details" in the panel below to see more information if there is a failure.

If our pre-commit code quality check fails, any trivial fixes will automatically be pushed to your PR unless it is a draft.

Don't hesitate to ask questions on the aeon Slack channel if you have any.

[aadya940]

@MatthewMiddlehurst let's check if this improves coverage or not.

[MatthewMiddlehurst]

You will have to push another commit for the test to run unfortunately.

[TonyBagnall]

I think we can put this in without testing coverage, seems a good thing to do!

[MatthewMiddlehurst]

Taking a look, I don't think this is needed anymore actually, see line 33 of the new file where it skips individual estimators if there is a missing module.

Honestly, this test could be improved a lot though, the skip list and try/catch are a bit hacky. Should try to integrate _check_soft_dependencies without skipping the whole test.

[aadya940]

I think we could skip the complete test if the tensorflow is not present. However, we could skip the networks which use tensorflow-addons?

[MatthewMiddlehurst]

Ideally we shouldnt skip anything other than the base class, the only good reason to skip currently is it has a dependency which is not installed

[MatthewMiddlehurst]

It could be good to remove the current skipped classes and replace the except with one using the _check_soft_dependencies(soft_dependencies) and _check_python_version(python_version) methods. You may have to edit the new base class a bit to get the correct inputs.

If you can think of anything else we would want to add to this test, that would be good. Currently it does not check anything from the output.

[aadya940]

It could be good to remove the current skipped classes and replace the except with one using the _check_soft_dependencies(soft_dependencies) and _check_python_version(python_version) methods. You may have to edit the new base class a bit to get the correct inputs.

If you can think of anything else we would want to add to this test, that would be good. Currently it does not check anything from the output.

Could you please elaborate on this? I didn't quite understand. Thanks :)

[aadya940]

How does this look? @MatthewMiddlehurst

[aadya940]

Do we skip this using pytest.mark.skipif if the version is 3.10? Looks like tensorflow.keras is incompatible with 3.10.

[aadya940]

Got it 👍 , shall update it once we take a decision. Thanks :))

[MatthewMiddlehurst]

I have updated my recommendation above.

[MatthewMiddlehurst]

LGTM. Still some minor issues but does not need to be done here.

[hadifawaz1999]

LGTM! would be good to check coverage now too