[ENH] Standardize testing of estimator outputs and skip tests for non-conformant estimators

[PranavBhatP]

Reference Issues/PRs

#1966

What does this implement/fix? Explain your changes.

Fix #1966.
Uncovered non-conformances:

1. TimeXer
2. NBeatsKAN (newest estimator)
3. NBeats
4. xLSTMTime

This PR enforces testing of the estimators' output format on only one case - [batch_size, prediction_length, loss_dim]. The non-conformance in the above mentioned models, was the presence of 2D output for single targets case.

What should a reviewer concentrate their feedback on?

Did you add any tests for the change?

Any other comments?

PR checklist

• The PR title starts with either [ENH], [MNT], [DOC], or [BUG]. [BUG] - bugfix, [MNT] - CI, test framework, [ENH] - adding or improving code, [DOC] - writing or improving documentation or docstrings.
• Added/modified tests
• Used pre-commit hooks when committing to ensure that code is compliant with hooks. Install hooks with pre-commit install.
  To run hooks independent of commit, execute pre-commit run --all-files

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (main@a3bc452). Learn more about missing BASE report.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1971   +/-   ##
=======================================
  Coverage        ?   85.32%           
=======================================
  Files           ?      158           
  Lines           ?     9296           
  Branches        ?        0           
=======================================
  Hits            ?     7932           
  Misses          ?     1364           
  Partials        ?        0           

Flag	Coverage Δ
cpu	85.32% <100.00%> (?)
pytest	85.32% <100.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[phoeenniixx]

you could also add "tests:skip_by_name" tag in corresponding _pkg classes. There if you add test_integration it should skip the integration tests for that model class. This would not skip other tests for that model, like it is now, where framework is skipping the whole _pkg class.

[fkiraly]

yes, although afaik we would have to add the machinery for that tag in pytorch-forecasting first. I do not think it works out of the box right now, since I have not added that feature to scikit-base yet. There is code in sktime that could be copied for this.

[phoeenniixx]

Yes, it exists! we already brought it (and modified it for _pkg usage) from sktime. See here
We are already using it to skip specific params for some buggy fixtures(like LogNormalDistributionLoss for DecoderMLP, see here). So, I think we could use it here as well...