[MNT] Improve coverage of test framework for ptf-v2 and fix TiDE categorical embedding bug

[harshsomankar123-tech]

Reference Issues/PRs

Fixes #2328.

What does this implement/fix? Explain your changes.

This PR increases the test coverage of the V2 testing framework by introducing new data scenarios, parameterizing integration tests, adding validation-check tests, and resolving a categorical embedding bug in the TiDE V2 model.

Specifically:

1. Diverse V2 Data Scenarios:
   • Added make_datasets_v2_without_covariates to verify estimator behaviors on datasets containing only targets and group IDs without any covariates.
   • Added make_datasets_v2_with_categoricals to verify estimators handling categorical features alongside continuous features.
2. Parameterized Integration Tests:
   • Parameterized V2 model integration tests in test_all_estimators_v2.py across three data scenarios: with_covariates, without_covariates, and with_categoricals.
3. New Test Validation Cases:
   • Added test_uninitialized_predict_error to assert that calling .predict() on an uninitialized model correctly raises a RuntimeError.
   • Added test_predict_save_to_dir to assert that calling .predict() with an output_dir successfully saves the prediction dictionary to a predictions.pkl file.
4. Fixed TiDE Categorical Embedding Bug:
   • Updated TslibDataModule and EncoderDecoderTimeSeriesDataModule metadata builders to compute and include "categorical_cardinalities" (derived from the dataset's unique values/max values).
   • Updated TIDE model constructor to automatically resolve categorical cardinalities from the datamodule metadata when embs is not explicitly provided. This prevents the IndexError on categorical feature embedding projection.

What should a reviewer concentrate their feedback on?

• Review the integration of "categorical_cardinalities" metadata propagation inside TslibDataModule and EncoderDecoderTimeSeriesDataModule.
• Inspect the fallback mechanism in TIDE.__init__ that dynamically maps these cardinalities into embedding dimensions when embs is not explicitly supplied in model configuration.

Did you add any tests for the change?

Yes:

• Added parameterized integration tests for all V2 models across the three new and existing data scenarios (with_covariates, without_covariates, and with_categoricals).
• Added test_uninitialized_predict_error and test_predict_save_to_dir check cases in test_all_estimators_v2.py.

Any other comments?

All V2 estimators (DLinear, TimeXer, TFT, TiDE) were verified to pass the integration and validation test suites successfully (201 total tests passed).

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@e7c7965). Learn more about missing BASE report.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2345   +/-   ##
=======================================
  Coverage        ?   87.60%           
=======================================
  Files           ?      171           
  Lines           ?    10157           
  Branches        ?        0           
=======================================
  Hits            ?     8898           
  Misses          ?     1259           
  Partials        ?        0           

Flag	Coverage Δ
cpu	87.60% <100.00%> (?)
pytest	87.60% <100.00%> (?)

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

☔ View full report in Codecov by Harness.
📢 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.

[harshsomankar123-tech]

@phoeenniixx @fkiraly could you please take a look!
Thanks!

[phoeenniixx]

Thanks!
This is a difficult task! Actually i was thinking of adding scenario classes and updating the test framework in a similar way as in sktime.

(also the tests are failing)