For now, tests mostly check whether the parallel and sequential versions give the same result. This is important, but we also should verify other aspects, for example:
Shape
Value range
Returned data type
Invalid input validation (expected exceptions)
Additionally, we should have Scikit-learn API compliance tests:
Input/output format
Pipelines compatibility
Saving/loading transformers, including those in pipelines
For now, tests mostly check whether the parallel and sequential versions give the same result. This is important, but we also should verify other aspects, for example:
Additionally, we should have Scikit-learn API compliance tests: