[Fix] add dtype attribute to python oneDAL table objects

[icfaust]

Description

Taken from #2126 convert_to_supported will set the data to float, but the fptype param may still be double. This is most simply solved by querying the data for its dtype at the parameter setting stage. Ideally fptype should be instead managed entirely in the backend. Unfortunately, that change is very core to the backend offloading and should be done more carefully.

---

PR should start as a draft, then move to ready for review state after CI is passed and all applicable checkboxes are closed.
This approach ensures that reviewers don't spend extra time asking for regular requirements.

You can remove a checkbox as not applicable only if it doesn't relate to this PR in any way.
For example, PR with docs update doesn't require checkboxes for performance while PR with any change in actual code should have checkboxes and justify how this code change is expected to affect performance (or justification should be self-evident).

Checklist to comply with before moving PR from draft:

PR completeness and readability

• I have reviewed my changes thoroughly before submitting this pull request.
• I have commented my code, particularly in hard-to-understand areas.
• I have updated the documentation to reflect the changes or created a separate PR with update and provided its number in the description, if necessary.
• Git commit message contains an appropriate signed-off-by string (see CONTRIBUTING.md for details).
• I have added a respective label(s) to PR if I have a permission for that.
• I have resolved any merge conflicts that might occur with the base branch.

Testing

• I have run it locally and tested the changes extensively.
• All CI jobs are green or I have provided justification why they aren't.
• I have extended testing suite if new functionality was introduced in this PR.

Performance

• I have measured performance for affected algorithms using scikit-learn_bench and provided at least summary table with measured data, if performance change is expected.
• I have provided justification why performance has changed or why changes are not expected.
• I have provided justification why quality metrics have changed or why changes are not expected.
• I have extended benchmarking suite and provided corresponding scikit-learn_bench PR if new measurable functionality was introduced in this PR.

[david-cortes-intel]

convert_to_supported will set the data to float, but the fptype param may still be double

I'm not understanding the issue. What would be fp64 when the data is fp32?

[icfaust]

convert_to_supported will set the data to float, but the fptype param may still be double

I'm not understanding the issue. What would be fp64 when the data is fp32?

our tables can ingest complex float32 types (https://github.com/intel/scikit-learn-intelex/blob/main/onedal/datatypes/utils/numpy_helpers.hpp#L71) and will return float32 types: https://github.com/intel/scikit-learn-intelex/blob/main/onedal/datatypes/utils/numpy_helpers.cpp#L34, our checks for dtype are weak. We assume that dtype can only be of float32 or float64 which must be enforced at the very beginning of the code on sklearnex side using validate_data (just as an example from Linear Regression: https://github.com/intel/scikit-learn-intelex/blob/main/onedal/linear_model/linear_model.py#L46). We should use the datatype of the oneDAL table to describe how oneDAL is to operate.

[david-cortes-intel]

convert_to_supported will set the data to float, but the fptype param may still be double

I'm not understanding the issue. What would be fp64 when the data is fp32?

our tables can ingest complex float32 types (https://github.com/intel/scikit-learn-intelex/blob/main/onedal/datatypes/utils/numpy_helpers.hpp#L71) and will return float32 types: https://github.com/intel/scikit-learn-intelex/blob/main/onedal/datatypes/utils/numpy_helpers.cpp#L34, our checks for dtype are weak. We assume that dtype can only be of float32 or float64 which must be enforced at the very beginning of the code on sklearnex side using validate_data (just as an example from Linear Regression: https://github.com/intel/scikit-learn-intelex/blob/main/onedal/linear_model/linear_model.py#L46). We should use the datatype of the oneDAL table to describe how oneDAL is to operate.

Got it. But if complex numbers are the only issue, wouldn't it be better to create a helper function along the lines of "is_fp32_dtype" or so, and convert the data in python before it reaches oneDAL? It'd avoid double copies or conversions when the inputs are not in array formats (e.g. arrow[complex64] -> numpy[complex64] -> array[float32]).

Perhaps it could just use np.finfo(data.dtype).dtype.

[david-cortes-intel]

Another question in that regard: could the dtype potentially be used as part of the fit condition checks for oneDAL algorithm support?

[samir-nasibli]

Suggested change

	table_obj.def_property_readonly("dtype", [](const table& t){
	table_obj.def_property_readonly("get_numpy_dtype", [](const table& t){

I suggest to use this or just numpy_dtype naming, just not to be confused

[samir-nasibli]

And again, from my point of view for tables API, this seems unnecessary and looks like a workaround or a hack.