scikit-learn-intelex
scikit-learn-intelex copied to clipboard
intel
[enhancement, testing] add initial array_api testing capabilities using array-api-strict
Description
Array-api-strict is a module which exposes only the bare essentials defined by the array-api standard. This wraps over top numpy. This will allow for testing of array_api compatability in sklearnex (only on CPU). This will be enabled by either setting get_dataframes_and_queues specifically with "array_api" in the string, or by setting a sklearn context (via set_context or config_context) with array_api_dispatch=True. As this context is available only starting in sklearn1.2, the two options are necessary. Ideally the first modules will be able to use specific new tests, and later then a context can be modified for global testing, similar to how GPU testing is done.
Changes proposed in this pull request:
- add array-api-strict to test requirements
- add import checks in _dataframes_support
- extend get_dataframes_and_queues
- extend convert_to_supported
