DOC: Corrections about passing different array API classes to methods

[david-cortes-intel]

Description

The docs currently state that it should be possible to fit a model on some array API class and then pass inputs of a different array API class on the same device, but it came up during reviews of PRs from @icfaust that this is not actually supposed to work. This PR corrects the docs to reflect this.

---

Checklist:

Completeness and readability

• I have updated the documentation to reflect the changes or created a separate PR with updates 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 resolved any merge conflicts that might occur with the base branch.

Testing

• All CI jobs are green or I have provided justification why they aren't.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Flag	Coverage Δ
azure	?
github	82.10% <ø> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 31 files with indirect coverage changes

🚀 New features to boost your workflow:

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

[Vika-F]

Is it possible to apply some workarounds, for example model serialization and deserialization, in order to make the predictions on another device possible?

[david-cortes-intel]

Is it possible to apply some workarounds, for example model serialization and deserialization, in order to make the predictions on another device possible?

It would require (a) knowing all of the model object attributes that can be on device; (b) having pytorch with xpu support installed and array api enabled; (c) moving all necessary data to a new pytorch array, which might be a larger load than the data to make predictions on.

Would be easier if oneDAL implemented functionalities to query and move models across devices.