Simon Adorf

**Original comment by Carl Simon Adorf (Bitbucket: [csadorf](https://bitbucket.org/csadorf), GitHub: [csadorf](https://github.com/csadorf)).** ---------------------------------------- This would require a synchronization across some server backend. That's certainly possible and I'm definitely open to this. We...

@martinRenou Any chance that this long-standing bug could be addressed?

/ok to test 0ce12f781

Merge after https://github.com/rapidsai/cuml/pull/6892

This approach works when selecting tests by marker or expression (`-m` / `-k`), but not when selecting by test filename(s) (e.g. `./run-tests.sh /path/to/test_file.py`). In my opinion, this is acceptable since...

Blocked on https://github.com/numba/numba/issues/10101

The affected tests are xfailed in https://github.com/rapidsai/cuml/pull/6865. Once https://github.com/numba/numba/issues/10101 is resolved, we should remove the xfail marker and close this issue.

The root cause has been fixed with the [Python 3.13.5 release](https://pythoninsider.blogspot.com/2025/06/python-3135-is-now-available.html).

> If in cuml there is precedent for having mixed types for fitted attributes then using a numpy array is the simplest thing. We may have precedent, but those still...

> I'd like us to move to a state where we explicitly document what attributes respect the output type setting, and what ones have a fixed type. I think _consistency_...