Olivier Grisel

Package level-zero on the conda-forge distribution

2

At the moment it is not possible to install `dpctl`, `dpnp` and/or `numba-dpex` from conda-forge with Intel GPU support because `level-zero` has not been packaged for that distribution. Would you...

Found on the joblib CI: ```python ================================== FAILURES =================================== _______ test_multithreaded_parallel_termination_resource_tracker_silent _______ @with_numpy @with_multiprocessing def test_multithreaded_parallel_termination_resource_tracker_silent(): # test that concurrent termination attempts of a same executor does not # emit...

Fix AttributeError use_fallback_lbfgs_solve for newton-cholesky when fitting with max_iter=0

2

As found when trying to run [`benchopt`](https://benchopt.github.io/) on LogisticRegression with the `newton-cholesky` solver. When writing the non-regression test, I found that both sag/saga and LBFGS also had problems with `max_iter=0`...

Array API support for k-nearest neighbors models with the brute force method

1

This issue is a sibling of a similar issue for k-means: #26585 with similar purpose but likely different constraints. In particular an efficient implementation of k-NN on the GPU would...

Array API support for k-means

1

This is an early issue to publicly discuss the possibility (or not) to use the Array API (see #22352) for k-means and make it run on GPUs using PyTorch in...

SplineTransformer(extrapolate="periodic") outputs nan values for constant features

6

While reviewing #24145 I discovered the following bug: ```python In [1]: import numpy as np In [2]: from sklearn.preprocessing import SplineTransformer In [3]: SplineTransformer(extrapolation="periodic").fit_transform(np.ones(shape=(5, 1))) Out[3]: array([[nan, nan, nan, nan],...

[WIP] Implement Gini coefficient for model selection with positive regression GLMs

2

Fixes #10003. This PR builds on top of #14300. I plan to rebase the last 2 commits on top of #14300 from time to time. TODO: - [ ] Add...

DOC: improve plot_compare_calibration.py

6

Related to the ongoing discussion on improving the user guide section that is based on this example: #28171. I realized that using the default value of C=1 was yielding under-confident...

feat: make it possible to pass a seed parameter to `ibis.random`

12

### Is your feature request related to a problem? It would be neat to get repeatable queries that use `ibis.random`, for instance when randomly reordering results (`t.order_by(ibis.random())`): - #7689 has...

Consider adding `POLARS_MAX_THREADS` to `joblib._parallel_backends.MAX_NUM_THREADS_VARS`

1

This would allow to protect against over-subscription when nesting polars multithreaded operations below `joblib.Parallel` calls when using the process-based loky backend. See: https://github.com/pola-rs/polars/issues/7922 Similarly to what we already do for...