Merge branch 'main' into dev/pytorch_testing_2

david-cortes-intel · david-cortes-intel · commit 3a6254e889aa · 2025-12-10T08:23:39.000+01:00
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -16,7 +16,7 @@
 
 repos:
   - repo: https://github.com/psf/black
-    rev: 25.11.0
+    rev: 25.12.0
     hooks:
       - id: black
   - repo: https://github.com/PyCQA/isort
diff --git a/deselected_tests.yaml b/deselected_tests.yaml
@@ -34,10 +34,6 @@ deselected_tests:
   - neighbors/tests/test_neighbors.py::test_nan_euclidean_support[KNeighborsClassifier-params0]
   - neighbors/tests/test_neighbors.py::test_nan_euclidean_support[KNeighborsRegressor-params1]
   - neighbors/tests/test_neighbors.py::test_nan_euclidean_support[LocalOutlierFactor-params6]
-  - neighbors/tests/test_neighbors.py::test_neighbor_classifiers_loocv[ball_tree-nn_model0]
-  - neighbors/tests/test_neighbors.py::test_neighbor_classifiers_loocv[brute-nn_model0]
-  - neighbors/tests/test_neighbors.py::test_neighbor_classifiers_loocv[kd_tree-nn_model0]
-  - neighbors/tests/test_neighbors.py::test_neighbor_classifiers_loocv[auto-nn_model0]
   # sklearn 1.7 unsupported features
   - tests/test_common.py::test_estimators[LinearRegression()-check_sample_weight_equivalence_on_dense_data]
 
diff --git a/doc/sources/distributed-mode.rst b/doc/sources/distributed-mode.rst
@@ -26,7 +26,7 @@ Several :doc:`GPU-supported algorithms <oneapi-gpu>`
 also provide distributed, multi-GPU computing capabilities via integration with |mpi4py|. The prerequisites
 match those of GPU computing, along with an MPI backend of your choice (`Intel MPI recommended
 <https://www.intel.com/content/www/us/en/developer/tools/oneapi/mpi-library.html>`_, available
-via the ``impi_rt`` python/conda package) and the |mpi4py| python package. If using |sklearnex|
+via the ``impi_rt`` / ``impi-rt`` python/conda package) and the |mpi4py| python package. If using |sklearnex|
 `installed from sources <https://github.com/uxlfoundation/scikit-learn-intelex/blob/main/INSTALL.md#build-from-sources>`_,
 ensure that the spmd_backend is built.
 
@@ -49,9 +49,9 @@ ensure that the spmd_backend is built.
         .. tab:: From Intel's pip Index
             ::
 
-                pip install --index-url https://software.repos.intel.com/python/pypi mpi4py impi_rt
+                pip install --index-url https://software.repos.intel.com/python/pypi mpi4py impi-rt
 
-  It also requires the MPI runtime executable (``mpiexec`` / ``mpirun``) to be from the same library that was used to compile |sklearnex| or from a compatible library. Intel's MPI runtime library is offered as a Python package ``impi_rt`` and will be installed together with the ``mpi4py`` package if executing the commands above, but otherwise, it can be installed separately from different distribution channels:
+  It also requires the MPI runtime executable (``mpiexec`` / ``mpirun``) to be from the same library that was used to compile |sklearnex| or from a compatible library. Intel's MPI runtime library is offered as a Python package ``impi_rt`` (conda) / ``impi-rt`` (PyPI) and will be installed together with the ``mpi4py`` package if executing the commands above, but otherwise, it can be installed separately from different distribution channels:
 
     .. tabs::
         .. tab:: From conda-forge
@@ -67,12 +67,12 @@ ensure that the spmd_backend is built.
         .. tab:: From PyPI
             ::
 
-                pip install impi_rt
+                pip install impi-rt
 
         .. tab:: From Intel's pip Index
             ::
 
-                pip install --index-url https://software.repos.intel.com/python/pypi impi_rt
+                pip install --index-url https://software.repos.intel.com/python/pypi impi-rt
 
 
   Using other MPI backends that are not MPICH-compatible (e.g. OpenMPI) requires building |sklearnex| from source with that backend, and using an |mpi4py| built with that same backend.
diff --git a/doc/sources/distributed_daal4py.rst b/doc/sources/distributed_daal4py.rst
@@ -45,7 +45,7 @@ same algorithms to much larger problem sizes.
     the MPI runtime library managing the computations to be the same MPI backend library
     with which the |sklearnex| library was compiled, or to be ABI compatible with it.
     Distributions of the |sklearnex| in PyPI and conda-forge are both compiled with `Intel's MPI <https://www.intel.com/content/www/us/en/developer/tools/oneapi/mpi-library.html>`__
-    as MPI backend (offered as Python package ``impi_rt`` in both PyPI and conda): ::
+    as MPI backend (offered as Python package ``impi_rt`` in conda, or ``impi-rt`` in PyPI): ::
 
         conda install -c conda-forge impi_rt mpi=*=impi
 
@@ -77,7 +77,7 @@ same algorithms to much larger problem sizes.
         .. tab:: From Intel's pip Index
             ::
 
-                pip install --index-url https://software.repos.intel.com/python/pypi mpi4py impi_rt
+                pip install --index-url https://software.repos.intel.com/python/pypi mpi4py impi-rt
 
 
 Using distributed mode
diff --git a/onedal/covariance/incremental_covariance.py b/onedal/covariance/incremental_covariance.py
@@ -18,13 +18,10 @@
 
 from daal4py.sklearn._utils import daal_check_version
 
-from .._config import _get_config
 from .._device_offload import supports_queue
 from ..common._backend import bind_default_backend
 from ..datatypes import from_table, return_type_constructor, to_table
 from ..utils import _sycl_queue_manager as QM
-from ..utils._array_api import _get_sycl_namespace
-from ..utils.validation import _check_array
 from .covariance import BaseEmpiricalCovariance
 
 
diff --git a/onedal/datatypes/_dlpack.py b/onedal/datatypes/_dlpack.py
@@ -14,8 +14,6 @@
 # limitations under the License.
 # ==============================================================================
 
-from collections.abc import Iterable
-
 import numpy as np
 
 from onedal import _default_backend as backend
diff --git a/onedal/decomposition/incremental_pca.py b/onedal/decomposition/incremental_pca.py
@@ -14,13 +14,10 @@
 # limitations under the License.
 # ==============================================================================
 
-import numpy as np
-
 from onedal._device_offload import supports_queue
 from onedal.common._backend import bind_default_backend
 from onedal.utils import _sycl_queue_manager as QM
 
-from .._config import _get_config
 from ..datatypes import from_table, return_type_constructor, to_table
 from .pca import PCA
 
diff --git a/onedal/decomposition/pca.py b/onedal/decomposition/pca.py
@@ -14,11 +14,8 @@
 # limitations under the License.
 # ==============================================================================
 
-import numbers
 from abc import ABCMeta
 
-import numpy as np
-
 from onedal._device_offload import supports_queue
 from onedal.common._backend import bind_default_backend
 
diff --git a/onedal/neighbors/neighbors.py b/onedal/neighbors/neighbors.py
@@ -472,7 +472,7 @@ def predict_proba(self, X, queue=None):
             _y = self._y.reshape((-1, 1))
             classes_ = [self.classes_]
 
-        n_queries = _num_samples(X)
+        n_queries = _num_samples(X if X is not None else self._fit_X)
 
         weights = self._get_weights(neigh_dist, self.weights)
         if weights is None:
diff --git a/requirements-test.txt b/requirements-test.txt
@@ -1,5 +1,5 @@
 pytest==7.4.4 ; python_version <= '3.10'
-pytest==9.0.1 ; python_version >= '3.11'
+pytest==9.0.2 ; python_version >= '3.11'
 pytest-json-report==1.5.0
 pytest-cov==7.0.0
 numpy>=1.19.5 ; python_version <= '3.9'
@@ -13,7 +13,7 @@ pandas==2.3.3 ; python_version >= '3.11'
 xgboost==2.1.4 ; python_version < '3.10'
 xgboost==3.1.2 ; python_version >= '3.10'
 lightgbm==4.6.0
-catboost==1.2.8
+catboost==1.2.8 ; python_version < '3.14' # TODO: Remove condition when catboost releases py3.14 package
 shap==0.49.1 ; python_version < '3.11'
 shap==0.50.0 ; python_version >= '3.11'
 treelite==4.6.1
diff --git a/sklearnex/covariance/incremental_covariance.py b/sklearnex/covariance/incremental_covariance.py
@@ -19,7 +19,6 @@
 from functools import partial
 
 import numpy as np
-from scipy import linalg
 from sklearn.base import BaseEstimator, clone
 from sklearn.covariance import EmpiricalCovariance as _sklearn_EmpiricalCovariance
 from sklearn.utils import gen_batches
diff --git a/sklearnex/decomposition/pca.py b/sklearnex/decomposition/pca.py
@@ -19,7 +19,6 @@
 from daal4py.sklearn._utils import daal_check_version
 
 if daal_check_version((2024, "P", 100)):
-    from math import sqrt
     from numbers import Integral
     from warnings import warn
 
diff --git a/sklearnex/linear_model/linear.py b/sklearnex/linear_model/linear.py
@@ -23,7 +23,6 @@
 
 from daal4py.sklearn._n_jobs_support import control_n_jobs
 from daal4py.sklearn._utils import daal_check_version, sklearn_check_version
-from onedal.common.hyperparameters import get_hyperparameters
 from onedal.linear_model import LinearRegression as onedal_LinearRegression
 from onedal.utils.validation import _num_features, _num_samples
 
diff --git a/sklearnex/neighbors/common.py b/sklearnex/neighbors/common.py
@@ -156,6 +156,19 @@ def _onedal_supported(self, device, method_name, *data):
         patching_status = PatchingConditionsChain(
             f"sklearn.neighbors.{class_name}.{method_name}"
         )
+        # TODO: with verbosity enabled, here it would emit a log saying that it fell
+        # back to sklearn, but internally, sklearn will end up calling 'kneighbors'
+        # which is overridden in the sklearnex classes, thus it will end up calling
+        # oneDAL in the end, but the log will say otherwise. Find a way to make the
+        # log consistent with what happens in practice.
+        patching_status.and_conditions(
+            [
+                (
+                    not (data[0] is None and method_name in ["predict", "score"]),
+                    "Predictions on 'None' data are handled by internal sklearn methods.",
+                )
+            ]
+        )
         if not patching_status.and_condition(
             "radius" not in method_name, "RadiusNeighbors not implemented in sklearnex"
         ):
diff --git a/sklearnex/neighbors/knn_classification.py b/sklearnex/neighbors/knn_classification.py
@@ -79,7 +79,8 @@ def fit(self, X, y):
     @wrap_output_data
     def predict(self, X):
         check_is_fitted(self)
-        check_feature_names(self, X, reset=False)
+        if X is not None:
+            check_feature_names(self, X, reset=False)
         return dispatch(
             self,
             "predict",
@@ -93,7 +94,8 @@ def predict(self, X):
     @wrap_output_data
     def predict_proba(self, X):
         check_is_fitted(self)
-        check_feature_names(self, X, reset=False)
+        if X is not None:
+            check_feature_names(self, X, reset=False)
         return dispatch(
             self,
             "predict_proba",
@@ -107,7 +109,8 @@ def predict_proba(self, X):
     @wrap_output_data
     def score(self, X, y, sample_weight=None):
         check_is_fitted(self)
-        check_feature_names(self, X, reset=False)
+        if X is not None:
+            check_feature_names(self, X, reset=False)
         return dispatch(
             self,
             "score",
diff --git a/sklearnex/preview/decomposition/incremental_pca.py b/sklearnex/preview/decomposition/incremental_pca.py
@@ -14,7 +14,6 @@
 # limitations under the License.
 # ===============================================================================
 
-import numpy as np
 import scipy.sparse as sp
 from sklearn.decomposition import IncrementalPCA as _sklearn_IncrementalPCA
 from sklearn.utils import check_array, gen_batches
diff --git a/sklearnex/tests/test_common.py b/sklearnex/tests/test_common.py
@@ -24,7 +24,7 @@
 import sys
 import trace
 from contextlib import redirect_stdout
-from multiprocessing import Pipe, Process, get_context
+from multiprocessing import get_context
 
 import pytest
 from sklearn.base import BaseEstimator
diff --git a/sklearnex/utils/_array_api.py b/sklearnex/utils/_array_api.py
@@ -14,7 +14,7 @@
 # limitations under the License.
 # ==============================================================================
 
-"""Tools to support array_api."""
+"""Tools to support array API."""
 
 import math
 from collections.abc import Callable
@@ -27,6 +27,7 @@
 from daal4py.sklearn._utils import sklearn_check_version
 from onedal.utils._array_api import _get_sycl_namespace, _is_numpy_namespace
 
+from .._config import get_config
 from ..base import oneDALEstimator
 
 if sklearn_check_version("1.6"):
@@ -83,11 +84,19 @@ def get_namespace(*arrays):
         True of the arrays are containers that implement the Array API spec.
     """
 
-    sycl_type, xp, is_array_api_compliant = _get_sycl_namespace(*arrays)
-
-    if sycl_type:
-        return xp, is_array_api_compliant
-    elif sklearn_check_version("1.2"):
+    # check required because _get_sycl_namespace only verifies that *arrays
+    # are of the same sycl namespace, not of the same array namespace.
+    # When array_api_dispatch is enabled, then sklearn's version is required
+    # for the additional array namespace check. This is now possible with
+    # dpnp and dpctl as they both support `__array_namespace__`.
+    if not get_config().get("array_api_dispatch", False):
+        sycl_type, xp, is_array_api_compliant = _get_sycl_namespace(*arrays)
+        if sycl_type:
+            return xp, is_array_api_compliant
+
+    # sklearn contains a specially patched numpy wrapper that should be
+    # reused which is yielded from sklearn's get_namespace.
+    if sklearn_check_version("1.2"):
         return sklearn_get_namespace(*arrays)
     else:
         return np, False
