scikit-learn-intelex 2024.6.0__py312-none-manylinux1_x86_64.whl → 2024.7.0__py312-none-manylinux1_x86_64.whl

sklearnex/neighbors/common.py

@@ -19,14 +19,16 @@ import warnings
 import numpy as np
 from scipy import sparse as sp
 from sklearn.neighbors._ball_tree import BallTree
-from sklearn.neighbors._base import VALID_METRICS
+from sklearn.neighbors._base import VALID_METRICS, KNeighborsMixin
 from sklearn.neighbors._base import NeighborsBase as sklearn_NeighborsBase
 from sklearn.neighbors._kd_tree import KDTree
+from sklearn.utils.validation import check_is_fitted
 
 from daal4py.sklearn._utils import sklearn_check_version
 from onedal.utils import _check_array, _num_features, _num_samples
 
 from .._utils import PatchingConditionsChain
+from ..utils import get_namespace
 
 
 class KNeighborsDispatchingBase:
@@ -147,6 +149,10 @@ class KNeighborsDispatchingBase:
         patching_status = PatchingConditionsChain(
             f"sklearn.neighbors.{class_name}.{method_name}"
         )
+        if not patching_status.and_condition(
+            "radius" not in method_name, "RadiusNeighbors not implemented in sklearnex"
+        ):
+            return patching_status
 
         if not patching_status.and_condition(
             not isinstance(data[0], (KDTree, BallTree, sklearn_NeighborsBase)),
@@ -264,3 +270,41 @@ class KNeighborsDispatchingBase:
 
     def _onedal_cpu_supported(self, method_name, *data):
         return self._onedal_supported("cpu", method_name, *data)
+
+    def kneighbors_graph(self, X=None, n_neighbors=None, mode="connectivity"):
+        check_is_fitted(self)
+        if n_neighbors is None:
+            n_neighbors = self.n_neighbors
+
+        # check the input only in self.kneighbors
+
+        # construct CSR matrix representation of the k-NN graph
+        if mode == "connectivity":
+            A_ind = self.kneighbors(X, n_neighbors, return_distance=False)
+            xp, _ = get_namespace(A_ind)
+            n_queries = A_ind.shape[0]
+            A_data = xp.ones(n_queries * n_neighbors)
+
+        elif mode == "distance":
+            A_data, A_ind = self.kneighbors(X, n_neighbors, return_distance=True)
+            xp, _ = get_namespace(A_ind)
+            A_data = xp.reshape(A_data, (-1,))
+
+        else:
+            raise ValueError(
+                'Unsupported mode, must be one of "connectivity", '
+                f'or "distance" but got "{mode}" instead'
+            )
+
+        n_queries = A_ind.shape[0]
+        n_samples_fit = self.n_samples_fit_
+        n_nonzero = n_queries * n_neighbors
+        A_indptr = xp.arange(0, n_nonzero + 1, n_neighbors)
+
+        kneighbors_graph = sp.csr_matrix(
+            (A_data, xp.reshape(A_ind, (-1,)), A_indptr), shape=(n_queries, n_samples_fit)
+        )
+
+        return kneighbors_graph
+
+    kneighbors_graph.__doc__ = KNeighborsMixin.kneighbors_graph.__doc__

sklearnex/neighbors/knn_classification.py

@@ -32,7 +32,7 @@ from .common import KNeighborsDispatchingBase
 @control_n_jobs(
     decorated_methods=["fit", "predict", "predict_proba", "kneighbors", "score"]
 )
-class KNeighborsClassifier(sklearn_KNeighborsClassifier, KNeighborsDispatchingBase):
+class KNeighborsClassifier(KNeighborsDispatchingBase, sklearn_KNeighborsClassifier):
     __doc__ = sklearn_KNeighborsClassifier.__doc__
     if sklearn_check_version("1.2"):
         _parameter_constraints: dict = {
@@ -169,24 +169,6 @@ class KNeighborsClassifier(sklearn_KNeighborsClassifier, KNeighborsDispatchingBa
             return_distance=return_distance,
         )
 
-    @wrap_output_data
-    def radius_neighbors(
-        self, X=None, radius=None, return_distance=True, sort_results=False
-    ):
-        _onedal_estimator = getattr(self, "_onedal_estimator", None)
-
-        if (
-            _onedal_estimator is not None
-            or getattr(self, "_tree", 0) is None
-            and self._fit_method == "kd_tree"
-        ):
-            sklearn_NearestNeighbors.fit(self, self._fit_X, getattr(self, "_y", None))
-        result = sklearn_NearestNeighbors.radius_neighbors(
-            self, X, radius, return_distance, sort_results
-        )
-
-        return result
-
     def _onedal_fit(self, X, y, queue=None):
         onedal_params = {
             "n_neighbors": self.n_neighbors,
@@ -242,4 +224,3 @@ class KNeighborsClassifier(sklearn_KNeighborsClassifier, KNeighborsDispatchingBa
     predict_proba.__doc__ = sklearn_KNeighborsClassifier.predict_proba.__doc__
     score.__doc__ = sklearn_KNeighborsClassifier.score.__doc__
     kneighbors.__doc__ = sklearn_KNeighborsClassifier.kneighbors.__doc__
-    radius_neighbors.__doc__ = sklearn_NearestNeighbors.radius_neighbors.__doc__

sklearnex/neighbors/knn_regression.py

@@ -30,7 +30,7 @@ from .common import KNeighborsDispatchingBase
 
 
 @control_n_jobs(decorated_methods=["fit", "predict", "kneighbors"])
-class KNeighborsRegressor(sklearn_KNeighborsRegressor, KNeighborsDispatchingBase):
+class KNeighborsRegressor(KNeighborsDispatchingBase, sklearn_KNeighborsRegressor):
     __doc__ = sklearn_KNeighborsRegressor.__doc__
     if sklearn_check_version("1.2"):
         _parameter_constraints: dict = {
@@ -152,24 +152,6 @@ class KNeighborsRegressor(sklearn_KNeighborsRegressor, KNeighborsDispatchingBase
             return_distance=return_distance,
         )
 
-    @wrap_output_data
-    def radius_neighbors(
-        self, X=None, radius=None, return_distance=True, sort_results=False
-    ):
-        _onedal_estimator = getattr(self, "_onedal_estimator", None)
-
-        if (
-            _onedal_estimator is not None
-            or getattr(self, "_tree", 0) is None
-            and self._fit_method == "kd_tree"
-        ):
-            sklearn_NearestNeighbors.fit(self, self._fit_X, getattr(self, "_y", None))
-        result = sklearn_NearestNeighbors.radius_neighbors(
-            self, X, radius, return_distance, sort_results
-        )
-
-        return result
-
     def _onedal_fit(self, X, y, queue=None):
         onedal_params = {
             "n_neighbors": self.n_neighbors,
@@ -218,5 +200,4 @@ class KNeighborsRegressor(sklearn_KNeighborsRegressor, KNeighborsDispatchingBase
     fit.__doc__ = sklearn_KNeighborsRegressor.__doc__
     predict.__doc__ = sklearn_KNeighborsRegressor.predict.__doc__
     kneighbors.__doc__ = sklearn_KNeighborsRegressor.kneighbors.__doc__
-    radius_neighbors.__doc__ = sklearn_NearestNeighbors.radius_neighbors.__doc__
     score.__doc__ = sklearn_KNeighborsRegressor.score.__doc__

sklearnex/neighbors/knn_unsupervised.py

@@ -26,7 +26,7 @@ from .common import KNeighborsDispatchingBase
 
 
 @control_n_jobs(decorated_methods=["fit", "kneighbors"])
-class NearestNeighbors(sklearn_NearestNeighbors, KNeighborsDispatchingBase):
+class NearestNeighbors(KNeighborsDispatchingBase, sklearn_NearestNeighbors):
     __doc__ = sklearn_NearestNeighbors.__doc__
     if sklearn_check_version("1.2"):
         _parameter_constraints: dict = {**sklearn_NearestNeighbors._parameter_constraints}
@@ -88,19 +88,40 @@ class NearestNeighbors(sklearn_NearestNeighbors, KNeighborsDispatchingBase):
     def radius_neighbors(
         self, X=None, radius=None, return_distance=True, sort_results=False
     ):
-        _onedal_estimator = getattr(self, "_onedal_estimator", None)
-
         if (
-            _onedal_estimator is not None
+            hasattr(self, "_onedal_estimator")
             or getattr(self, "_tree", 0) is None
             and self._fit_method == "kd_tree"
         ):
             sklearn_NearestNeighbors.fit(self, self._fit_X, getattr(self, "_y", None))
-        result = sklearn_NearestNeighbors.radius_neighbors(
-            self, X, radius, return_distance, sort_results
+        return dispatch(
+            self,
+            "radius_neighbors",
+            {
+                "onedal": None,
+                "sklearn": sklearn_NearestNeighbors.radius_neighbors,
+            },
+            X,
+            radius=radius,
+            return_distance=return_distance,
+            sort_results=sort_results,
         )
 
-        return result
+    def radius_neighbors_graph(
+        self, X=None, radius=None, mode="connectivity", sort_results=False
+    ):
+        return dispatch(
+            self,
+            "radius_neighbors_graph",
+            {
+                "onedal": None,
+                "sklearn": sklearn_NearestNeighbors.radius_neighbors_graph,
+            },
+            X,
+            radius=radius,
+            mode=mode,
+            sort_results=sort_results,
+        )
 
     def _onedal_fit(self, X, y=None, queue=None):
         onedal_params = {
@@ -144,3 +165,6 @@ class NearestNeighbors(sklearn_NearestNeighbors, KNeighborsDispatchingBase):
     fit.__doc__ = sklearn_NearestNeighbors.__doc__
     kneighbors.__doc__ = sklearn_NearestNeighbors.kneighbors.__doc__
     radius_neighbors.__doc__ = sklearn_NearestNeighbors.radius_neighbors.__doc__
+    radius_neighbors_graph.__doc__ = (
+        sklearn_NearestNeighbors.radius_neighbors_graph.__doc__
+    )

sklearnex/preview/__init__.py

@@ -14,4 +14,4 @@
 # limitations under the License.
 # ==============================================================================
 
-__all__ = ["cluster", "covariance", "decomposition"]
+__all__ = ["cluster", "covariance", "decomposition", "linear_model"]

sklearnex/preview/linear_model/__init__.py

@@ -0,0 +1,19 @@
+# ===============================================================================
+# Copyright 2024 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ===============================================================================
+
+from .ridge import Ridge
+
+__all__ = ["Ridge"]

sklearnex/preview/linear_model/ridge.py

@@ -0,0 +1,419 @@
+# ===============================================================================
+# Copyright 2024 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ===============================================================================
+
+import logging
+
+from daal4py.sklearn._utils import daal_check_version, sklearn_check_version
+
+if daal_check_version((2024, "P", 600)):
+    import numbers
+
+    import numpy as np
+    from scipy.sparse import issparse
+    from sklearn.linear_model import Ridge as sklearn_Ridge
+    from sklearn.metrics import r2_score
+    from sklearn.utils.validation import check_is_fitted, check_X_y
+
+    from daal4py.sklearn.linear_model._ridge import _fit_ridge as daal4py_fit_ridge
+
+    if sklearn_check_version("1.0") and not sklearn_check_version("1.2"):
+        from sklearn.linear_model._base import _deprecate_normalize
+    if sklearn_check_version("1.1") and not sklearn_check_version("1.2"):
+        from sklearn.utils import check_scalar
+
+    from onedal.linear_model import Ridge as onedal_Ridge
+    from onedal.utils import _num_features, _num_samples
+
+    from ..._device_offload import dispatch, wrap_output_data
+    from ..._utils import PatchingConditionsChain
+
+    def _is_numeric_scalar(value):
+        """
+        Determines if the provided value is a numeric scalar.
+
+        Args:
+        value: The value to be checked.
+
+        Returns:
+        bool: True if the value is a numeric scalar, False otherwise.
+        """
+        return isinstance(value, numbers.Real)
+
+    class Ridge(sklearn_Ridge):
+        __doc__ = sklearn_Ridge.__doc__
+
+        if sklearn_check_version("1.2"):
+            _parameter_constraints: dict = {**sklearn_Ridge._parameter_constraints}
+
+            def __init__(
+                self,
+                alpha=1.0,
+                fit_intercept=True,
+                copy_X=True,
+                max_iter=None,
+                tol=1e-4,
+                solver="auto",
+                positive=False,
+                random_state=None,
+            ):
+                super().__init__(
+                    alpha=alpha,
+                    fit_intercept=fit_intercept,
+                    copy_X=copy_X,
+                    max_iter=max_iter,
+                    tol=tol,
+                    solver=solver,
+                    positive=positive,
+                    random_state=random_state,
+                )
+
+        elif sklearn_check_version("1.0"):
+
+            def __init__(
+                self,
+                alpha=1.0,
+                fit_intercept=True,
+                normalize="deprecated",
+                copy_X=True,
+                max_iter=None,
+                tol=1e-3,
+                solver="auto",
+                positive=False,
+                random_state=None,
+            ):
+                super().__init__(
+                    alpha=alpha,
+                    fit_intercept=fit_intercept,
+                    normalize=normalize,
+                    copy_X=copy_X,
+                    max_iter=max_iter,
+                    solver=solver,
+                    tol=tol,
+                    positive=positive,
+                    random_state=random_state,
+                )
+
+        else:
+
+            def __init__(
+                self,
+                alpha=1.0,
+                fit_intercept=True,
+                normalize=False,
+                copy_X=True,
+                max_iter=None,
+                tol=1e-3,
+                solver="auto",
+                random_state=None,
+            ):
+                super().__init__(
+                    alpha=alpha,
+                    fit_intercept=fit_intercept,
+                    normalize=normalize,
+                    copy_X=copy_X,
+                    max_iter=max_iter,
+                    tol=tol,
+                    solver=solver,
+                    random_state=random_state,
+                )
+
+        def fit(self, X, y, sample_weight=None):
+            # It is necessary to properly update coefs for predict if we
+            # fallback to sklearn in dispatch
+            if hasattr(self, "_onedal_estimator"):
+                del self._onedal_estimator
+
+            dispatch(
+                self,
+                "fit",
+                {
+                    "onedal": self.__class__._onedal_fit,
+                    "sklearn": sklearn_Ridge.fit,
+                },
+                X,
+                y,
+                sample_weight,
+            )
+            return self
+
+        @wrap_output_data
+        def predict(self, X):
+            check_is_fitted(self)
+
+            return dispatch(
+                self,
+                "predict",
+                {
+                    "onedal": self.__class__._onedal_predict,
+                    "sklearn": sklearn_Ridge.predict,
+                },
+                X,
+            )
+
+        @wrap_output_data
+        def score(self, X, y, sample_weight=None):
+            check_is_fitted(self)
+
+            return dispatch(
+                self,
+                "score",
+                {
+                    "onedal": self.__class__._onedal_score,
+                    "sklearn": sklearn_Ridge.score,
+                },
+                X,
+                y,
+                sample_weight=sample_weight,
+            )
+
+        def _onedal_fit_supported(self, patching_status, method_name, *data):
+            assert method_name == "fit"
+            assert len(data) == 3
+            X, y, sample_weight = data
+
+            normalize_is_set = (
+                hasattr(self, "normalize")
+                and self.normalize
+                and self.normalize != "deprecated"
+            )
+            positive_is_set = hasattr(self, "positive") and self.positive
+
+            n_samples = _num_samples(X)
+            n_features = _num_features(X, fallback_1d=True)
+
+            # Check if equations are well defined
+            is_underdetermined = n_samples < (n_features + int(self.fit_intercept))
+
+            patching_status.and_conditions(
+                [
+                    (
+                        self.solver == "auto",
+                        f"'{self.solver}' solver is not supported. "
+                        "Only 'auto' solver is supported.",
+                    ),
+                    (
+                        not issparse(X) and not issparse(y),
+                        "Sparse input is not supported.",
+                    ),
+                    (
+                        not is_underdetermined,
+                        "The shape of X (fitting) does not satisfy oneDAL requirements:"
+                        "Number of features + 1 >= number of samples.",
+                    ),
+                    (sample_weight is None, "Sample weight is not supported."),
+                    (not normalize_is_set, "Normalization is not supported."),
+                    (
+                        not positive_is_set,
+                        "Forced positive coefficients are not supported.",
+                    ),
+                ]
+            )
+
+            return patching_status
+
+        def _onedal_predict_supported(self, patching_status, method_name, *data):
+            assert method_name in ["predict", "score"]
+            assert len(data) <= 2
+
+            n_samples = _num_samples(data[0])
+            model_is_sparse = issparse(self.coef_) or (
+                self.fit_intercept and issparse(self.intercept_)
+            )
+            patching_status.and_conditions(
+                [
+                    (
+                        self.solver == "auto",
+                        f"'{self.solver}' solver is not supported. "
+                        "Only 'auto' solver is supported.",
+                    ),
+                    (n_samples > 0, "Number of samples is less than 1."),
+                    (not issparse(data[0]), "Sparse input is not supported."),
+                    (not model_is_sparse, "Sparse coefficients are not supported."),
+                ]
+            )
+
+            return patching_status
+
+        def _onedal_gpu_supported(self, method_name, *data):
+            patching_status = PatchingConditionsChain(
+                f"sklearn.linear_model.{self.__class__.__name__}.fit"
+            )
+
+            if method_name == "fit":
+                patching_status.and_condition(
+                    _is_numeric_scalar(self.alpha),
+                    "Non-scalar alpha is not supported for GPU.",
+                )
+
+                return self._onedal_fit_supported(patching_status, method_name, *data)
+
+            if method_name in ["predict", "score"]:
+                return self._onedal_predict_supported(patching_status, method_name, *data)
+
+            raise RuntimeError(
+                f"Unknown method {method_name} in {self.__class__.__name__}"
+            )
+
+        def _onedal_cpu_supported(self, method_name, *data):
+            patching_status = PatchingConditionsChain(
+                f"sklearn.linear_model.{self.__class__.__name__}.fit"
+            )
+
+            if method_name == "fit":
+                return self._onedal_fit_supported(patching_status, method_name, *data)
+
+            if method_name in ["predict", "score"]:
+                return self._onedal_predict_supported(patching_status, method_name, *data)
+
+            raise RuntimeError(
+                f"Unknown method {method_name} in {self.__class__.__name__}"
+            )
+
+        def _initialize_onedal_estimator(self):
+            onedal_params = {
+                "fit_intercept": self.fit_intercept,
+                "alpha": self.alpha,
+                "copy_X": self.copy_X,
+            }
+            self._onedal_estimator = onedal_Ridge(**onedal_params)
+
+        def _daal_fit(self, X, y, sample_weight=None):
+            daal4py_fit_ridge(self, X, y, sample_weight)
+            self._onedal_estimator.n_features_in_ = _num_features(X, fallback_1d=True)
+            self._onedal_estimator.coef_ = self.coef_
+            self._onedal_estimator.intercept_ = self.intercept_
+
+        def _onedal_fit(self, X, y, sample_weight, queue=None):
+            # `Sample weight` is not supported. Expected to be None value.
+            assert sample_weight is None
+
+            if sklearn_check_version("1.2"):
+                self._validate_params()
+            elif sklearn_check_version("1.1"):
+                if self.max_iter is not None:
+                    self.max_iter = check_scalar(
+                        self.max_iter, "max_iter", target_type=numbers.Integral, min_val=1
+                    )
+                self.tol = check_scalar(
+                    self.tol, "tol", target_type=numbers.Real, min_val=0.0
+                )
+                if self.alpha is not None and not isinstance(
+                    self.alpha, (np.ndarray, tuple)
+                ):
+                    self.alpha = check_scalar(
+                        self.alpha,
+                        "alpha",
+                        target_type=numbers.Real,
+                        min_val=0.0,
+                        include_boundaries="left",
+                    )
+
+            check_params = {
+                "X": X,
+                "y": y,
+                "dtype": [np.float64, np.float32],
+                "accept_sparse": ["csr", "csc", "coo"],
+                "y_numeric": True,
+                "multi_output": True,
+            }
+            if sklearn_check_version("1.0"):
+                X, y = self._validate_data(**check_params)
+            else:
+                X, y = check_X_y(**check_params)
+
+            if sklearn_check_version("1.0") and not sklearn_check_version("1.2"):
+                self._normalize = _deprecate_normalize(
+                    self.normalize,
+                    default=False,
+                    estimator_name=self.__class__.__name__,
+                )
+
+            self._initialize_onedal_estimator()
+
+            # Falling back to daal4py if the device is CPU and alpha is array-like
+            # since onedal does not yet support non-scalars for alpha, thus
+            # should only be used for GPU/CPU with scalar alpha to not limit the functionality
+            cpu_device = queue is None or queue.sycl_device.is_cpu
+            if cpu_device and not _is_numeric_scalar(self.alpha):
+                self._daal_fit(X, y)
+            else:
+                self._onedal_estimator.fit(X, y, queue=queue)
+
+            self._save_attributes()
+
+        def _onedal_predict(self, X, queue=None):
+            if sklearn_check_version("1.0"):
+                X = self._validate_data(X, accept_sparse=False, reset=False)
+
+            if not hasattr(self, "_onedal_estimator"):
+                self._initialize_onedal_estimator()
+                self._onedal_estimator.coef_ = self.coef_
+                self._onedal_estimator.intercept_ = self.intercept_
+
+            res = self._onedal_estimator.predict(X, queue=queue)
+            return res
+
+        def _onedal_score(self, X, y, sample_weight=None, queue=None):
+            return r2_score(
+                y, self._onedal_predict(X, queue=queue), sample_weight=sample_weight
+            )
+
+        @property
+        def coef_(self):
+            return self._coef
+
+        @coef_.setter
+        def coef_(self, value):
+            if hasattr(self, "_onedal_estimator"):
+                self._onedal_estimator.coef_ = value
+                # checking if the model is already fitted and if so, deleting the model
+                if hasattr(self._onedal_estimator, "_onedal_model"):
+                    del self._onedal_estimator._onedal_model
+            self._coef = value
+
+        @property
+        def intercept_(self):
+            return self._intercept
+
+        @intercept_.setter
+        def intercept_(self, value):
+            if hasattr(self, "_onedal_estimator"):
+                self._onedal_estimator.intercept_ = value
+                # checking if the model is already fitted and if so, deleting the model
+                if hasattr(self._onedal_estimator, "_onedal_model"):
+                    del self._onedal_estimator._onedal_model
+            self._intercept = value
+
+        def _save_attributes(self):
+            self.n_features_in_ = self._onedal_estimator.n_features_in_
+            self._coef = self._onedal_estimator.coef_
+            self._intercept = self._onedal_estimator.intercept_
+
+        fit.__doc__ = sklearn_Ridge.fit.__doc__
+        predict.__doc__ = sklearn_Ridge.predict.__doc__
+        score.__doc__ = sklearn_Ridge.score.__doc__
+
+else:
+    from daal4py.sklearn.linear_model._ridge import Ridge
+    from onedal._device_offload import support_usm_ndarray
+
+    Ridge.fit = support_usm_ndarray(queue_param=False)(Ridge.fit)
+    Ridge.predict = support_usm_ndarray(queue_param=False)(Ridge.predict)
+    Ridge.score = support_usm_ndarray(queue_param=False)(Ridge.score)
+
+    logging.warning(
+        "Preview Ridge requires oneDAL version >= 2024.6 but it was not found"
+    )