scikit-learn-intelex 2024.5.0__py39-none-manylinux1_x86_64.whl → 2024.7.0__py39-none-manylinux1_x86_64.whl

sklearnex/linear_model/tests/test_incremental_linear.py

@@ -47,7 +47,7 @@ def test_sklearnex_fit_on_gold_data(dataframe, queue, fit_intercept, macro_block
 
     y_pred = inclin.predict(X_df)
 
-    tol = 2e-6 if dtype == np.float32 else 1e-7
+    tol = 2e-6 if y_pred.dtype == np.float32 else 1e-7
     assert_allclose(inclin.coef_, [1], atol=tol)
     if fit_intercept:
         assert_allclose(inclin.intercept_, [0], atol=tol)
@@ -82,15 +82,15 @@ def test_sklearnex_partial_fit_on_gold_data(
         )
         inclin.partial_fit(X_split_df, y_split_df)
 
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    y_pred = inclin.predict(X_df)
+
     assert inclin.n_features_in_ == 1
-    tol = 2e-6 if dtype == np.float32 else 1e-7
+    tol = 2e-6 if y_pred.dtype == np.float32 else 1e-7
     assert_allclose(inclin.coef_, [[1]], atol=tol)
     if fit_intercept:
         assert_allclose(inclin.intercept_, 3, atol=tol)
 
-    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
-    y_pred = inclin.predict(X_df)
-
     assert_allclose(_as_numpy(y_pred), y, atol=tol)
 
 
@@ -122,15 +122,15 @@ def test_sklearnex_partial_fit_multitarget_on_gold_data(
         )
         inclin.partial_fit(X_split_df, y_split_df)
 
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    y_pred = inclin.predict(X_df)
+
     assert inclin.n_features_in_ == 2
-    tol = 7e-6 if dtype == np.float32 else 1e-7
+    tol = 7e-6 if y_pred.dtype == np.float32 else 1e-7
     assert_allclose(inclin.coef_, [1.0, 2.0], atol=tol)
     if fit_intercept:
         assert_allclose(inclin.intercept_, 3.0, atol=tol)
 
-    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
-    y_pred = inclin.predict(X_df)
-
     assert_allclose(_as_numpy(y_pred), y, atol=tol)
 
 
@@ -181,7 +181,7 @@ def test_sklearnex_partial_fit_on_random_data(
         )
         inclin.partial_fit(X_split_df, y_split_df)
 
-    tol = 1e-4 if dtype == np.float32 else 1e-7
+    tol = 1e-4 if inclin.coef_.dtype == np.float32 else 1e-7
     assert_allclose(coef, inclin.coef_.T, atol=tol)
 
     if fit_intercept:

sklearnex/linear_model/tests/test_linear.py

@@ -20,6 +20,10 @@ from numpy.testing import assert_allclose
 from sklearn.datasets import make_regression
 
 from daal4py.sklearn._utils import daal_check_version
+from daal4py.sklearn.linear_model.tests.test_ridge import (
+    _test_multivariate_ridge_alpha_shape,
+    _test_multivariate_ridge_coefficients,
+)
 from onedal.tests.utils._dataframes_support import (
     _as_numpy,
     _convert_to_dataframe,
@@ -52,37 +56,46 @@ def test_sklearnex_import_linear(dataframe, queue, dtype, macro_block):
     assert "sklearnex" in linreg.__module__
     assert linreg.n_features_in_ == 2
 
-    tol = 1e-5 if dtype == np.float32 else 1e-7
+    tol = 1e-5 if _as_numpy(linreg.coef_).dtype == np.float32 else 1e-7
     assert_allclose(_as_numpy(linreg.intercept_), 3.0, rtol=tol)
     assert_allclose(_as_numpy(linreg.coef_), [1.0, 2.0], rtol=tol)
 
 
-def test_sklearnex_import_ridge():
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+def test_sklearnex_import_ridge(dataframe, queue):
     from sklearnex.linear_model import Ridge
 
     X = np.array([[1, 1], [1, 2], [2, 2], [2, 3]])
     y = np.dot(X, np.array([1, 2])) + 3
+    X = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    y = _convert_to_dataframe(y, sycl_queue=queue, target_df=dataframe)
     ridgereg = Ridge().fit(X, y)
     assert "daal4py" in ridgereg.__module__
     assert_allclose(ridgereg.intercept_, 4.5)
     assert_allclose(ridgereg.coef_, [0.8, 1.4])
 
 
-def test_sklearnex_import_lasso():
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+def test_sklearnex_import_lasso(dataframe, queue):
     from sklearnex.linear_model import Lasso
 
     X = [[0, 0], [1, 1], [2, 2]]
     y = [0, 1, 2]
+    X = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    y = _convert_to_dataframe(y, sycl_queue=queue, target_df=dataframe)
     lasso = Lasso(alpha=0.1).fit(X, y)
     assert "daal4py" in lasso.__module__
     assert_allclose(lasso.intercept_, 0.15)
     assert_allclose(lasso.coef_, [0.85, 0.0])
 
 
-def test_sklearnex_import_elastic():
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+def test_sklearnex_import_elastic(dataframe, queue):
     from sklearnex.linear_model import ElasticNet
 
     X, y = make_regression(n_features=2, random_state=0)
+    X = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    y = _convert_to_dataframe(y, sycl_queue=queue, target_df=dataframe)
     elasticnet = ElasticNet(random_state=0).fit(X, y)
     assert "daal4py" in elasticnet.__module__
     assert_allclose(elasticnet.intercept_, 1.451, atol=1e-3)
@@ -113,5 +126,17 @@ def test_sklearnex_reconstruct_model(dataframe, queue, dtype):
 
     y_pred = linreg.predict(X)
 
-    tol = 1e-5 if dtype == np.float32 else 1e-7
+    tol = 1e-5 if _as_numpy(y_pred).dtype == np.float32 else 1e-7
     assert_allclose(gtr, _as_numpy(y_pred), rtol=tol)
+
+
+def test_sklearnex_multivariate_ridge_coefs():
+    from sklearnex.linear_model import Ridge
+
+    _test_multivariate_ridge_coefficients(Ridge, random_state=0)
+
+
+def test_sklearnex_multivariate_ridge_alpha_shape():
+    from sklearnex.linear_model import Ridge
+
+    _test_multivariate_ridge_alpha_shape(Ridge, random_state=0)

sklearnex/linear_model/tests/test_logreg.py

@@ -14,8 +14,11 @@
 # limitations under the License.
 # ===============================================================================
 
+import numpy as np
 import pytest
-from sklearn.datasets import load_breast_cancer, load_iris
+from numpy.testing import assert_allclose, assert_array_equal
+from scipy.sparse import csr_matrix
+from sklearn.datasets import load_breast_cancer, load_iris, make_classification
 from sklearn.metrics import accuracy_score
 from sklearn.model_selection import train_test_split
 
@@ -24,7 +27,9 @@ from onedal.tests.utils._dataframes_support import (
     _as_numpy,
     _convert_to_dataframe,
     get_dataframes_and_queues,
+    get_queues,
 )
+from sklearnex import config_context
 
 
 def prepare_input(X, y, dataframe, queue):
@@ -38,8 +43,7 @@ def prepare_input(X, y, dataframe, queue):
 
 
 @pytest.mark.parametrize(
-    "dataframe,queue",
-    get_dataframes_and_queues(device_filter_="cpu"),
+    "dataframe,queue", get_dataframes_and_queues(device_filter_="cpu")
 )
 def test_sklearnex_multiclass_classification(dataframe, queue):
     from sklearnex.linear_model import LogisticRegression
@@ -89,3 +93,41 @@ def test_sklearnex_binary_classification(dataframe, queue):
 
     y_pred = _as_numpy(logreg.predict(X_test))
     assert accuracy_score(y_test, y_pred) > 0.95
+
+
+if daal_check_version((2024, "P", 700)):
+
+    @pytest.mark.parametrize("queue", get_queues("gpu"))
+    @pytest.mark.parametrize("dtype", [np.float32, np.float64])
+    @pytest.mark.parametrize(
+        "dims", [(3007, 17, 0.05), (50000, 100, 0.01), (512, 10, 0.5)]
+    )
+    def test_csr(queue, dtype, dims):
+        from sklearnex.linear_model import LogisticRegression
+
+        n, p, density = dims
+
+        # Create sparse dataset for classification
+        X, y = make_classification(n, p, random_state=42)
+        X = X.astype(dtype)
+        y = y.astype(dtype)
+        np.random.seed(2007 + n + p)
+        mask = np.random.binomial(1, density, (n, p))
+        X = X * mask
+        X_sp = csr_matrix(X)
+
+        model = LogisticRegression(fit_intercept=True, solver="newton-cg")
+        model_sp = LogisticRegression(fit_intercept=True, solver="newton-cg")
+
+        with config_context(target_offload="gpu:0"):
+            model.fit(X, y)
+            pred = model.predict(X)
+            prob = model.predict_proba(X)
+            model_sp.fit(X_sp, y)
+            pred_sp = model_sp.predict(X_sp)
+            prob_sp = model_sp.predict_proba(X_sp)
+
+        assert_allclose(pred, pred_sp)
+        assert_allclose(prob, prob_sp)
+        assert_allclose(model.coef_, model_sp.coef_, rtol=1e-4)
+        assert_allclose(model.intercept_, model_sp.intercept_, rtol=1e-4)

sklearnex/manifold/t_sne.py

@@ -15,3 +15,7 @@
 # ===============================================================================
 
 from daal4py.sklearn.manifold import TSNE
+from onedal._device_offload import support_usm_ndarray
+
+TSNE.fit = support_usm_ndarray(queue_param=False)(TSNE.fit)
+TSNE.fit_transform = support_usm_ndarray(queue_param=False)(TSNE.fit_transform)

sklearnex/metrics/pairwise.py

@@ -15,3 +15,8 @@
 # ===============================================================================
 
 from daal4py.sklearn.metrics import pairwise_distances
+from onedal._device_offload import support_usm_ndarray
+
+pairwise_distances = support_usm_ndarray(freefunc=True, queue_param=False)(
+    pairwise_distances
+)

sklearnex/metrics/ranking.py

@@ -15,3 +15,6 @@
 # ===============================================================================
 
 from daal4py.sklearn.metrics import roc_auc_score
+from onedal._device_offload import support_usm_ndarray
+
+roc_auc_score = support_usm_ndarray(freefunc=True, queue_param=False)(roc_auc_score)

sklearnex/model_selection/split.py

@@ -15,3 +15,6 @@
 # ===============================================================================
 
 from daal4py.sklearn.model_selection import train_test_split
+from onedal._device_offload import support_usm_ndarray
+
+train_test_split = support_usm_ndarray(freefunc=True, queue_param=False)(train_test_split)

sklearnex/neighbors/_lof.py

@@ -97,6 +97,15 @@ class LocalOutlierFactor(KNeighborsDispatchingBase, sklearn_LocalOutlierFactor):
                 self.negative_outlier_factor_, 100.0 * self.contamination
             )
 
+        # adoption of warning for data with duplicated samples from
+        # https://github.com/scikit-learn/scikit-learn/pull/28773
+        if sklearn_check_version("1.6"):
+            if np.min(self.negative_outlier_factor_) < -1e7 and not self.novelty:
+                warnings.warn(
+                    "Duplicate values are leading to incorrect results. "
+                    "Increase the number of neighbors for more accurate results."
+                )
+
         return self
 
     def fit(self, X, y=None):

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

@@ -14,6 +14,7 @@
 # limitations under the License.
 # ==============================================================================
 
+from sklearn.metrics import r2_score
 from sklearn.neighbors._regression import (
     KNeighborsRegressor as sklearn_KNeighborsRegressor,
 )
@@ -29,7 +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 = {
@@ -117,6 +118,23 @@ class KNeighborsRegressor(sklearn_KNeighborsRegressor, KNeighborsDispatchingBase
             X,
         )
 
+    @wrap_output_data
+    def score(self, X, y, sample_weight=None):
+        check_is_fitted(self)
+        if sklearn_check_version("1.0"):
+            self._check_feature_names(X, reset=False)
+        return dispatch(
+            self,
+            "score",
+            {
+                "onedal": self.__class__._onedal_score,
+                "sklearn": sklearn_KNeighborsRegressor.score,
+            },
+            X,
+            y,
+            sample_weight=sample_weight,
+        )
+
     @wrap_output_data
     def kneighbors(self, X=None, n_neighbors=None, return_distance=True):
         check_is_fitted(self)
@@ -134,24 +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,
@@ -184,6 +184,11 @@ class KNeighborsRegressor(sklearn_KNeighborsRegressor, KNeighborsDispatchingBase
             X, n_neighbors, return_distance, queue=queue
         )
 
+    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
+        )
+
     def _save_attributes(self):
         self.n_features_in_ = self._onedal_estimator.n_features_in_
         self.n_samples_fit_ = self._onedal_estimator.n_samples_fit_
@@ -195,4 +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"]
+__all__ = ["cluster", "covariance", "decomposition", "linear_model"]

sklearnex/preview/decomposition/__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 .incremental_pca import IncrementalPCA
+
+__all__ = ["IncrementalPCA"]