scikit-learn-intelex 2024.3.0__py311-none-manylinux1_x86_64.whl → 2024.5.0__py311-none-manylinux1_x86_64.whl

sklearnex/svm/_common.py

@@ -76,7 +76,7 @@ class BaseSVM(ABC):
         inference_methods = (
             ["predict"]
             if class_name.endswith("R")
-            else ["predict", "predict_proba", "decision_function"]
+            else ["predict", "predict_proba", "decision_function", "score"]
         )
         if method_name in inference_methods:
             patching_status.and_conditions(
@@ -111,12 +111,9 @@ class BaseSVC(BaseSVM):
             cv = StratifiedKFold(
                 n_splits=n_splits, shuffle=True, random_state=self.random_state
             )
-            if sklearn_check_version("0.24"):
-                self.clf_prob = CalibratedClassifierCV(
-                    clf_base, ensemble=False, cv=cv, method="sigmoid", n_jobs=n_jobs
-                )
-            else:
-                self.clf_prob = CalibratedClassifierCV(clf_base, cv=cv, method="sigmoid")
+            self.clf_prob = CalibratedClassifierCV(
+                clf_base, ensemble=False, cv=cv, method="sigmoid", n_jobs=n_jobs
+            )
             self.clf_prob.fit(X, y, sample_weight)
         except ValueError:
             clf_base = clf_base.fit(X, y, sample_weight)

sklearnex/svm/nusvc.py

@@ -14,12 +14,15 @@
 # limitations under the License.
 # ==============================================================================
 
+import numpy as np
 from sklearn.exceptions import NotFittedError
+from sklearn.metrics import accuracy_score
 from sklearn.svm import NuSVC as sklearn_NuSVC
 from sklearn.utils.validation import _deprecate_positional_args
 
 from daal4py.sklearn._n_jobs_support import control_n_jobs
 from daal4py.sklearn._utils import sklearn_check_version
+from sklearnex.utils import get_namespace
 
 from .._device_offload import dispatch, wrap_output_data
 from ._common import BaseSVC
@@ -31,7 +34,7 @@ from onedal.svm import NuSVC as onedal_NuSVC
 
 
 @control_n_jobs(
-    decorated_methods=["fit", "predict", "_predict_proba", "decision_function"]
+    decorated_methods=["fit", "predict", "_predict_proba", "decision_function", "score"]
 )
 class NuSVC(sklearn_NuSVC, BaseSVC):
     __doc__ = sklearn_NuSVC.__doc__
@@ -78,39 +81,6 @@ class NuSVC(sklearn_NuSVC, BaseSVC):
         )
 
     def fit(self, X, y, sample_weight=None):
-        """
-        Fit the SVM model according to the given training data.
-
-        Parameters
-        ----------
-        X : {array-like, sparse matrix} of shape (n_samples, n_features) \
-                or (n_samples, n_samples)
-            Training vectors, where `n_samples` is the number of samples
-            and `n_features` is the number of features.
-            For kernel="precomputed", the expected shape of X is
-            (n_samples, n_samples).
-
-        y : array-like of shape (n_samples,)
-            Target values (class labels in classification, real numbers in
-            regression).
-
-        sample_weight : array-like of shape (n_samples,), default=None
-            Per-sample weights. Rescale C per sample. Higher weights
-            force the classifier to put more emphasis on these points.
-
-        Returns
-        -------
-        self : object
-            Fitted estimator.
-
-        Notes
-        -----
-        If X and y are not C-ordered and contiguous arrays of np.float64 and
-        X is not a scipy.sparse.csr_matrix, X and/or y may be copied.
-
-        If X is a dense array, then the other methods will not support sparse
-        matrices as input.
-        """
         if sklearn_check_version("1.2"):
             self._validate_params()
         if sklearn_check_version("1.0"):
@@ -131,22 +101,6 @@ class NuSVC(sklearn_NuSVC, BaseSVC):
 
     @wrap_output_data
     def predict(self, X):
-        """
-        Perform regression on samples in X.
-
-        For an one-class model, +1 (inlier) or -1 (outlier) is returned.
-
-        Parameters
-        ----------
-        X : {array-like, sparse matrix} of shape (n_samples, n_features)
-            For kernel="precomputed", the expected shape of X is
-            (n_samples_test, n_samples_train).
-
-        Returns
-        -------
-        y_pred : ndarray of shape (n_samples,)
-            The predicted values.
-        """
         if sklearn_check_version("1.0"):
             self._check_feature_names(X, reset=False)
         return dispatch(
@@ -159,6 +113,22 @@ class NuSVC(sklearn_NuSVC, BaseSVC):
             X,
         )
 
+    @wrap_output_data
+    def score(self, X, y, sample_weight=None):
+        if sklearn_check_version("1.0"):
+            self._check_feature_names(X, reset=False)
+        return dispatch(
+            self,
+            "score",
+            {
+                "onedal": self.__class__._onedal_score,
+                "sklearn": sklearn_NuSVC.score,
+            },
+            X,
+            y,
+            sample_weight=sample_weight,
+        )
+
     if sklearn_check_version("1.0"):
 
         @available_if(sklearn_NuSVC._check_proba)
@@ -191,6 +161,38 @@ class NuSVC(sklearn_NuSVC, BaseSVC):
             """
             return self._predict_proba(X)
 
+        @available_if(sklearn_NuSVC._check_proba)
+        def predict_log_proba(self, X):
+            """Compute log probabilities of possible outcomes for samples in X.
+
+            The model need to have probability information computed at training
+            time: fit with attribute `probability` set to True.
+
+            Parameters
+            ----------
+            X : array-like of shape (n_samples, n_features) or \
+                    (n_samples_test, n_samples_train)
+                For kernel="precomputed", the expected shape of X is
+                (n_samples_test, n_samples_train).
+
+            Returns
+            -------
+            T : ndarray of shape (n_samples, n_classes)
+                Returns the log-probabilities of the sample for each class in
+                the model. The columns correspond to the classes in sorted
+                order, as they appear in the attribute :term:`classes_`.
+
+            Notes
+            -----
+            The probability model is created using cross validation, so
+            the results can be slightly different than those obtained by
+            predict. Also, it will produce meaningless results on very small
+            datasets.
+            """
+            xp, _ = get_namespace(X)
+
+            return xp.log(self.predict_proba(X))
+
     else:
 
         @property
@@ -198,6 +200,10 @@ class NuSVC(sklearn_NuSVC, BaseSVC):
             self._check_proba()
             return self._predict_proba
 
+        def _predict_log_proba(self, X):
+            xp, _ = get_namespace(X)
+            return xp.log(self.predict_proba(X))
+
         predict_proba.__doc__ = sklearn_NuSVC.predict_proba.__doc__
 
     @wrap_output_data
@@ -278,3 +284,13 @@ class NuSVC(sklearn_NuSVC, BaseSVC):
 
     def _onedal_decision_function(self, X, queue=None):
         return self._onedal_estimator.decision_function(X, queue=queue)
+
+    def _onedal_score(self, X, y, sample_weight=None, queue=None):
+        return accuracy_score(
+            y, self._onedal_predict(X, queue=queue), sample_weight=sample_weight
+        )
+
+    fit.__doc__ = sklearn_NuSVC.fit.__doc__
+    predict.__doc__ = sklearn_NuSVC.predict.__doc__
+    decision_function.__doc__ = sklearn_NuSVC.decision_function.__doc__
+    score.__doc__ = sklearn_NuSVC.score.__doc__

sklearnex/svm/nusvr.py

@@ -63,39 +63,6 @@ class NuSVR(sklearn_NuSVR, BaseSVR):
         )
 
     def fit(self, X, y, sample_weight=None):
-        """
-        Fit the SVM model according to the given training data.
-
-        Parameters
-        ----------
-        X : {array-like, sparse matrix} of shape (n_samples, n_features) \
-                or (n_samples, n_samples)
-            Training vectors, where `n_samples` is the number of samples
-            and `n_features` is the number of features.
-            For kernel="precomputed", the expected shape of X is
-            (n_samples, n_samples).
-
-        y : array-like of shape (n_samples,)
-            Target values (class labels in classification, real numbers in
-            regression).
-
-        sample_weight : array-like of shape (n_samples,), default=None
-            Per-sample weights. Rescale C per sample. Higher weights
-            force the classifier to put more emphasis on these points.
-
-        Returns
-        -------
-        self : object
-            Fitted estimator.
-
-        Notes
-        -----
-        If X and y are not C-ordered and contiguous arrays of np.float64 and
-        X is not a scipy.sparse.csr_matrix, X and/or y may be copied.
-
-        If X is a dense array, then the other methods will not support sparse
-        matrices as input.
-        """
         if sklearn_check_version("1.2"):
             self._validate_params()
         if sklearn_check_version("1.0"):
@@ -115,22 +82,6 @@ class NuSVR(sklearn_NuSVR, BaseSVR):
 
     @wrap_output_data
     def predict(self, X):
-        """
-        Perform regression on samples in X.
-
-        For an one-class model, +1 (inlier) or -1 (outlier) is returned.
-
-        Parameters
-        ----------
-        X : {array-like, sparse matrix} of shape (n_samples, n_features)
-            For kernel="precomputed", the expected shape of X is
-            (n_samples_test, n_samples_train).
-
-        Returns
-        -------
-        y_pred : ndarray of shape (n_samples,)
-            The predicted values.
-        """
         if sklearn_check_version("1.0"):
             self._check_feature_names(X, reset=False)
         return dispatch(
@@ -163,3 +114,6 @@ class NuSVR(sklearn_NuSVR, BaseSVR):
 
     def _onedal_predict(self, X, queue=None):
         return self._onedal_estimator.predict(X, queue=queue)
+
+    fit.__doc__ = sklearn_NuSVR.fit.__doc__
+    predict.__doc__ = sklearn_NuSVR.predict.__doc__

sklearnex/svm/svc.py

@@ -17,11 +17,13 @@
 import numpy as np
 from scipy import sparse as sp
 from sklearn.exceptions import NotFittedError
+from sklearn.metrics import accuracy_score
 from sklearn.svm import SVC as sklearn_SVC
 from sklearn.utils.validation import _deprecate_positional_args
 
 from daal4py.sklearn._n_jobs_support import control_n_jobs
 from daal4py.sklearn._utils import sklearn_check_version
+from sklearnex.utils import get_namespace
 
 from .._device_offload import dispatch, wrap_output_data
 from .._utils import PatchingConditionsChain
@@ -34,7 +36,7 @@ from onedal.svm import SVC as onedal_SVC
 
 
 @control_n_jobs(
-    decorated_methods=["fit", "predict", "_predict_proba", "decision_function"]
+    decorated_methods=["fit", "predict", "_predict_proba", "decision_function", "score"]
 )
 class SVC(sklearn_SVC, BaseSVC):
     __doc__ = sklearn_SVC.__doc__
@@ -81,39 +83,6 @@ class SVC(sklearn_SVC, BaseSVC):
         )
 
     def fit(self, X, y, sample_weight=None):
-        """
-        Fit the SVM model according to the given training data.
-
-        Parameters
-        ----------
-        X : {array-like, sparse matrix} of shape (n_samples, n_features) \
-                or (n_samples, n_samples)
-            Training vectors, where `n_samples` is the number of samples
-            and `n_features` is the number of features.
-            For kernel="precomputed", the expected shape of X is
-            (n_samples, n_samples).
-
-        y : array-like of shape (n_samples,)
-            Target values (class labels in classification, real numbers in
-            regression).
-
-        sample_weight : array-like of shape (n_samples,), default=None
-            Per-sample weights. Rescale C per sample. Higher weights
-            force the classifier to put more emphasis on these points.
-
-        Returns
-        -------
-        self : object
-            Fitted estimator.
-
-        Notes
-        -----
-        If X and y are not C-ordered and contiguous arrays of np.float64 and
-        X is not a scipy.sparse.csr_matrix, X and/or y may be copied.
-
-        If X is a dense array, then the other methods will not support sparse
-        matrices as input.
-        """
         if sklearn_check_version("1.2"):
             self._validate_params()
         if sklearn_check_version("1.0"):
@@ -133,22 +102,6 @@ class SVC(sklearn_SVC, BaseSVC):
 
     @wrap_output_data
     def predict(self, X):
-        """
-        Perform regression on samples in X.
-
-        For an one-class model, +1 (inlier) or -1 (outlier) is returned.
-
-        Parameters
-        ----------
-        X : {array-like, sparse matrix} of shape (n_samples, n_features)
-            For kernel="precomputed", the expected shape of X is
-            (n_samples_test, n_samples_train).
-
-        Returns
-        -------
-        y_pred : ndarray of shape (n_samples,)
-            The predicted values.
-        """
         if sklearn_check_version("1.0"):
             self._check_feature_names(X, reset=False)
         return dispatch(
@@ -161,6 +114,22 @@ class SVC(sklearn_SVC, BaseSVC):
             X,
         )
 
+    @wrap_output_data
+    def score(self, X, y, sample_weight=None):
+        if sklearn_check_version("1.0"):
+            self._check_feature_names(X, reset=False)
+        return dispatch(
+            self,
+            "score",
+            {
+                "onedal": self.__class__._onedal_score,
+                "sklearn": sklearn_SVC.score,
+            },
+            X,
+            y,
+            sample_weight=sample_weight,
+        )
+
     if sklearn_check_version("1.0"):
 
         @available_if(sklearn_SVC._check_proba)
@@ -193,6 +162,38 @@ class SVC(sklearn_SVC, BaseSVC):
             """
             return self._predict_proba(X)
 
+        @available_if(sklearn_SVC._check_proba)
+        def predict_log_proba(self, X):
+            """Compute log probabilities of possible outcomes for samples in X.
+
+            The model need to have probability information computed at training
+            time: fit with attribute `probability` set to True.
+
+            Parameters
+            ----------
+            X : array-like of shape (n_samples, n_features) or \
+                    (n_samples_test, n_samples_train)
+                For kernel="precomputed", the expected shape of X is
+                (n_samples_test, n_samples_train).
+
+            Returns
+            -------
+            T : ndarray of shape (n_samples, n_classes)
+                Returns the log-probabilities of the sample for each class in
+                the model. The columns correspond to the classes in sorted
+                order, as they appear in the attribute :term:`classes_`.
+
+            Notes
+            -----
+            The probability model is created using cross validation, so
+            the results can be slightly different than those obtained by
+            predict. Also, it will produce meaningless results on very small
+            datasets.
+            """
+            xp, _ = get_namespace(X)
+
+            return xp.log(self.predict_proba(X))
+
     else:
 
         @property
@@ -200,6 +201,10 @@ class SVC(sklearn_SVC, BaseSVC):
             self._check_proba()
             return self._predict_proba
 
+        def _predict_log_proba(self, X):
+            xp, _ = get_namespace(X)
+            return xp.log(self.predict_proba(X))
+
         predict_proba.__doc__ = sklearn_SVC.predict_proba.__doc__
 
     @wrap_output_data
@@ -257,7 +262,7 @@ class SVC(sklearn_SVC, BaseSVC):
         if method_name == "fit":
             patching_status.and_conditions(conditions)
             return patching_status
-        if method_name in ["predict", "predict_proba", "decision_function"]:
+        if method_name in ["predict", "predict_proba", "decision_function", "score"]:
             conditions.append(
                 (hasattr(self, "_onedal_estimator"), "oneDAL model was not trained")
             )
@@ -307,3 +312,13 @@ class SVC(sklearn_SVC, BaseSVC):
 
     def _onedal_decision_function(self, X, queue=None):
         return self._onedal_estimator.decision_function(X, queue=queue)
+
+    def _onedal_score(self, X, y, sample_weight=None, queue=None):
+        return accuracy_score(
+            y, self._onedal_predict(X, queue=queue), sample_weight=sample_weight
+        )
+
+    fit.__doc__ = sklearn_SVC.fit.__doc__
+    predict.__doc__ = sklearn_SVC.predict.__doc__
+    decision_function.__doc__ = sklearn_SVC.decision_function.__doc__
+    score.__doc__ = sklearn_SVC.score.__doc__

sklearnex/svm/svr.py

@@ -63,39 +63,6 @@ class SVR(sklearn_SVR, BaseSVR):
         )
 
     def fit(self, X, y, sample_weight=None):
-        """
-        Fit the SVM model according to the given training data.
-
-        Parameters
-        ----------
-        X : {array-like, sparse matrix} of shape (n_samples, n_features) \
-                or (n_samples, n_samples)
-            Training vectors, where `n_samples` is the number of samples
-            and `n_features` is the number of features.
-            For kernel="precomputed", the expected shape of X is
-            (n_samples, n_samples).
-
-        y : array-like of shape (n_samples,)
-            Target values (class labels in classification, real numbers in
-            regression).
-
-        sample_weight : array-like of shape (n_samples,), default=None
-            Per-sample weights. Rescale C per sample. Higher weights
-            force the classifier to put more emphasis on these points.
-
-        Returns
-        -------
-        self : object
-            Fitted estimator.
-
-        Notes
-        -----
-        If X and y are not C-ordered and contiguous arrays of np.float64 and
-        X is not a scipy.sparse.csr_matrix, X and/or y may be copied.
-
-        If X is a dense array, then the other methods will not support sparse
-        matrices as input.
-        """
         if sklearn_check_version("1.2"):
             self._validate_params()
         if sklearn_check_version("1.0"):
@@ -116,22 +83,6 @@ class SVR(sklearn_SVR, BaseSVR):
 
     @wrap_output_data
     def predict(self, X):
-        """
-        Perform regression on samples in X.
-
-        For an one-class model, +1 (inlier) or -1 (outlier) is returned.
-
-        Parameters
-        ----------
-        X : {array-like, sparse matrix} of shape (n_samples, n_features)
-            For kernel="precomputed", the expected shape of X is
-            (n_samples_test, n_samples_train).
-
-        Returns
-        -------
-        y_pred : ndarray of shape (n_samples,)
-            The predicted values.
-        """
         if sklearn_check_version("1.0"):
             self._check_feature_names(X, reset=False)
         return dispatch(
@@ -164,3 +115,6 @@ class SVR(sklearn_SVR, BaseSVR):
 
     def _onedal_predict(self, X, queue=None):
         return self._onedal_estimator.predict(X, queue=queue)
+
+    fit.__doc__ = sklearn_SVR.fit.__doc__
+    predict.__doc__ = sklearn_SVR.predict.__doc__

sklearnex/tests/_utils.py

@@ -17,6 +17,7 @@
 from inspect import isclass
 
 import numpy as np
+from sklearn import clone
 from sklearn.base import (
     BaseEstimator,
     ClassifierMixin,
@@ -30,6 +31,7 @@ from sklearn.neighbors._base import KNeighborsMixin
 
 from onedal.tests.utils._dataframes_support import _convert_to_dataframe
 from sklearnex import get_patch_map, patch_sklearn, sklearn_is_patched, unpatch_sklearn
+from sklearnex.linear_model import LogisticRegression
 from sklearnex.neighbors import (
     KNeighborsClassifier,
     KNeighborsRegressor,
@@ -86,26 +88,38 @@ mixin_map = [
 ]
 
 
-SPECIAL_INSTANCES = {
-    str(i): i
-    for i in [
-        LocalOutlierFactor(novelty=True),
-        SVC(probability=True),
-        NuSVC(probability=True),
-        KNeighborsClassifier(algorithm="brute"),
-        KNeighborsRegressor(algorithm="brute"),
-        NearestNeighbors(algorithm="brute"),
-    ]
-}
+class _sklearn_clone_dict(dict):
+
+    def __getitem__(self, key):
+        return clone(super().__getitem__(key))
+
+
+SPECIAL_INSTANCES = _sklearn_clone_dict(
+    {
+        str(i): i
+        for i in [
+            LocalOutlierFactor(novelty=True),
+            SVC(probability=True),
+            NuSVC(probability=True),
+            KNeighborsClassifier(algorithm="brute"),
+            KNeighborsRegressor(algorithm="brute"),
+            NearestNeighbors(algorithm="brute"),
+            LogisticRegression(solver="newton-cg"),
+        ]
+    }
+)
 
 
 def gen_models_info(algorithms):
     output = []
     for i in algorithms:
-        # split handles SPECIAL_INSTANCES or custom inputs
-        # custom sklearn inputs must be a dict of estimators
-        # with keys set by the __str__ method
-        est = PATCHED_MODELS[i.split("(")[0]]
+
+        if i in PATCHED_MODELS:
+            est = PATCHED_MODELS[i]
+        elif i in SPECIAL_INSTANCES:
+            est = SPECIAL_INSTANCES[i].__class__
+        else:
+            raise KeyError(f"Unrecognized sklearnex estimator: {i}")
 
         methods = set()
         candidates = set(
@@ -116,7 +130,11 @@ def gen_models_info(algorithms):
             if issubclass(est, mixin):
                 methods |= candidates & set(method)
 
-        output += [[i, j] for j in methods]
+        output += [[i, j] for j in methods] if methods else [[i, None]]
+
+    # In the case that no methods are available, set method to None.
+    # This will allow estimators without mixins to still test the fit
+    # method in various tests.
     return output
 
 

sklearnex/tests/test_memory_usage.py

@@ -96,7 +96,11 @@ def remove_duplicated_estimators(estimators_list):
     return estimators_map.values()
 
 
-BANNED_ESTIMATORS = ("TSNE",)  # too slow for using in testing on common data size
+BANNED_ESTIMATORS = (
+    "IncrementalEmpiricalCovariance",  # dataframe_f issues
+    "IncrementalLinearRegression",  # TODO fix memory leak issue in private CI for data_shape = (1000, 100), data_transform_function = dataframe_f
+    "TSNE",  # too slow for using in testing on common data size
+)
 estimators = [
     TrainTestSplitEstimator,
     FiniteCheckEstimator,

sklearnex/tests/test_n_jobs_support.py

@@ -22,6 +22,7 @@ import pytest
 from sklearn.base import BaseEstimator
 from sklearn.datasets import make_classification
 
+from sklearnex.decomposition import PCA
 from sklearnex.dispatcher import get_patch_map
 from sklearnex.svm import SVC, NuSVC
 
@@ -73,6 +74,9 @@ def test_n_jobs_support(caplog, estimator_class, n_jobs):
     # by default, [Nu]SVC.predict_proba is restricted by @available_if decorator
     if estimator_class in [SVC, NuSVC]:
         estimator_kwargs["probability"] = True
+    # explicitly request oneDAL's PCA-Covariance algorithm
+    if estimator_class == PCA:
+        estimator_kwargs["svd_solver"] = "covariance_eigh"
     estimator_instance = estimator_class(**estimator_kwargs)
     # check `n_jobs` parameter doc entry
     check_estimator_doc(estimator_class)
@@ -84,10 +88,16 @@ def test_n_jobs_support(caplog, estimator_class, n_jobs):
         if method_name == "fit":
             continue
         method = getattr(estimator_instance, method_name)
-        if len(inspect.signature(method).parameters) == 0:
+        argdict = inspect.signature(method).parameters
+        argnum = len(
+            [i for i in argdict if argdict[i].default == inspect.Parameter.empty]
+        )
+        if argnum == 0:
             check_method(method=method, caplog=caplog)
-        else:
+        elif argnum == 1:
             check_method(X, method=method, caplog=caplog)
+        else:
+            check_method(X, Y, method=method, caplog=caplog)
     # check if correct methods were decorated
     check_methods_decoration(estimator_class)
     check_methods_decoration(estimator_instance)