scikit-learn-intelex 2024.0.1__py311-none-manylinux1_x86_64.whl → 2024.4.0__py311-none-manylinux1_x86_64.whl

sklearnex/linear_model/logistic_regression.py

@@ -0,0 +1,385 @@
+# ===============================================================================
+# 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 abc import ABC
+
+from daal4py.sklearn._utils import daal_check_version
+from daal4py.sklearn.linear_model.logistic_path import (
+    LogisticRegression as LogisticRegression_daal4py,
+)
+from daal4py.sklearn.linear_model.logistic_path import daal4py_fit, daal4py_predict
+
+
+class BaseLogisticRegression(ABC):
+    def _save_attributes(self):
+        assert hasattr(self, "_onedal_estimator")
+        self.classes_ = self._onedal_estimator.classes_
+        self.coef_ = self._onedal_estimator.coef_
+        self.intercept_ = self._onedal_estimator.intercept_
+        self.n_features_in_ = self._onedal_estimator.n_features_in_
+        self.n_iter_ = self._onedal_estimator.n_iter_
+
+
+if daal_check_version((2024, "P", 1)):
+    import numpy as np
+    from scipy.sparse import issparse
+    from sklearn.linear_model import LogisticRegression as sklearn_LogisticRegression
+    from sklearn.metrics import accuracy_score
+    from sklearn.utils.multiclass import type_of_target
+    from sklearn.utils.validation import check_X_y
+
+    from daal4py.sklearn._n_jobs_support import control_n_jobs
+    from daal4py.sklearn._utils import sklearn_check_version
+    from onedal.linear_model import LogisticRegression as onedal_LogisticRegression
+    from onedal.utils import _num_samples
+
+    from .._device_offload import dispatch, wrap_output_data
+    from .._utils import PatchingConditionsChain, get_patch_message
+    from ..utils.validation import _assert_all_finite
+
+    @control_n_jobs(
+        decorated_methods=[
+            "fit",
+            "predict",
+            "predict_proba",
+            "predict_log_proba",
+            "score",
+        ]
+    )
+    class LogisticRegression(sklearn_LogisticRegression, BaseLogisticRegression):
+        __doc__ = sklearn_LogisticRegression.__doc__
+        intercept_, coef_, n_iter_ = None, None, None
+
+        if sklearn_check_version("1.2"):
+            _parameter_constraints: dict = {
+                **sklearn_LogisticRegression._parameter_constraints
+            }
+
+        def __init__(
+            self,
+            penalty="l2",
+            *,
+            dual=False,
+            tol=1e-4,
+            C=1.0,
+            fit_intercept=True,
+            intercept_scaling=1,
+            class_weight=None,
+            random_state=None,
+            solver="lbfgs",
+            max_iter=100,
+            multi_class="auto",
+            verbose=0,
+            warm_start=False,
+            n_jobs=None,
+            l1_ratio=None,
+        ):
+            super().__init__(
+                penalty=penalty,
+                dual=dual,
+                tol=tol,
+                C=C,
+                fit_intercept=fit_intercept,
+                intercept_scaling=intercept_scaling,
+                class_weight=class_weight,
+                random_state=random_state,
+                solver=solver,
+                max_iter=max_iter,
+                multi_class=multi_class,
+                verbose=verbose,
+                warm_start=warm_start,
+                n_jobs=n_jobs,
+                l1_ratio=l1_ratio,
+            )
+
+        _onedal_cpu_fit = daal4py_fit
+
+        def fit(self, X, y, sample_weight=None):
+            if sklearn_check_version("1.0"):
+                self._check_feature_names(X, reset=True)
+            if sklearn_check_version("1.2"):
+                self._validate_params()
+            dispatch(
+                self,
+                "fit",
+                {
+                    "onedal": self.__class__._onedal_fit,
+                    "sklearn": sklearn_LogisticRegression.fit,
+                },
+                X,
+                y,
+                sample_weight,
+            )
+            return self
+
+        @wrap_output_data
+        def predict(self, X):
+            if sklearn_check_version("1.0"):
+                self._check_feature_names(X, reset=False)
+            return dispatch(
+                self,
+                "predict",
+                {
+                    "onedal": self.__class__._onedal_predict,
+                    "sklearn": sklearn_LogisticRegression.predict,
+                },
+                X,
+            )
+
+        @wrap_output_data
+        def predict_proba(self, X):
+            if sklearn_check_version("1.0"):
+                self._check_feature_names(X, reset=False)
+            return dispatch(
+                self,
+                "predict",
+                {
+                    "onedal": self.__class__._onedal_predict_proba,
+                    "sklearn": sklearn_LogisticRegression.predict_proba,
+                },
+                X,
+            )
+
+        @wrap_output_data
+        def predict_log_proba(self, X):
+            if sklearn_check_version("1.0"):
+                self._check_feature_names(X, reset=False)
+            return dispatch(
+                self,
+                "predict",
+                {
+                    "onedal": self.__class__._onedal_predict_log_proba,
+                    "sklearn": sklearn_LogisticRegression.predict_log_proba,
+                },
+                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_LogisticRegression.score,
+                },
+                X,
+                y,
+                sample_weight=sample_weight,
+            )
+
+        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
+            )
+
+        def _test_type_and_finiteness(self, X_in):
+            X = np.asarray(X_in)
+
+            if np.iscomplexobj(X):
+                return False
+            try:
+                _assert_all_finite(X)
+            except BaseException:
+                return False
+            return True
+
+        def _onedal_gpu_fit_supported(self, method_name, *data):
+            assert method_name == "fit"
+            assert len(data) == 3
+            X, y, sample_weight = data
+
+            class_name = self.__class__.__name__
+            patching_status = PatchingConditionsChain(
+                f"sklearn.linear_model.{class_name}.fit"
+            )
+
+            dal_ready = patching_status.and_conditions(
+                [
+                    (self.penalty == "l2", "Only l2 penalty is supported."),
+                    (self.dual == False, "dual=True is not supported."),
+                    (
+                        self.intercept_scaling == 1,
+                        "Intercept scaling is not supported.",
+                    ),
+                    (self.class_weight is None, "Class weight is not supported"),
+                    (self.solver == "newton-cg", "Only newton-cg solver is supported."),
+                    (
+                        self.multi_class != "multinomial",
+                        "multi_class parameter is not supported.",
+                    ),
+                    (self.warm_start == False, "Warm start is not supported."),
+                    (self.l1_ratio is None, "l1 ratio is not supported."),
+                    (sample_weight is None, "Sample weight is not supported."),
+                    (
+                        type_of_target(y) == "binary",
+                        "Only binary classification is supported",
+                    ),
+                ]
+            )
+
+            if not dal_ready:
+                return patching_status
+
+            if not patching_status.and_condition(
+                self._test_type_and_finiteness(X), "Input X is not supported."
+            ):
+                return patching_status
+
+            patching_status.and_condition(
+                self._test_type_and_finiteness(y), "Input y is not supported."
+            )
+
+            return patching_status
+
+        def _onedal_gpu_predict_supported(self, method_name, *data):
+            assert method_name in [
+                "predict",
+                "predict_proba",
+                "predict_log_proba",
+                "score",
+            ]
+
+            class_name = self.__class__.__name__
+            patching_status = PatchingConditionsChain(
+                f"sklearn.linear_model.{class_name}.{method_name}"
+            )
+
+            n_samples = _num_samples(data[0])
+            model_is_sparse = issparse(self.coef_) or (
+                self.fit_intercept and issparse(self.intercept_)
+            )
+            dal_ready = patching_status.and_conditions(
+                [
+                    (n_samples > 0, "Number of samples is less than 1."),
+                    (
+                        not any([issparse(i) for i in data]),
+                        "Sparse input is not supported.",
+                    ),
+                    (not model_is_sparse, "Sparse coefficients are not supported."),
+                    (
+                        hasattr(self, "_onedal_estimator"),
+                        "oneDAL model was not trained.",
+                    ),
+                ]
+            )
+            if not dal_ready:
+                return patching_status
+
+            patching_status.and_condition(
+                self._test_type_and_finiteness(*data), "Input X is not supported."
+            )
+
+            return patching_status
+
+        def _onedal_gpu_supported(self, method_name, *data):
+            if method_name == "fit":
+                return self._onedal_gpu_fit_supported(method_name, *data)
+            if method_name in ["predict", "predict_proba", "predict_log_proba", "score"]:
+                return self._onedal_gpu_predict_supported(method_name, *data)
+            raise RuntimeError(
+                f"Unknown method {method_name} in {self.__class__.__name__}"
+            )
+
+        def _onedal_cpu_supported(self, method_name, *data):
+            class_name = self.__class__.__name__
+            patching_status = PatchingConditionsChain(
+                f"sklearn.linear_model.{class_name}.{method_name}"
+            )
+
+            return patching_status
+
+        def _initialize_onedal_estimator(self):
+            onedal_params = {
+                "tol": self.tol,
+                "C": self.C,
+                "fit_intercept": self.fit_intercept,
+                "solver": self.solver,
+                "max_iter": self.max_iter,
+            }
+            self._onedal_estimator = onedal_LogisticRegression(**onedal_params)
+
+        def _onedal_fit(self, X, y, sample_weight, queue=None):
+            if queue is None or queue.sycl_device.is_cpu:
+                return self._onedal_cpu_fit(X, y, sample_weight)
+
+            assert sample_weight is None
+
+            check_params = {
+                "X": X,
+                "y": y,
+                "dtype": [np.float64, np.float32],
+                "accept_sparse": False,
+                "multi_output": False,
+                "force_all_finite": True,
+            }
+            if sklearn_check_version("1.2"):
+                X, y = self._validate_data(**check_params)
+            else:
+                X, y = check_X_y(**check_params)
+            self._initialize_onedal_estimator()
+            try:
+                self._onedal_estimator.fit(X, y, queue=queue)
+                self._save_attributes()
+            except RuntimeError:
+                logging.getLogger("sklearnex").info(
+                    f"{self.__class__.__name__}.fit "
+                    + get_patch_message("sklearn_after_onedal")
+                )
+
+                del self._onedal_estimator
+                super().fit(X, y)
+
+        def _onedal_predict(self, X, queue=None):
+            if queue is None or queue.sycl_device.is_cpu:
+                return daal4py_predict(self, X, "computeClassLabels")
+
+            X = self._validate_data(X, accept_sparse=False, reset=False)
+            assert hasattr(self, "_onedal_estimator")
+            return self._onedal_estimator.predict(X, queue=queue)
+
+        def _onedal_predict_proba(self, X, queue=None):
+            if queue is None or queue.sycl_device.is_cpu:
+                return daal4py_predict(self, X, "computeClassProbabilities")
+
+            X = self._validate_data(X, accept_sparse=False, reset=False)
+            assert hasattr(self, "_onedal_estimator")
+            return self._onedal_estimator.predict_proba(X, queue=queue)
+
+        def _onedal_predict_log_proba(self, X, queue=None):
+            if queue is None or queue.sycl_device.is_cpu:
+                return daal4py_predict(self, X, "computeClassLogProbabilities")
+
+            X = self._validate_data(X, accept_sparse=False, reset=False)
+            assert hasattr(self, "_onedal_estimator")
+            return self._onedal_estimator.predict_log_proba(X, queue=queue)
+
+        fit.__doc__ = sklearn_LogisticRegression.fit.__doc__
+        predict.__doc__ = sklearn_LogisticRegression.predict.__doc__
+        predict_proba.__doc__ = sklearn_LogisticRegression.predict_proba.__doc__
+        predict_log_proba.__doc__ = sklearn_LogisticRegression.predict_log_proba.__doc__
+        score.__doc__ = sklearn_LogisticRegression.score.__doc__
+
+else:
+    LogisticRegression = LogisticRegression_daal4py
+
+    logging.warning(
+        "Sklearnex LogisticRegression requires oneDAL version >= 2024.0.1 "
+        "but it was not found"
+    )

sklearnex/linear_model/ridge.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/linear_model/tests/test_linear.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #
@@ -29,20 +28,33 @@ from onedal.tests.utils._dataframes_support import (
 
 
 @pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
-def test_sklearnex_import_linear(dataframe, queue):
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+@pytest.mark.parametrize("macro_block", [None, 1024])
+def test_sklearnex_import_linear(dataframe, queue, dtype, macro_block):
     from sklearnex.linear_model import LinearRegression
 
     X = np.array([[1, 1], [1, 2], [2, 2], [2, 3]])
     y = np.dot(X, np.array([1, 2])) + 3
+    X = X.astype(dtype=dtype)
+    y = y.astype(dtype=dtype)
     X = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
     y = _convert_to_dataframe(y, sycl_queue=queue, target_df=dataframe)
-    linreg = LinearRegression().fit(X, y)
-    if daal_check_version((2023, "P", 100)):
-        assert hasattr(linreg, "_onedal_estimator")
+
+    linreg = LinearRegression()
+    if daal_check_version((2024, "P", 0)) and macro_block is not None:
+        hparams = linreg.get_hyperparameters("fit")
+        hparams.cpu_macro_block = macro_block
+        hparams.gpu_macro_block = macro_block
+
+    linreg.fit(X, y)
+
+    assert hasattr(linreg, "_onedal_estimator")
     assert "sklearnex" in linreg.__module__
     assert linreg.n_features_in_ == 2
-    assert_allclose(_as_numpy(linreg.intercept_), 3.0)
-    assert_allclose(_as_numpy(linreg.coef_), [1.0, 2.0])
+
+    tol = 1e-5 if X.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():
@@ -75,3 +87,31 @@ def test_sklearnex_import_elastic():
     assert "daal4py" in elasticnet.__module__
     assert_allclose(elasticnet.intercept_, 1.451, atol=1e-3)
     assert_allclose(elasticnet.coef_, [18.838, 64.559], atol=1e-3)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_sklearnex_reconstruct_model(dataframe, queue, dtype):
+    from sklearnex.linear_model import LinearRegression
+
+    seed = 42
+    num_samples = 3500
+    num_features, num_targets = 14, 9
+
+    gen = np.random.default_rng(seed)
+    intercept = gen.random(size=num_targets, dtype=dtype)
+    coef = gen.random(size=(num_targets, num_features), dtype=dtype).T
+
+    X = gen.random(size=(num_samples, num_features), dtype=dtype)
+    gtr = X @ coef + intercept[np.newaxis, :]
+
+    X = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+
+    linreg = LinearRegression(fit_intercept=True)
+    linreg.coef_ = coef.T
+    linreg.intercept_ = intercept
+
+    y_pred = linreg.predict(X)
+
+    tol = 1e-5 if X.dtype == np.float32 else 1e-7
+    assert_allclose(gtr, _as_numpy(y_pred), rtol=tol)

sklearnex/linear_model/tests/test_logreg.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #
@@ -15,15 +14,78 @@
 # limitations under the License.
 # ===============================================================================
 
-import numpy as np
-from numpy.testing import assert_allclose
-from sklearn.datasets import load_iris
+import pytest
+from sklearn.datasets import load_breast_cancer, load_iris
+from sklearn.metrics import accuracy_score
+from sklearn.model_selection import train_test_split
 
+from daal4py.sklearn._utils import daal_check_version
+from onedal.tests.utils._dataframes_support import (
+    _as_numpy,
+    _convert_to_dataframe,
+    get_dataframes_and_queues,
+)
 
-def test_sklearnex_import():
+
+def prepare_input(X, y, dataframe, queue):
+    X_train, X_test, y_train, y_test = train_test_split(
+        X, y, train_size=0.8, random_state=42
+    )
+    X_train = _convert_to_dataframe(X_train, sycl_queue=queue, target_df=dataframe)
+    y_train = _convert_to_dataframe(y_train, sycl_queue=queue, target_df=dataframe)
+    X_test = _convert_to_dataframe(X_test, sycl_queue=queue, target_df=dataframe)
+    return X_train, X_test, y_train, y_test
+
+
+@pytest.mark.parametrize(
+    "dataframe,queue",
+    get_dataframes_and_queues(device_filter_="cpu"),
+)
+def test_sklearnex_multiclass_classification(dataframe, queue):
     from sklearnex.linear_model import LogisticRegression
 
     X, y = load_iris(return_X_y=True)
-    logreg = LogisticRegression(random_state=0, max_iter=200).fit(X, y)
-    assert "daal4py" in logreg.__module__
-    assert_allclose(logreg.score(X, y), 0.9733, atol=1e-3)
+    X_train, X_test, y_train, y_test = prepare_input(X, y, dataframe, queue)
+
+    logreg = LogisticRegression(fit_intercept=True, solver="lbfgs", max_iter=200).fit(
+        X_train, y_train
+    )
+
+    if daal_check_version((2024, "P", 1)):
+        assert "sklearnex" in logreg.__module__
+    else:
+        assert "daal4py" in logreg.__module__
+
+    y_pred = _as_numpy(logreg.predict(X_test))
+    assert accuracy_score(y_test, y_pred) > 0.99
+
+
+@pytest.mark.parametrize(
+    "dataframe,queue",
+    get_dataframes_and_queues(),
+)
+def test_sklearnex_binary_classification(dataframe, queue):
+    from sklearnex.linear_model import LogisticRegression
+
+    X, y = load_breast_cancer(return_X_y=True)
+    X_train, X_test, y_train, y_test = prepare_input(X, y, dataframe, queue)
+
+    logreg = LogisticRegression(fit_intercept=True, solver="newton-cg", max_iter=100).fit(
+        X_train, y_train
+    )
+
+    if daal_check_version((2024, "P", 1)):
+        assert "sklearnex" in logreg.__module__
+    else:
+        assert "daal4py" in logreg.__module__
+    if (
+        dataframe != "numpy"
+        and queue is not None
+        and queue.sycl_device.is_gpu
+        and daal_check_version((2024, "P", 1))
+    ):
+        # fit was done on gpu
+        assert hasattr(logreg, "_onedal_estimator")
+
+    y_pred = _as_numpy(logreg.predict(X_test))
+    assert accuracy_score(y_test, y_pred) > 0.95

sklearnex/manifold/__init__.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/manifold/t_sne.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/manifold/tests/test_tsne.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/metrics/__init__.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/metrics/pairwise.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/metrics/ranking.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/metrics/tests/test_metrics.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/model_selection/__init__.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/model_selection/split.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/model_selection/tests/test_model_selection.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #

sklearnex/neighbors/__init__.py

@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 # ===============================================================================
 # Copyright 2021 Intel Corporation
 #
@@ -15,10 +14,10 @@
 # limitations under the License.
 # ===============================================================================
 
+from ._lof import LocalOutlierFactor
 from .knn_classification import KNeighborsClassifier
 from .knn_regression import KNeighborsRegressor
 from .knn_unsupervised import NearestNeighbors
-from .lof import LocalOutlierFactor
 
 __all__ = [
     "KNeighborsClassifier",