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

sklearnex/basic_statistics/tests/test_basic_statistics.py

@@ -0,0 +1,251 @@
+# ==============================================================================
+# Copyright 2023 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 numpy as np
+import pytest
+from numpy.testing import assert_allclose
+
+from onedal.basic_statistics.tests.test_basic_statistics import (
+    expected_max,
+    expected_mean,
+    expected_sum,
+    options_and_tests,
+)
+from onedal.tests.utils._dataframes_support import (
+    _convert_to_dataframe,
+    get_dataframes_and_queues,
+)
+from sklearnex.basic_statistics import BasicStatistics
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+def test_sklearnex_import_basic_statistics(dataframe, queue):
+    X = np.array([[0, 0], [1, 1]])
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+
+    weights = np.array([1, 0.5])
+    weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+
+    result = BasicStatistics().fit(X_df)
+
+    expected_mean = np.array([0.5, 0.5])
+    expected_min = np.array([0, 0])
+    expected_max = np.array([1, 1])
+
+    assert_allclose(expected_mean, result.mean)
+    assert_allclose(expected_max, result.max)
+    assert_allclose(expected_min, result.min)
+
+    result = BasicStatistics().fit(X_df, sample_weight=weights_df)
+
+    expected_weighted_mean = np.array([0.25, 0.25])
+    expected_weighted_min = np.array([0, 0])
+    expected_weighted_max = np.array([0.5, 0.5])
+
+    assert_allclose(expected_weighted_mean, result.mean)
+    assert_allclose(expected_weighted_min, result.min)
+    assert_allclose(expected_weighted_max, result.max)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_multiple_options_on_gold_data(dataframe, queue, weighted, dtype):
+    X = np.array([[0, 0], [1, 1]])
+    X = X.astype(dtype=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = np.array([1, 0.5])
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics()
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    if weighted:
+        expected_weighted_mean = np.array([0.25, 0.25])
+        expected_weighted_min = np.array([0, 0])
+        expected_weighted_max = np.array([0.5, 0.5])
+        assert_allclose(expected_weighted_mean, result.mean)
+        assert_allclose(expected_weighted_max, result.max)
+        assert_allclose(expected_weighted_min, result.min)
+    else:
+        expected_mean = np.array([0.5, 0.5])
+        expected_min = np.array([0, 0])
+        expected_max = np.array([1, 1])
+        assert_allclose(expected_mean, result.mean)
+        assert_allclose(expected_max, result.max)
+        assert_allclose(expected_min, result.min)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("option", options_and_tests)
+@pytest.mark.parametrize("row_count", [100, 1000])
+@pytest.mark.parametrize("column_count", [10, 100])
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_single_option_on_random_data(
+    dataframe, queue, option, row_count, column_count, weighted, dtype
+):
+    result_option, function, tols = option
+    fp32tol, fp64tol = tols
+    seed = 77
+    gen = np.random.default_rng(seed)
+    X = gen.uniform(low=-0.3, high=+0.7, size=(row_count, column_count))
+    X = X.astype(dtype=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = gen.uniform(low=-0.5, high=1.0, size=row_count)
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics(result_options=result_option)
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    res = getattr(result, result_option)
+    if weighted:
+        weighted_data = np.diag(weights) @ X
+        gtr = function(weighted_data)
+    else:
+        gtr = function(X)
+
+    tol = fp32tol if res.dtype == np.float32 else fp64tol
+    assert_allclose(gtr, res, atol=tol)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("row_count", [100, 1000])
+@pytest.mark.parametrize("column_count", [10, 100])
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_multiple_options_on_random_data(
+    dataframe, queue, row_count, column_count, weighted, dtype
+):
+    seed = 77
+    gen = np.random.default_rng(seed)
+    X = gen.uniform(low=-0.3, high=+0.7, size=(row_count, column_count))
+    X = X.astype(dtype=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = gen.uniform(low=-0.5, high=1.0, size=row_count)
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics(result_options=["mean", "max", "sum"])
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    res_mean, res_max, res_sum = result.mean, result.max, result.sum
+    if weighted:
+        weighted_data = np.diag(weights) @ X
+        gtr_mean, gtr_max, gtr_sum = (
+            expected_mean(weighted_data),
+            expected_max(weighted_data),
+            expected_sum(weighted_data),
+        )
+    else:
+        gtr_mean, gtr_max, gtr_sum = (
+            expected_mean(X),
+            expected_max(X),
+            expected_sum(X),
+        )
+
+    tol = 5e-4 if res_mean.dtype == np.float32 else 1e-7
+    assert_allclose(gtr_mean, res_mean, atol=tol)
+    assert_allclose(gtr_max, res_max, atol=tol)
+    assert_allclose(gtr_sum, res_sum, atol=tol)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("row_count", [100, 1000])
+@pytest.mark.parametrize("column_count", [10, 100])
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_all_option_on_random_data(
+    dataframe, queue, row_count, column_count, weighted, dtype
+):
+    seed = 77
+    gen = np.random.default_rng(seed)
+    X = gen.uniform(low=-0.3, high=+0.7, size=(row_count, column_count))
+    X = X.astype(dtype=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = gen.uniform(low=-0.5, high=+1.0, size=row_count)
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics(result_options="all")
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    if weighted:
+        weighted_data = np.diag(weights) @ X
+
+    for option in options_and_tests:
+        result_option, function, tols = option
+        fp32tol, fp64tol = tols
+        res = getattr(result, result_option)
+        if weighted:
+            gtr = function(weighted_data)
+        else:
+            gtr = function(X)
+        tol = fp32tol if res.dtype == np.float32 else fp64tol
+        assert_allclose(gtr, res, atol=tol)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("option", options_and_tests)
+@pytest.mark.parametrize("data_size", [100, 1000])
+@pytest.mark.parametrize("weighted", [True, False])
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+def test_1d_input_on_random_data(dataframe, queue, option, data_size, weighted, dtype):
+    result_option, function, tols = option
+    fp32tol, fp64tol = tols
+    seed = 77
+    gen = np.random.default_rng(seed)
+    X = gen.uniform(low=-0.3, high=+0.7, size=data_size)
+    X = X.astype(dtype=dtype)
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+    if weighted:
+        weights = gen.uniform(low=-0.5, high=1.0, size=data_size)
+        weights = weights.astype(dtype=dtype)
+        weights_df = _convert_to_dataframe(weights, sycl_queue=queue, target_df=dataframe)
+    basicstat = BasicStatistics(result_options=result_option)
+
+    if weighted:
+        result = basicstat.fit(X_df, sample_weight=weights_df)
+    else:
+        result = basicstat.fit(X_df)
+
+    res = getattr(result, result_option)
+    if weighted:
+        weighted_data = weights * X
+        gtr = function(weighted_data)
+    else:
+        gtr = function(X)
+
+    tol = fp32tol if res.dtype == np.float32 else fp64tol
+    assert_allclose(gtr, res, atol=tol)

sklearnex/basic_statistics/tests/test_incremental_basic_statistics.py

@@ -18,7 +18,7 @@ import numpy as np
 import pytest
 from numpy.testing import assert_allclose
 
-from onedal.basic_statistics.tests.test_incremental_basic_statistics import (
+from onedal.basic_statistics.tests.test_basic_statistics import (
     expected_max,
     expected_mean,
     expected_sum,

sklearnex/cluster/dbscan.py

@@ -17,7 +17,6 @@
 import numbers
 from abc import ABC
 
-import numpy as np
 from scipy import sparse as sp
 from sklearn.cluster import DBSCAN as sklearn_DBSCAN
 from sklearn.utils.validation import _check_sample_weight
@@ -85,6 +84,9 @@ class DBSCAN(sklearn_DBSCAN, BaseDBSCAN):
         self.n_jobs = n_jobs
 
     def _onedal_fit(self, X, y, sample_weight=None, queue=None):
+        if sklearn_check_version("1.0"):
+            X = self._validate_data(X, force_all_finite=False)
+
         onedal_params = {
             "eps": self.eps,
             "min_samples": self.min_samples,

sklearnex/cluster/k_means.py

@@ -15,3 +15,11 @@
 # ===============================================================================
 
 from daal4py.sklearn.cluster import KMeans
+from onedal._device_offload import support_usm_ndarray
+
+# Note: `sklearnex.cluster.KMeans` only has functional
+# sycl GPU support. No GPU device will be offloaded.
+KMeans.fit = support_usm_ndarray(queue_param=False)(KMeans.fit)
+KMeans.fit_predict = support_usm_ndarray(queue_param=False)(KMeans.fit_predict)
+KMeans.predict = support_usm_ndarray(queue_param=False)(KMeans.predict)
+KMeans.score = support_usm_ndarray(queue_param=False)(KMeans.score)

sklearnex/cluster/tests/test_dbscan.py

@@ -18,16 +18,18 @@ import numpy as np
 import pytest
 from numpy.testing import assert_allclose
 
+from onedal.tests.utils._dataframes_support import (
+    _convert_to_dataframe,
+    get_dataframes_and_queues,
+)
 
-# TODO:
-# adding this parameterized testing
-# somehow breaks other test with preview module patch:
-# sklearnex/tests/test_monkeypatch.py::test_preview_namespace.
-# @pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
-def test_sklearnex_import_dbscan():
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+def test_sklearnex_import_dbscan(dataframe, queue):
     from sklearnex.cluster import DBSCAN
 
     X = np.array([[1, 2], [2, 2], [2, 3], [8, 7], [8, 8], [25, 80]])
+    X = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
     dbscan = DBSCAN(eps=3, min_samples=2).fit(X)
     assert "sklearnex" in dbscan.__module__
 

sklearnex/cluster/tests/test_kmeans.py

@@ -15,16 +15,28 @@
 # ===============================================================================
 
 import numpy as np
+import pytest
 from numpy.testing import assert_allclose
 
+from onedal.tests.utils._dataframes_support import (
+    _as_numpy,
+    _convert_to_dataframe,
+    get_dataframes_and_queues,
+)
+
+
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+def test_sklearnex_import(dataframe, queue):
 
-def test_sklearnex_import():
     from sklearnex.cluster import KMeans
 
     X = np.array([[1, 2], [1, 4], [1, 0], [10, 2], [10, 4], [10, 0]])
+    X = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
     kmeans = KMeans(n_clusters=2, random_state=0).fit(X)
     assert "daal4py" in kmeans.__module__
 
-    result = kmeans.predict([[0, 0], [12, 3]])
+    X_test = [[0, 0], [12, 3]]
+    X_test = _convert_to_dataframe(X_test, sycl_queue=queue, target_df=dataframe)
+    result = kmeans.predict(X_test)
     expected = np.array([1, 0], dtype=np.int32)
-    assert_allclose(expected, result)
+    assert_allclose(expected, _as_numpy(result))

sklearnex/conftest.py

@@ -19,7 +19,8 @@ import logging
 
 import pytest
 
-from sklearnex import patch_sklearn, unpatch_sklearn
+from daal4py.sklearn._utils import sklearn_check_version
+from sklearnex import config_context, patch_sklearn, unpatch_sklearn
 
 
 def pytest_configure(config):
@@ -61,3 +62,12 @@ def with_sklearnex():
     patch_sklearn()
     yield
     unpatch_sklearn()
+
+
+@pytest.fixture
+def with_array_api():
+    if sklearn_check_version("1.2"):
+        with config_context(array_api_dispatch=True):
+            yield
+    else:
+        yield

sklearnex/covariance/incremental_covariance.py

@@ -19,13 +19,14 @@ import warnings
 
 import numpy as np
 from scipy import linalg
-from sklearn.base import BaseEstimator
+from sklearn.base import BaseEstimator, clone
 from sklearn.covariance import EmpiricalCovariance as sklearn_EmpiricalCovariance
+from sklearn.covariance import log_likelihood
 from sklearn.utils import check_array, gen_batches
+from sklearn.utils.validation import _num_features
 
 from daal4py.sklearn._n_jobs_support import control_n_jobs
 from daal4py.sklearn._utils import daal_check_version, sklearn_check_version
-from onedal._device_offload import support_usm_ndarray
 from onedal.covariance import (
     IncrementalEmpiricalCovariance as onedal_IncrementalEmpiricalCovariance,
 )
@@ -34,6 +35,7 @@ from sklearnex import config_context
 from .._device_offload import dispatch, wrap_output_data
 from .._utils import PatchingConditionsChain, register_hyperparameters
 from ..metrics import pairwise_distances
+from ..utils import get_namespace
 
 if sklearn_check_version("1.2"):
     from sklearn.utils._param_validation import Interval
@@ -98,7 +100,6 @@ class IncrementalEmpiricalCovariance(BaseEstimator):
 
     get_precision = sklearn_EmpiricalCovariance.get_precision
     error_norm = wrap_output_data(sklearn_EmpiricalCovariance.error_norm)
-    score = wrap_output_data(sklearn_EmpiricalCovariance.score)
 
     def __init__(
         self, *, store_precision=False, assume_centered=False, batch_size=None, copy=True
@@ -197,6 +198,43 @@ class IncrementalEmpiricalCovariance(BaseEstimator):
 
         return self
 
+    @wrap_output_data
+    def score(self, X_test, y=None):
+        xp, _ = get_namespace(X_test)
+
+        location = self.location_
+        if sklearn_check_version("1.0"):
+            X = self._validate_data(
+                X_test,
+                dtype=[np.float64, np.float32],
+                reset=False,
+            )
+        else:
+            X = check_array(
+                X_test,
+                dtype=[np.float64, np.float32],
+            )
+
+        if "numpy" not in xp.__name__:
+            location = xp.asarray(location, device=X_test.device)
+            # depending on the sklearn version, check_array
+            # and validate_data will return only numpy arrays
+            # which will break dpnp/dpctl support. If the
+            # array namespace isn't from numpy and the data
+            # is now a numpy array, it has been validated and
+            # the original can be used.
+            if isinstance(X, np.ndarray):
+                X = X_test
+
+        est = clone(self)
+        est.set_params(**{"assume_centered": True})
+
+        # test_cov is a numpy array, but calculated on device
+        test_cov = est.fit(X - location).covariance_
+        res = log_likelihood(test_cov, self.get_precision())
+
+        return res
+
     def partial_fit(self, X, y=None, check_input=True):
         """
         Incremental fit with X. All of X is processed as a single batch.
@@ -293,21 +331,34 @@ class IncrementalEmpiricalCovariance(BaseEstimator):
         return self
 
     # expose sklearnex pairwise_distances if mahalanobis distance eventually supported
-    @wrap_output_data
     def mahalanobis(self, X):
         if sklearn_check_version("1.0"):
-            self._validate_data(X, reset=False, copy=self.copy)
-        else:
-            check_array(X, copy=self.copy)
+            self._check_feature_names(X, reset=False)
 
+        xp, _ = get_namespace(X)
         precision = self.get_precision()
-        with config_context(assume_finite=True):
-            # compute mahalanobis distances
-            dist = pairwise_distances(
-                X, self.location_[np.newaxis, :], metric="mahalanobis", VI=precision
-            )
+        # compute mahalanobis distances
+        # pairwise_distances will check n_features (via n_feature matching with
+        # self.location_) , and will check for finiteness via check array
+        # check_feature_names will match _validate_data functionally
+        location = self.location_[np.newaxis, :]
+        if "numpy" not in xp.__name__:
+            # Guarantee that inputs to pairwise_distances match in type and location
+            location = xp.asarray(location, device=X.device)
+
+        try:
+            dist = pairwise_distances(X, location, metric="mahalanobis", VI=precision)
+        except ValueError as e:
+            # Throw the expected sklearn error in an n_feature length violation
+            if "Incompatible dimension for X and Y matrices: X.shape[1] ==" in str(e):
+                raise ValueError(
+                    f"X has {_num_features(X)} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_in_} features as input."
+                )
+            else:
+                raise e
 
-        return np.reshape(dist, (len(X),)) ** 2
+        return (xp.reshape(dist, (-1,))) ** 2
 
     _onedal_cpu_supported = _onedal_supported
     _onedal_gpu_supported = _onedal_supported

sklearnex/covariance/tests/test_incremental_covariance.py

@@ -16,13 +16,18 @@
 
 import numpy as np
 import pytest
+from numpy.linalg import slogdet
 from numpy.testing import assert_allclose
+from scipy.linalg import pinvh
 from sklearn.covariance.tests.test_covariance import (
     test_covariance,
     test_EmpiricalCovariance_validates_mahalanobis,
 )
+from sklearn.datasets import load_diabetes
+from sklearn.decomposition import PCA
 
 from onedal.tests.utils._dataframes_support import (
+    _as_numpy,
     _convert_to_dataframe,
     get_dataframes_and_queues,
 )
@@ -163,6 +168,36 @@ def test_sklearnex_fit_on_random_data(
     assert_allclose(expected_means, result.location_, atol=1e-6)
 
 
+@pytest.mark.parametrize("dataframe,queue", get_dataframes_and_queues())
+def test_whitened_toy_score(dataframe, queue):
+    from sklearnex.covariance import IncrementalEmpiricalCovariance
+
+    # Load a sklearn toy dataset with sufficient data
+    X, _ = load_diabetes(return_X_y=True)
+    n = X.shape[1]
+
+    # Transform the data into uncorrelated, unity variance components
+    X = PCA(whiten=True).fit_transform(X)
+
+    # change dataframe
+    X_df = _convert_to_dataframe(X, sycl_queue=queue, target_df=dataframe)
+
+    # fit data
+    est = IncrementalEmpiricalCovariance()
+    est.fit(X_df)
+    # location_ attribute approximately zero (10,), covariance_ identity (10,10)
+
+    # The log-likelihood can be calculated simply due to covariance_
+    # use of scipy.linalg.pinvh, np.linalg.sloget and np.cov for estimator
+    # independence
+    expected_result = (
+        -(n - slogdet(pinvh(np.cov(X.T, bias=1)))[1] + n * np.log(2 * np.pi)) / 2
+    )
+    # expected_result = -14.1780602988
+    result = _as_numpy(est.score(X_df))
+    assert_allclose(expected_result, result, atol=1e-6)
+
+
 # Monkeypatch IncrementalEmpiricalCovariance into relevant sklearn.covariance tests
 @pytest.mark.allow_sklearn_fallback
 @pytest.mark.parametrize(

sklearnex/decomposition/pca.py

@@ -32,6 +32,7 @@ if daal_check_version((2024, "P", 100)):
 
     from .._device_offload import dispatch, wrap_output_data
     from .._utils import PatchingConditionsChain
+    from ..utils import get_namespace
 
     if sklearn_check_version("1.1") and not sklearn_check_version("1.2"):
         from sklearn.utils import check_scalar
@@ -42,7 +43,6 @@ if daal_check_version((2024, "P", 100)):
     from sklearn.decomposition import PCA as sklearn_PCA
 
     from onedal.decomposition import PCA as onedal_PCA
-    from sklearnex.utils import get_namespace
 
     @control_n_jobs(decorated_methods=["fit", "transform", "fit_transform"])
     class PCA(sklearn_PCA):
@@ -210,6 +210,29 @@ if daal_check_version((2024, "P", 100)):
                 # Scikit-learn PCA["covariance_eigh"] was fit
                 return self._transform(X_fit, xp, x_is_centered=x_is_centered)
 
+        @wrap_output_data
+        def inverse_transform(self, X):
+            xp, _ = get_namespace(X)
+
+            mean = self.mean_
+            if self.whiten:
+                components = (
+                    xp.sqrt(self.explained_variance_[:, np.newaxis]) * self.components_
+                )
+            else:
+                components = self.components_
+
+            if "numpy" not in xp.__name__:
+                # DPCtl and dpnp require inputs to be on the same device for
+                # matrix multiplication and division. The type and location
+                # of the components and mean are dependent on the sklearn
+                # version, this makes sure it is of the same type and on the
+                # same device as the data (compute follows data).
+                components = xp.asarray(components, device=X.device)
+                mean = xp.asarray(mean, device=X.device)
+
+            return X @ components + mean
+
         def _onedal_supported(self, method_name, X):
             class_name = self.__class__.__name__
             patching_status = PatchingConditionsChain(
@@ -381,6 +404,7 @@ if daal_check_version((2024, "P", 100)):
         fit.__doc__ = sklearn_PCA.fit.__doc__
         transform.__doc__ = sklearn_PCA.transform.__doc__
         fit_transform.__doc__ = sklearn_PCA.fit_transform.__doc__
+        inverse_transform.__doc__ = sklearn_PCA.inverse_transform.__doc__
 
 else:
     from daal4py.sklearn.decomposition import PCA

sklearnex/decomposition/tests/test_pca.py

@@ -51,6 +51,8 @@ def test_sklearnex_import(dataframe, queue):
         assert hasattr(pca, "_onedal_estimator")
     else:
         assert "daal4py" in pca.__module__
+
+    tol = 1e-5 if _as_numpy(X_transformed).dtype == np.float32 else 1e-7
     assert_allclose([6.30061232, 0.54980396], _as_numpy(pca.singular_values_))
-    assert_allclose(X_transformed_expected, _as_numpy(X_transformed))
-    assert_allclose(X_transformed_expected, _as_numpy(X_fit_transformed))
+    assert_allclose(X_transformed_expected, _as_numpy(X_transformed), rtol=tol)
+    assert_allclose(X_transformed_expected, _as_numpy(X_fit_transformed), rtol=tol)