scikit-learn-intelex 2024.4.0__py310-none-win_amd64.whl → 2024.6.0__py310-none-win_amd64.whl

{scikit_learn_intelex-2024.4.0.data → scikit_learn_intelex-2024.6.0.data}/data/Lib/site-packages/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)

{scikit_learn_intelex-2024.4.0.data → scikit_learn_intelex-2024.6.0.data}/data/Lib/site-packages/sklearnex/tests/test_patching.py

@@ -61,12 +61,15 @@ def test_pairwise_distances_patching(caplog, dataframe, queue, dtype, metric):
             pytest.skip("pairwise_distances does not support GPU queues")
 
         rng = nprnd.default_rng()
-        X = _convert_to_dataframe(
-            rng.random(size=1000).reshape(1, -1),
-            sycl_queue=queue,
-            target_df=dataframe,
-            dtype=dtype,
-        )
+        if dataframe == "pandas":
+            X = _convert_to_dataframe(
+                rng.random(size=1000).astype(dtype).reshape(1, -1),
+                target_df=dataframe,
+            )
+        else:
+            X = _convert_to_dataframe(
+                rng.random(size=1000), sycl_queue=queue, target_df=dataframe, dtype=dtype
+            )[None, :]
 
         _ = pairwise_distances(X, metric=metric)
     assert all(
@@ -90,14 +93,17 @@ def test_roc_auc_score_patching(caplog, dataframe, queue, dtype):
 
     with caplog.at_level(logging.WARNING, logger="sklearnex"):
         rng = nprnd.default_rng()
+        X = rng.integers(2, size=1000)
+        y = rng.integers(2, size=1000)
+
         X = _convert_to_dataframe(
-            rng.integers(2, size=1000),
+            X,
             sycl_queue=queue,
             target_df=dataframe,
             dtype=dtype,
         )
         y = _convert_to_dataframe(
-            rng.integers(2, size=1000),
+            y,
             sycl_queue=queue,
             target_df=dataframe,
             dtype=dtype,
@@ -142,10 +148,16 @@ def test_standard_estimator_patching(caplog, dataframe, queue, dtype, estimator,
             and dtype in [np.uint32, np.uint64]
         ):
             pytest.skip("Windows segmentation fault for Ridge.predict for unsigned ints")
+        elif estimator == "IncrementalLinearRegression" and np.issubdtype(
+            dtype, np.integer
+        ):
+            pytest.skip(
+                "IncrementalLinearRegression fails on oneDAL side with int types because dataset is filled by zeroes"
+            )
         elif method and not hasattr(est, method):
             pytest.skip(f"sklearn available_if prevents testing {estimator}.{method}")
 
-        X, y = gen_dataset(est, queue=queue, target_df=dataframe, dtype=dtype)
+        X, y = gen_dataset(est, queue=queue, target_df=dataframe, dtype=dtype)[0]
         est.fit(X, y)
 
         if method:
@@ -177,7 +189,7 @@ def test_special_estimator_patching(caplog, dataframe, queue, dtype, estimator,
         elif dtype == np.float64 and queue and not queue.sycl_device.has_aspect_fp64:
             pytest.skip("Hardware does not support fp64 SYCL testing")
 
-        X, y = gen_dataset(est, queue=queue, target_df=dataframe, dtype=dtype)
+        X, y = gen_dataset(est, queue=queue, target_df=dataframe, dtype=dtype)[0]
         est.fit(X, y)
 
         if method and not hasattr(est, method):

scikit_learn_intelex-2024.6.0.data/data/Lib/site-packages/sklearnex/tests/test_run_to_run_stability.py

@@ -0,0 +1,283 @@
+# ===============================================================================
+# Copyright 2020 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 random
+from collections.abc import Iterable
+from functools import partial
+from numbers import Number
+
+import numpy as np
+import pytest
+from _utils import (
+    PATCHED_MODELS,
+    SPECIAL_INSTANCES,
+    _sklearn_clone_dict,
+    gen_dataset,
+    gen_models_info,
+)
+from numpy.testing import assert_allclose
+from scipy import sparse
+from sklearn.datasets import (
+    load_breast_cancer,
+    load_diabetes,
+    load_iris,
+    make_classification,
+    make_regression,
+)
+
+import daal4py as d4p
+from onedal.tests.utils._dataframes_support import _as_numpy, get_dataframes_and_queues
+from sklearnex.cluster import DBSCAN, KMeans
+from sklearnex.decomposition import PCA
+from sklearnex.metrics import pairwise_distances, roc_auc_score
+from sklearnex.model_selection import train_test_split
+from sklearnex.neighbors import (
+    KNeighborsClassifier,
+    KNeighborsRegressor,
+    NearestNeighbors,
+)
+from sklearnex.svm import SVC
+
+# to reproduce errors even in CI
+d4p.daalinit(nthreads=100)
+
+_dataset_dict = {
+    "classification": [
+        partial(load_iris, return_X_y=True),
+        partial(load_breast_cancer, return_X_y=True),
+    ],
+    "regression": [
+        partial(load_diabetes, return_X_y=True),
+        partial(
+            make_regression, n_samples=500, n_features=10, noise=64.0, random_state=42
+        ),
+    ],
+}
+
+
+def eval_method(X, y, est, method):
+    res = []
+    est.fit(X, y)
+
+    if method:
+        if method != "score":
+            res = getattr(est, method)(X)
+        else:
+            res = est.score(X, y)
+
+    if not isinstance(res, Iterable):
+        res = [res]
+
+    # if estimator follows sklearn design rules, then set attributes should have a
+    # trailing underscore
+    attributes = [
+        i
+        for i in dir(est)
+        if hasattr(est, i) and not i.startswith("_") and i.endswith("_")
+    ]
+    results = [getattr(est, i) for i in attributes] + [_as_numpy(i) for i in res]
+    attributes += [method for i in res]
+    return results, attributes
+
+
+def _run_test(estimator, method, datasets):
+
+    for X, y in datasets:
+        baseline, attributes = eval_method(X, y, estimator, method)
+
+        for i in range(10):
+            res, _ = eval_method(X, y, estimator, method)
+
+            for r, b, n in zip(res, baseline, attributes):
+                if (
+                    isinstance(b, Number)
+                    or hasattr(b, "__array__")
+                    or hasattr(b, "__array_namespace__")
+                    or hasattr(b, "__sycl_usm_ndarray__")
+                ):
+                    assert_allclose(
+                        r, b, rtol=0.0, atol=0.0, err_msg=str(n + " is incorrect")
+                    )
+
+
+SPARSE_INSTANCES = _sklearn_clone_dict(
+    {
+        str(i): i
+        for i in [
+            SVC(),
+            KMeans(),
+            KMeans(init="random"),
+        ]
+    }
+)
+
+STABILITY_INSTANCES = _sklearn_clone_dict(
+    {
+        str(i): i
+        for i in [
+            KNeighborsClassifier(algorithm="brute", weights="distance"),
+            KNeighborsClassifier(algorithm="kd_tree", weights="distance"),
+            KNeighborsClassifier(algorithm="kd_tree"),
+            KNeighborsRegressor(algorithm="brute", weights="distance"),
+            KNeighborsRegressor(algorithm="kd_tree", weights="distance"),
+            KNeighborsRegressor(algorithm="kd_tree"),
+            NearestNeighbors(algorithm="kd_tree"),
+            DBSCAN(algorithm="brute"),
+            PCA(n_components=0.5, svd_solver="covariance_eigh"),
+            KMeans(init="random"),
+        ]
+    }
+)
+
+
+@pytest.mark.parametrize("dataframe, queue", get_dataframes_and_queues("numpy"))
+@pytest.mark.parametrize("estimator, method", gen_models_info(PATCHED_MODELS))
+def test_standard_estimator_stability(estimator, method, dataframe, queue):
+    if estimator in ["LogisticRegression", "TSNE"]:
+        pytest.skip(f"stability not guaranteed for {estimator}")
+    if estimator in ["KMeans", "PCA"] and method == "score" and queue == None:
+        pytest.skip(f"variation observed in {estimator}.score")
+
+    est = PATCHED_MODELS[estimator]()
+
+    if method and not hasattr(est, method):
+        pytest.skip(f"sklearn available_if prevents testing {estimator}.{method}")
+
+    params = est.get_params().copy()
+    if "random_state" in params:
+        params["random_state"] = 0
+        est.set_params(**params)
+
+    datasets = gen_dataset(est, datasets=_dataset_dict, queue=queue, target_df=dataframe)
+    _run_test(est, method, datasets)
+
+
+@pytest.mark.allow_sklearn_fallback
+@pytest.mark.parametrize("dataframe, queue", get_dataframes_and_queues("numpy"))
+@pytest.mark.parametrize("estimator, method", gen_models_info(SPECIAL_INSTANCES))
+def test_special_estimator_stability(estimator, method, dataframe, queue):
+    if queue is None and estimator in ["LogisticRegression(solver='newton-cg')"]:
+        pytest.skip(f"stability not guaranteed for {estimator}")
+    if "KMeans" in estimator and method == "score" and queue == None:
+        pytest.skip(f"variation observed in KMeans.score")
+
+    est = SPECIAL_INSTANCES[estimator]
+
+    if method and not hasattr(est, method):
+        pytest.skip(f"sklearn available_if prevents testing {estimator}.{method}")
+
+    params = est.get_params().copy()
+    if "random_state" in params:
+        params["random_state"] = 0
+        est.set_params(**params)
+
+    datasets = gen_dataset(est, datasets=_dataset_dict, queue=queue, target_df=dataframe)
+    _run_test(est, method, datasets)
+
+
+@pytest.mark.parametrize("dataframe, queue", get_dataframes_and_queues("numpy"))
+@pytest.mark.parametrize("estimator, method", gen_models_info(SPARSE_INSTANCES))
+def test_sparse_estimator_stability(estimator, method, dataframe, queue):
+    if "KMeans" in estimator and method == "score" and queue == None:
+        pytest.skip(f"variation observed in KMeans.score")
+
+    est = SPARSE_INSTANCES[estimator]
+
+    if method and not hasattr(est, method):
+        pytest.skip(f"sklearn available_if prevents testing {estimator}.{method}")
+
+    params = est.get_params().copy()
+    if "random_state" in params:
+        params["random_state"] = 0
+        est.set_params(**params)
+
+    datasets = gen_dataset(
+        est, sparse=True, datasets=_dataset_dict, queue=queue, target_df=dataframe
+    )
+    _run_test(est, method, datasets)
+
+
+@pytest.mark.parametrize("dataframe, queue", get_dataframes_and_queues("numpy"))
+@pytest.mark.parametrize("estimator, method", gen_models_info(STABILITY_INSTANCES))
+def test_other_estimator_stability(estimator, method, dataframe, queue):
+    if "KMeans" in estimator and method == "score" and queue == None:
+        pytest.skip(f"variation observed in KMeans.score")
+
+    est = STABILITY_INSTANCES[estimator]
+
+    if method and not hasattr(est, method):
+        pytest.skip(f"sklearn available_if prevents testing {estimator}.{method}")
+
+    params = est.get_params().copy()
+    if "random_state" in params:
+        params["random_state"] = 0
+        est.set_params(**params)
+
+    datasets = gen_dataset(est, datasets=_dataset_dict, queue=queue, target_df=dataframe)
+    _run_test(est, method, datasets)
+
+
+@pytest.mark.parametrize("features", range(5, 10))
+def test_train_test_split(features):
+    X, y = make_classification(
+        n_samples=4000,
+        n_features=features,
+        n_informative=features,
+        n_redundant=0,
+        n_clusters_per_class=8,
+        random_state=0,
+    )
+    (
+        baseline_X_train,
+        baseline_X_test,
+        baseline_y_train,
+        baseline_y_test,
+    ) = train_test_split(X, y, test_size=0.33, random_state=0)
+    baseline = [baseline_X_train, baseline_X_test, baseline_y_train, baseline_y_test]
+    for _ in range(10):
+        X_train, X_test, y_train, y_test = train_test_split(
+            X, y, test_size=0.33, random_state=0
+        )
+        res = [X_train, X_test, y_train, y_test]
+        for a, b in zip(res, baseline):
+            np.testing.assert_allclose(
+                a, b, rtol=0.0, atol=0.0, err_msg=str("train_test_split is incorrect")
+            )
+
+
+@pytest.mark.parametrize("metric", ["cosine", "correlation"])
+def test_pairwise_distances(metric):
+    X = np.random.rand(1000)
+    X = np.array(X, dtype=np.float64)
+    baseline = pairwise_distances(X.reshape(1, -1), metric=metric)
+    for _ in range(5):
+        res = pairwise_distances(X.reshape(1, -1), metric=metric)
+        for a, b in zip(res, baseline):
+            np.testing.assert_allclose(
+                a, b, rtol=0.0, atol=0.0, err_msg=str("pairwise_distances is incorrect")
+            )
+
+
+@pytest.mark.parametrize("array_size", [100, 1000, 10000])
+def test_roc_auc(array_size):
+    a = [random.randint(0, 1) for i in range(array_size)]
+    b = [random.randint(0, 1) for i in range(array_size)]
+    baseline = roc_auc_score(a, b)
+    for _ in range(5):
+        res = roc_auc_score(a, b)
+        np.testing.assert_allclose(
+            baseline, res, rtol=0.0, atol=0.0, err_msg=str("roc_auc is incorrect")
+        )

{scikit_learn_intelex-2024.4.0.data → scikit_learn_intelex-2024.6.0.data}/data/Lib/site-packages/sklearnex/utils/_namespace.py

@@ -94,4 +94,4 @@ def get_namespace(*arrays):
     elif sklearn_check_version("1.2"):
         return sklearn_get_namespace(*arrays)
     else:
-        return np, True
+        return np, False

scikit_learn_intelex-2024.6.0.data/data/Lib/site-packages/sklearnex/utils/tests/test_finite.py

@@ -0,0 +1,89 @@
+# ==============================================================================
+# 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 time
+
+import numpy as np
+import numpy.random as rand
+import pytest
+from numpy.testing import assert_raises
+
+from sklearnex.utils import _assert_all_finite
+
+
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+@pytest.mark.parametrize(
+    "shape",
+    [
+        [16, 2048],
+        [
+            2**16 + 3,
+        ],
+        [1000, 1000],
+    ],
+)
+@pytest.mark.parametrize("allow_nan", [False, True])
+def test_sum_infinite_actually_finite(dtype, shape, allow_nan):
+    X = np.array(shape, dtype=dtype)
+    X.fill(np.finfo(dtype).max)
+    _assert_all_finite(X, allow_nan=allow_nan)
+
+
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+@pytest.mark.parametrize(
+    "shape",
+    [
+        [16, 2048],
+        [
+            2**16 + 3,
+        ],
+        [1000, 1000],
+    ],
+)
+@pytest.mark.parametrize("allow_nan", [False, True])
+@pytest.mark.parametrize("check", ["inf", "NaN", None])
+@pytest.mark.parametrize("seed", [0, int(time.time())])
+def test_assert_finite_random_location(dtype, shape, allow_nan, check, seed):
+    rand.seed(seed)
+    X = rand.uniform(high=np.finfo(dtype).max, size=shape).astype(dtype)
+
+    if check:
+        loc = rand.randint(0, X.size - 1)
+        X.reshape((-1,))[loc] = float(check)
+
+    if check is None or (allow_nan and check == "NaN"):
+        _assert_all_finite(X, allow_nan=allow_nan)
+    else:
+        assert_raises(ValueError, _assert_all_finite, X, allow_nan=allow_nan)
+
+
+@pytest.mark.parametrize("dtype", [np.float32, np.float64])
+@pytest.mark.parametrize("allow_nan", [False, True])
+@pytest.mark.parametrize("check", ["inf", "NaN", None])
+@pytest.mark.parametrize("seed", [0, int(time.time())])
+def test_assert_finite_random_shape_and_location(dtype, allow_nan, check, seed):
+    lb, ub = 32768, 1048576  # lb is a patching condition, ub 2^20
+    rand.seed(seed)
+    X = rand.uniform(high=np.finfo(dtype).max, size=rand.randint(lb, ub)).astype(dtype)
+
+    if check:
+        loc = rand.randint(0, X.size - 1)
+        X[loc] = float(check)
+
+    if check is None or (allow_nan and check == "NaN"):
+        _assert_all_finite(X, allow_nan=allow_nan)
+    else:
+        assert_raises(ValueError, _assert_all_finite, X, allow_nan=allow_nan)