scikit-learn-intelex 2024.2.0__py310-none-manylinux1_x86_64.whl → 2024.4.0__py310-none-manylinux1_x86_64.whl

sklearnex/tests/test_patching.py

@@ -14,107 +14,306 @@
 # limitations under the License.
 # ==============================================================================
 
+
+import importlib
 import inspect
+import logging
 import os
-import pathlib
 import re
-import subprocess
 import sys
-from inspect import isclass
+from inspect import signature
 
+import numpy as np
+import numpy.random as nprnd
 import pytest
-from _models_info import TO_SKIP
 from sklearn.base import BaseEstimator
 
-from sklearnex import get_patch_map, is_patched_instance, patch_sklearn, unpatch_sklearn
-
-
-def get_branch(s):
-    if len(s) == 0:
-        return "NO INFO"
-    for i in s:
-        if "failed to run accelerated version, fallback to original Scikit-learn" in i:
-            return "was in OPT, but go in Scikit"
-    for i in s:
-        if "running accelerated version" in i:
-            return "OPT"
-    return "Scikit"
-
-
-def run_parse(mas, result):
-    name, dtype = mas[0].split()
-    temp = []
-    INFO_POS = 16
-    for i in range(1, len(mas)):
-        mas[i] = mas[i][INFO_POS:]  # remove 'SKLEARNEX INFO: '
-        if not mas[i].startswith("sklearn"):
-            ind = name + " " + dtype + " " + mas[i]
-            result[ind] = get_branch(temp)
-            temp.clear()
-        else:
-            temp.append(mas[i])
-
-
-def get_result_log():
-    os.environ["SKLEARNEX_VERBOSE"] = "INFO"
-    absolute_path = str(pathlib.Path(__file__).parent.absolute())
-    try:
-        process = subprocess.check_output(
-            [sys.executable, absolute_path + "/utils/_launch_algorithms.py"]
+from daal4py.sklearn._utils import sklearn_check_version
+from onedal.tests.utils._dataframes_support import (
+    _convert_to_dataframe,
+    get_dataframes_and_queues,
+)
+from sklearnex import is_patched_instance
+from sklearnex.dispatcher import _is_preview_enabled
+from sklearnex.metrics import pairwise_distances, roc_auc_score
+from sklearnex.tests._utils import (
+    DTYPES,
+    PATCHED_FUNCTIONS,
+    PATCHED_MODELS,
+    SPECIAL_INSTANCES,
+    UNPATCHED_FUNCTIONS,
+    UNPATCHED_MODELS,
+    gen_dataset,
+    gen_models_info,
+)
+
+
+@pytest.mark.parametrize("dtype", DTYPES)
+@pytest.mark.parametrize("dataframe, queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("metric", ["cosine", "correlation"])
+def test_pairwise_distances_patching(caplog, dataframe, queue, dtype, metric):
+    with caplog.at_level(logging.WARNING, logger="sklearnex"):
+        if dtype == np.float16 and queue and not queue.sycl_device.has_aspect_fp16:
+            pytest.skip("Hardware does not support fp16 SYCL testing")
+        elif dtype == np.float64 and queue and not queue.sycl_device.has_aspect_fp64:
+            pytest.skip("Hardware does not support fp64 SYCL testing")
+        elif queue and queue.sycl_device.is_gpu:
+            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,
+        )
+
+        _ = pairwise_distances(X, metric=metric)
+    assert all(
+        [
+            "running accelerated version" in i.message
+            or "fallback to original Scikit-learn" in i.message
+            for i in caplog.records
+        ]
+    ), f"sklearnex patching issue in pairwise_distances with log: \n{caplog.text}"
+
+
+@pytest.mark.parametrize(
+    "dtype", [i for i in DTYPES if "32" in i.__name__ or "64" in i.__name__]
+)
+@pytest.mark.parametrize("dataframe, queue", get_dataframes_and_queues())
+def test_roc_auc_score_patching(caplog, dataframe, queue, dtype):
+    if dtype in [np.uint32, np.uint64] and sys.platform == "win32":
+        pytest.skip("Windows issue with unsigned ints")
+    elif dtype == np.float64 and queue and not queue.sycl_device.has_aspect_fp64:
+        pytest.skip("Hardware does not support fp64 SYCL testing")
+
+    with caplog.at_level(logging.WARNING, logger="sklearnex"):
+        rng = nprnd.default_rng()
+        X = _convert_to_dataframe(
+            rng.integers(2, size=1000),
+            sycl_queue=queue,
+            target_df=dataframe,
+            dtype=dtype,
+        )
+        y = _convert_to_dataframe(
+            rng.integers(2, size=1000),
+            sycl_queue=queue,
+            target_df=dataframe,
+            dtype=dtype,
         )
-    except subprocess.CalledProcessError as e:
-        print(e)
-        exit(1)
-    mas = []
-    result = {}
-    for i in process.decode().split("\n"):
-        if i.startswith("SKLEARNEX WARNING"):
-            continue
-        if not i.startswith("SKLEARNEX INFO") and len(mas) != 0:
-            run_parse(mas, result)
-            mas.clear()
-            mas.append(i.strip())
-        else:
-            mas.append(i.strip())
-    del os.environ["SKLEARNEX_VERBOSE"]
-    return result
-
-
-result_log = get_result_log()
-
-
-@pytest.mark.parametrize("configuration", result_log)
-def test_patching(configuration):
-    if "OPT" in result_log[configuration]:
-        return
-    for skip in TO_SKIP:
-        if re.search(skip, configuration) is not None:
-            pytest.skip("SKIPPED", allow_module_level=False)
-    raise ValueError("Test patching failed: " + configuration)
-
-
-def _load_all_models(patched):
-    if patched:
-        patch_sklearn()
-
-    models = {}
-    for patch_infos in get_patch_map().values():
-        maybe_class = getattr(patch_infos[0][0][0], patch_infos[0][0][1], None)
-        if (
-            maybe_class is not None
-            and isclass(maybe_class)
-            and issubclass(maybe_class, BaseEstimator)
+
+        _ = roc_auc_score(X, y)
+    assert all(
+        [
+            "running accelerated version" in i.message
+            or "fallback to original Scikit-learn" in i.message
+            for i in caplog.records
+        ]
+    ), f"sklearnex patching issue in roc_auc_score with log: \n{caplog.text}"
+
+
+@pytest.mark.parametrize("dtype", DTYPES)
+@pytest.mark.parametrize("dataframe, queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("estimator, method", gen_models_info(PATCHED_MODELS))
+def test_standard_estimator_patching(caplog, dataframe, queue, dtype, estimator, method):
+    with caplog.at_level(logging.WARNING, logger="sklearnex"):
+        est = PATCHED_MODELS[estimator]()
+
+        if queue:
+            if dtype == np.float16 and not queue.sycl_device.has_aspect_fp16:
+                pytest.skip("Hardware does not support fp16 SYCL testing")
+            elif dtype == np.float64 and not queue.sycl_device.has_aspect_fp64:
+                pytest.skip("Hardware does not support fp64 SYCL testing")
+            elif queue.sycl_device.is_gpu and estimator in [
+                "KMeans",
+                "ElasticNet",
+                "Lasso",
+                "Ridge",
+            ]:
+                pytest.skip(f"{estimator} does not support GPU queues")
+
+        if estimator == "TSNE" and method == "fit_transform":
+            pytest.skip("TSNE.fit_transform is too slow for common testing")
+        elif (
+            estimator == "Ridge"
+            and method in ["predict", "score"]
+            and sys.platform == "win32"
+            and dtype in [np.uint32, np.uint64]
         ):
-            models[patch_infos[0][0][1]] = maybe_class
+            pytest.skip("Windows segmentation fault for Ridge.predict for unsigned ints")
+        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)
+        est.fit(X, y)
+
+        if method:
+            if method != "score":
+                getattr(est, method)(X)
+            else:
+                est.score(X, y)
+    assert all(
+        [
+            "running accelerated version" in i.message
+            or "fallback to original Scikit-learn" in i.message
+            for i in caplog.records
+        ]
+    ), f"sklearnex patching issue in {estimator}.{method} with log: \n{caplog.text}"
+
+
+@pytest.mark.parametrize("dtype", DTYPES)
+@pytest.mark.parametrize("dataframe, queue", get_dataframes_and_queues())
+@pytest.mark.parametrize("estimator, method", gen_models_info(SPECIAL_INSTANCES))
+def test_special_estimator_patching(caplog, dataframe, queue, dtype, estimator, method):
+    # prepare logging
+
+    with caplog.at_level(logging.WARNING, logger="sklearnex"):
+        est = SPECIAL_INSTANCES[estimator]
+
+        # Its not possible to get the dpnp/dpctl arrays to be in the proper dtype
+        if dtype == np.float16 and queue and not queue.sycl_device.has_aspect_fp16:
+            pytest.skip("Hardware does not support fp16 SYCL testing")
+        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)
+        est.fit(X, y)
+
+        if method and not hasattr(est, method):
+            pytest.skip(f"sklearn available_if prevents testing {estimator}.{method}")
+
+        if method:
+            if method != "score":
+                getattr(est, method)(X)
+            else:
+                est.score(X, y)
+
+    assert all(
+        [
+            "running accelerated version" in i.message
+            or "fallback to original Scikit-learn" in i.message
+            for i in caplog.records
+        ]
+    ), f"sklearnex patching issue in {estimator}.{method} with log: \n{caplog.text}"
+
+
+@pytest.mark.parametrize("estimator", UNPATCHED_MODELS.keys())
+def test_standard_estimator_signatures(estimator):
+    est = PATCHED_MODELS[estimator]()
+    unpatched_est = UNPATCHED_MODELS[estimator]()
+
+    # all public sklearn methods should have signature matches in sklearnex
+
+    unpatched_est_methods = [
+        i
+        for i in dir(unpatched_est)
+        if not i.startswith("_") and not i.endswith("_") and hasattr(unpatched_est, i)
+    ]
+    for method in unpatched_est_methods:
+        est_method = getattr(est, method)
+        unpatched_est_method = getattr(unpatched_est, method)
+        if callable(unpatched_est_method):
+            regex = rf"(?:sklearn|daal4py)\S*{estimator}"  # needed due to differences in module structure
+            patched_sig = re.sub(regex, estimator, str(signature(est_method)))
+            unpatched_sig = re.sub(regex, estimator, str(signature(unpatched_est_method)))
+            assert (
+                patched_sig == unpatched_sig
+            ), f"Signature of {estimator}.{method} does not match sklearn"
+
+
+@pytest.mark.parametrize("estimator", UNPATCHED_MODELS.keys())
+def test_standard_estimator_init_signatures(estimator):
+    # Several estimators have additional parameters that are user-accessible
+    # which are sklearnex-specific. They will fail and are removed from tests.
+    # remove n_jobs due to estimator patching for sklearnex (known deviation)
+    patched_sig = str(signature(PATCHED_MODELS[estimator].__init__))
+    unpatched_sig = str(signature(UNPATCHED_MODELS[estimator].__init__))
+
+    # Sklearnex allows for positional kwargs and n_jobs, when sklearn doesn't
+    for kwarg in ["n_jobs=None", "*"]:
+        patched_sig = patched_sig.replace(", " + kwarg, "")
+        unpatched_sig = unpatched_sig.replace(", " + kwarg, "")
+
+    # Special sklearnex-specific kwargs are removed from signatures here
+    if estimator in [
+        "RandomForestRegressor",
+        "RandomForestClassifier",
+        "ExtraTreesRegressor",
+        "ExtraTreesClassifier",
+    ]:
+        for kwarg in ["min_bin_size=1", "max_bins=256"]:
+            patched_sig = patched_sig.replace(", " + kwarg, "")
+
+    assert (
+        patched_sig == unpatched_sig
+    ), f"Signature of {estimator}.__init__ does not match sklearn"
+
+
+@pytest.mark.parametrize(
+    "function",
+    [
+        i
+        for i in UNPATCHED_FUNCTIONS.keys()
+        if i not in ["train_test_split", "set_config", "config_context"]
+    ],
+)
+def test_patched_function_signatures(function):
+    # certain functions are dropped from the test
+    # as they add functionality to the underlying sklearn function
+    if not sklearn_check_version("1.1") and function == "_assert_all_finite":
+        pytest.skip("Sklearn versioning not added to _assert_all_finite")
+    func = PATCHED_FUNCTIONS[function]
+    unpatched_func = UNPATCHED_FUNCTIONS[function]
+
+    if callable(unpatched_func):
+        assert str(signature(func)) == str(
+            signature(unpatched_func)
+        ), f"Signature of {func} does not match sklearn"
+
+
+def test_patch_map_match():
+    # This rule applies to functions and classes which are out of preview.
+    # Items listed in a matching submodule's __all__ attribute should be
+    # in get_patch_map. There should not be any missing or additional elements.
 
-    if patched:
-        unpatch_sklearn()
+    def list_all_attr(string):
+        try:
+            modules = set(importlib.import_module(string).__all__)
+        except ModuleNotFoundError:
+            modules = set([None])
+        return modules
 
-    return models
+    if _is_preview_enabled():
+        pytest.skip("preview sklearnex has been activated")
+    patched = {**PATCHED_MODELS, **PATCHED_FUNCTIONS}
 
+    sklearnex__all__ = list_all_attr("sklearnex")
+    sklearn__all__ = list_all_attr("sklearn")
 
-PATCHED_MODELS = _load_all_models(patched=True)
-UNPATCHED_MODELS = _load_all_models(patched=False)
+    module_map = {i: i for i in sklearnex__all__.intersection(sklearn__all__)}
+
+    # _assert_all_finite patches an internal sklearn function which isn't
+    # exposed via __all__ in sklearn. It is a special case where this rule
+    # is not applied (e.g. it is grandfathered in).
+    del patched["_assert_all_finite"]
+
+    # remove all scikit-learn-intelex-only estimators
+    for i in patched.copy():
+        if i not in UNPATCHED_MODELS and i not in UNPATCHED_FUNCTIONS:
+            del patched[i]
+
+    for module in module_map:
+        sklearn_module__all__ = list_all_attr("sklearn." + module_map[module])
+        sklearnex_module__all__ = list_all_attr("sklearnex." + module)
+        intersect = sklearnex_module__all__.intersection(sklearn_module__all__)
+
+        for i in intersect:
+            if i:
+                del patched[i]
+            else:
+                del patched[module]
+    assert patched == {}, f"{patched.keys()} were not properly patched"
 
 
 @pytest.mark.parametrize("estimator", UNPATCHED_MODELS.keys())
@@ -125,6 +324,42 @@ def test_is_patched_instance(estimator):
     assert not is_patched_instance(unpatched), f"{unpatched} is an unpatched instance"
 
 
+@pytest.mark.parametrize("estimator", PATCHED_MODELS.keys())
+def test_if_estimator_inherits_sklearn(estimator):
+    est = PATCHED_MODELS[estimator]
+    if estimator in UNPATCHED_MODELS:
+        assert issubclass(
+            est, UNPATCHED_MODELS[estimator]
+        ), f"{estimator} does not inherit from the patched sklearn estimator"
+    else:
+        assert issubclass(est, BaseEstimator)
+
+
+@pytest.mark.parametrize("estimator", UNPATCHED_MODELS.keys())
+def test_docstring_patching_match(estimator):
+    patched = PATCHED_MODELS[estimator]
+    unpatched = UNPATCHED_MODELS[estimator]
+    patched_docstrings = {
+        i: getattr(patched, i).__doc__
+        for i in dir(patched)
+        if not i.startswith("_") and not i.endswith("_") and hasattr(patched, i)
+    }
+    unpatched_docstrings = {
+        i: getattr(unpatched, i).__doc__
+        for i in dir(unpatched)
+        if not i.startswith("_") and not i.endswith("_") and hasattr(unpatched, i)
+    }
+
+    # check class docstring match if a docstring is available
+
+    assert (patched.__doc__ is None) == (unpatched.__doc__ is None)
+
+    # check class attribute docstrings
+
+    for i in unpatched_docstrings:
+        assert (patched_docstrings[i] is None) == (unpatched_docstrings[i] is None)
+
+
 @pytest.mark.parametrize("member", ["_onedal_cpu_supported", "_onedal_gpu_supported"])
 @pytest.mark.parametrize(
     "name",

sklearnex/utils/__init__.py

@@ -14,6 +14,7 @@
 # limitations under the License.
 # ===============================================================================
 
+from ._namespace import get_namespace
 from .validation import _assert_all_finite
 
-__all__ = ["_assert_all_finite"]
+__all__ = ["get_namespace", "_assert_all_finite"]

sklearnex/utils/_namespace.py

@@ -0,0 +1,97 @@
+# ==============================================================================
+# 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 numpy as np
+
+from daal4py.sklearn._utils import sklearn_check_version
+
+from .._device_offload import dpnp_available
+
+if sklearn_check_version("1.2"):
+    from sklearn.utils._array_api import get_namespace as sklearn_get_namespace
+
+if dpnp_available:
+    import dpnp
+
+
+def get_namespace(*arrays):
+    """Get namespace of arrays.
+
+    Introspect `arrays` arguments and return their common Array API
+    compatible namespace object, if any. NumPy 1.22 and later can
+    construct such containers using the `numpy.array_api` namespace
+    for instance.
+
+    This function will return the namespace of SYCL-related arrays
+    which define the __sycl_usm_array_interface__ attribute
+    regardless of array_api support, the configuration of
+    array_api_dispatch, or scikit-learn version.
+
+    See: https://numpy.org/neps/nep-0047-array-api-standard.html
+
+    If `arrays` are regular numpy arrays, an instance of the
+    `_NumPyApiWrapper` compatibility wrapper is returned instead.
+
+    Namespace support is not enabled by default. To enabled it
+    call:
+
+      sklearn.set_config(array_api_dispatch=True)
+
+    or:
+
+      with sklearn.config_context(array_api_dispatch=True):
+          # your code here
+
+    Otherwise an instance of the `_NumPyApiWrapper`
+    compatibility wrapper is always returned irrespective of
+    the fact that arrays implement the `__array_namespace__`
+    protocol or not.
+
+    Parameters
+    ----------
+    *arrays : array objects
+        Array objects.
+
+    Returns
+    -------
+    namespace : module
+        Namespace shared by array objects.
+
+    is_array_api : bool
+        True of the arrays are containers that implement the Array API spec.
+    """
+
+    # sycl support designed to work regardless of array_api_dispatch sklearn global value
+    sycl_type = {type(x): x for x in arrays if hasattr(x, "__sycl_usm_array_interface__")}
+
+    if len(sycl_type) > 1:
+        raise ValueError(f"Multiple SYCL types for array inputs: {sycl_type}")
+
+    if sycl_type:
+
+        (X,) = sycl_type.values()
+
+        if hasattr(X, "__array_namespace__"):
+            return X.__array_namespace__(), True
+        elif dpnp_available and isinstance(X, dpnp.ndarray):
+            return dpnp, False
+        else:
+            raise ValueError(f"SYCL type not recognized: {sycl_type}")
+
+    elif sklearn_check_version("1.2"):
+        return sklearn_get_namespace(*arrays)
+    else:
+        return np, True

sklearnex/preview/decomposition/__init__.py

@@ -1,19 +0,0 @@
-# ===============================================================================
-# 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.
-# ===============================================================================
-
-from .pca import PCA
-
-__all__ = ["PCA"]