Optimize import times a bit

python/cuml/cuml/accel/_wrappers/sklearn/svm.py

@@ -15,7 +15,7 @@
 import functools
 
 import numpy as np
-import sklearn.svm
+from sklearn.svm import SVC as _SVC
 from sklearn.utils.metaestimators import available_if
 
 import cuml.svm
@@ -76,12 +76,12 @@ def _gpu_decision_function(self, X):
     # ProxyBase lacks a builtin mechanism to do that, since this is the only
     # use case so far we manually define them for now.
     @available_if(_has_probability)
-    @functools.wraps(sklearn.svm.SVC.predict_proba)
+    @functools.wraps(_SVC.predict_proba)
     def predict_proba(self, X):
         return self._call_method("predict_proba", X)
 
     @available_if(_has_probability)
-    @functools.wraps(sklearn.svm.SVC.predict_log_proba)
+    @functools.wraps(_SVC.predict_log_proba)
     def predict_log_proba(self, X):
         return self._call_method("predict_log_proba", X)
 

python/cuml/cuml/explainer/kernel_shap.pyx

@@ -20,7 +20,6 @@ from random import randint
 
 import cupy as cp
 import numpy as np
-from sklearn.linear_model import LassoLarsIC, lars_path
 
 from cuml.explainer.base import SHAPBase
 from cuml.explainer.common import get_cai_ptr, model_func_call
@@ -565,6 +564,7 @@ def _l1_regularization(X,
     """
     Function calls LASSO or LARS if l1 regularization is needed.
     """
+    from sklearn.linear_model import LassoLarsIC, lars_path
 
     # create augmented dataset for feature selection
     s = cp.sum(X, axis=1)

python/cuml/cuml/explainer/tree_shap.pyx

@@ -13,24 +13,20 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 #
+import re
+
+import numpy as np
+import treelite
 
+import cuml
 from cuml.common import input_to_cuml_array
-from cuml.ensemble import RandomForestClassifier as curfc
-from cuml.ensemble import RandomForestRegressor as curfr
 from cuml.internals.array import CumlArray
 from cuml.internals.input_utils import determine_array_type
-
-from cuml.internals.treelite cimport *
 from cuml.internals.treelite import safe_treelite_call
 
 from libc.stdint cimport uintptr_t
 
-import re
-
-import numpy as np
-import treelite
-from sklearn.ensemble import RandomForestClassifier as sklrfc
-from sklearn.ensemble import RandomForestRegressor as sklrfr
+from cuml.internals.treelite cimport *
 
 
 cdef extern from "cuml/explainer/tree_shap.hpp" namespace "ML::Explainer" nogil:
@@ -178,15 +174,20 @@ cdef class TreeExplainer:
                 model = model.booster_
             tl_model = treelite.frontend.from_lightgbm(model)
         # cuML RF model object
-        elif isinstance(model, (curfr, curfc)):
+        elif isinstance(model, (cuml.RandomForestClassifier, cuml.RandomForestRegressor)):
             tl_model = model.convert_to_treelite_model()
         # scikit-learn RF model object
-        elif isinstance(model, (sklrfr, sklrfc)):
-            tl_model = treelite.sklearn.import_model(model)
         elif isinstance(model, treelite.Model):
             tl_model = model
         else:
-            raise ValueError(f"Unrecognized model object type: {type(model)}")
+            from sklearn.ensemble import (
+                RandomForestClassifier,
+                RandomForestRegressor,
+            )
+            if isinstance(model, (RandomForestClassifier, RandomForestRegressor)):
+                tl_model = treelite.sklearn.import_model(model)
+            else:
+                raise ValueError(f"Unrecognized model object type: {type(model)}")
 
         # Get num_class
         self.num_class = tl_model.get_header_accessor().get_field("num_class").copy()

python/cuml/cuml/internals/base.py

@@ -18,7 +18,6 @@
 import os
 
 import pylibraft.common.handle
-from sklearn.utils import estimator_html_repr
 
 import cuml
 import cuml.common
@@ -458,10 +457,11 @@ def _more_tags(self):
 
     def _repr_mimebundle_(self, **kwargs):
         """Prepare representations used by jupyter kernels to display estimator"""
-        if estimator_html_repr is not None:
-            output = {"text/plain": repr(self)}
-            output["text/html"] = estimator_html_repr(self)
-            return output
+        from sklearn.utils import estimator_html_repr
+
+        output = {"text/plain": repr(self)}
+        output["text/html"] = estimator_html_repr(self)
+        return output
 
     def set_nvtx_annotations(self):
         for func_name in [

python/cuml/cuml/model_selection/__init__.py

@@ -13,19 +13,20 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 #
-
-from sklearn.model_selection import GridSearchCV
-
-from cuml.model_selection._split import StratifiedKFold, train_test_split
-
-GridSearchCV.__doc__ = (
-    """
-This code is developed and maintained by scikit-learn and imported
+"""This code is developed and maintained by scikit-learn and imported
 by cuML to maintain the familiar sklearn namespace structure.
 cuML includes tests to ensure full compatibility of these wrappers
 with CUDA-based data and cuML estimators, but all of the underlying code
-is due to the scikit-learn developers.\n\n"""
-    + GridSearchCV.__doc__
-)
+is due to the scikit-learn developers."""
+
+from cuml.model_selection._split import StratifiedKFold, train_test_split
 
 __all__ = ["train_test_split", "GridSearchCV", "StratifiedKFold"]
+
+
+def __getattr__(name):
+    if name == "GridSearchCV":
+        from sklearn.model_selection import GridSearchCV
+
+        return GridSearchCV
+    raise AttributeError(f"module {__name__} has no attribute {name}")

python/cuml/cuml/pipeline/__init__.py

@@ -13,17 +13,22 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 #
-
-from sklearn.pipeline import Pipeline, make_pipeline
-
-disclaimer = """
-This code is developed and maintained by scikit-learn and imported
+"""This code is developed and maintained by scikit-learn and imported
 by cuML to maintain the familiar sklearn namespace structure.
 cuML includes tests to ensure full compatibility of these wrappers
 with CUDA-based data and cuML estimators, but all of the underlying code
-is due to the scikit-learn developers.\n\n"""
-
-Pipeline.__doc__ = disclaimer + Pipeline.__doc__
-make_pipeline.__doc__ = disclaimer + make_pipeline.__doc__
+is due to the scikit-learn developers."""
 
 __all__ = ["Pipeline", "make_pipeline"]
+
+
+def __getattr__(name):
+    if name == "Pipeline":
+        from sklearn.pipeline import Pipeline
+
+        return Pipeline
+    elif name == "make_pipeline":
+        from sklearn.pipeline import make_pipeline
+
+        return make_pipeline
+    raise AttributeError(f"module {__name__} has no attribute {name}")

python/cuml/cuml/random_projection/random_projection.py

@@ -15,7 +15,6 @@
 import cupyx.scipy.sparse as cp_sp
 import numpy as np
 import scipy.sparse as sp
-import sklearn.random_projection
 
 import cuml
 from cuml.common.array_descriptor import CumlArrayDescriptor
@@ -51,6 +50,8 @@ def johnson_lindenstrauss_min_dim(n_samples, eps=0.1):
         The minimal number of components to guarantee with good probability
         an eps-embedding with n_samples.
     """
+    import sklearn.random_projection
+
     return sklearn.random_projection.johnson_lindenstrauss_min_dim(
         n_samples, eps=eps
     )

python/cuml/cuml_accel_tests/upstream/scikit-learn/xfail-list.yaml

@@ -404,7 +404,6 @@
   - "sklearn.tests.test_common::test_estimators[Pipeline(steps=[('scaler',StandardScaler()),('final_estimator',LogisticRegression())])-check_estimators_overwrite_params]"
   - "sklearn.tests.test_common::test_estimators[Pipeline(steps=[('scaler',StandardScaler()),('final_estimator',Ridge())])-check_dont_overwrite_parameters]"
   - "sklearn.tests.test_common::test_estimators[Pipeline(steps=[('scaler',StandardScaler()),('final_estimator',Ridge())])-check_estimators_overwrite_params]"
-  - "sklearn.tests.test_docstrings::test_docstring[GridSearchCV-None]"
   - "sklearn.tests.test_multioutput::test_multi_output_classes_[estimator0]"
   - "sklearn.tests.test_multioutput::test_multi_output_classification_sample_weights"
   - "sklearn.tests.test_multioutput::test_multi_target_sparse_regression[bsr_array]"