Forward-merge release/26.04 into main

python/cuml/cuml/internals/outputs.py

@@ -357,9 +357,11 @@ def reflect(
         provide ``None`` to disable this inference entirely; in this case the
         output type is expected to be specified manually either internal or
         external to the method.
-    reset : bool, default=False
-        Set to True for methods like ``fit`` that reset the reflected type on
-        an estimator.
+    reset : bool or "type", default=False
+        If True, both the features and reflected type are reset on the estimator.
+        If ``"type"``, only the reflected type is reset on the estimator.
+        Defaults to False, to not reset anything. Most estimators should set
+        ``reset=True`` on any fit-like methods.
     """
     # Local to avoid circular imports
     import cuml.accel
@@ -391,9 +393,12 @@ def reflect(
     if array is not None:
         array = _get_param(sig, array)
 
-    if reset and (model is None or array is None):
+    if reset not in (True, False, "type"):
+        raise ValueError(f"reset={reset!r} is not supported")
+
+    if (reset is not False) and (model is None or array is None):
         raise ValueError(
-            "`reset=True` is not valid with `array=None` or `model=None`"
+            f"`reset={reset}` is not valid with `array=None` or `model=None`"
         )
 
     @functools.wraps(func)
@@ -411,8 +416,9 @@ def inner(*args, **kwargs):
             array_arg = np.asarray(array_arg)
 
         with enter_internal_context() as was_external:
-            if reset:
+            if reset is not False:
                 model_arg._set_output_type(array_arg)
+            if reset is True:
                 check_features(model_arg, array_arg, reset=True)
 
             res = func(*args, **kwargs)

python/cuml/cuml/internals/validation.py

@@ -71,7 +71,6 @@ def _get_n_features(X):
                 return len(row)
             except Exception:
                 pass
-        return 1
 
     if hasattr(X, "shape"):
         shape = X.shape
@@ -82,9 +81,38 @@ def _get_n_features(X):
     else:
         shape = np.asarray(X).shape
 
-    # TODO: Can remove the fallback to 1 when we finish dropping support
-    # for 1D X inputs
-    return shape[1] if len(shape) >= 2 else 1
+    ndim = len(shape)
+
+    if ndim != 2:
+        import cuml.accel
+
+        if isinstance(X, (cudf.Series, pd.Series)):
+            msg = (
+                f"Expected a 2-dimensional container but got {type(X).__name__} "
+                "instead. Pass a DataFrame containing a single row (i.e. "
+                "single sample) or a single column (i.e. single feature) "
+                "instead."
+            )
+        else:
+            kind = "scalar" if ndim == 0 else f"{ndim}D"
+            msg = (
+                f"Expected 2D array, got {kind} array instead. Reshape your data "
+                "using array.reshape(-1, 1) if your data has a single feature, "
+                "or array.reshape(1, -1) if it contains a single sample."
+            )
+
+        if cuml.accel.enabled() or ndim > 2:
+            raise ValueError(msg)
+        else:
+            warnings.warn(
+                "Support for passing non-2-dimensional X was deprecated in 26.04 "
+                "and will be removed in version 26.06 of cuML. In version 26.06 this will error "
+                f"with the following message:\n\n{msg}",
+                FutureWarning,
+            )
+            # Fallback to 1 feature until the deprecation is completed
+            return 1
+    return shape[1]
 
 
 def _warn_or_error(exc_cls, msg):

python/cuml/cuml/preprocessing/TargetEncoder.py

@@ -174,8 +174,8 @@ class TargetEncoder(Base, InteropMixin):
     >>> test = DataFrame({'category': ['a', 'c', 'b', 'a']})
 
     >>> encoder = TargetEncoder(output_type='numpy')
-    >>> train_encoded = encoder.fit_transform(train.category, train.label)
-    >>> test_encoded = encoder.transform(test.category)
+    >>> train_encoded = encoder.fit_transform(train[["category"]], train.label)
+    >>> test_encoded = encoder.transform(test[["category"]])
     >>> print(train_encoded)
     [1. 1. 0. 1.]
     >>> print(test_encoded)

python/cuml/cuml/preprocessing/label.py

@@ -157,7 +157,7 @@ def __init__(
         self.sparse_output = sparse_output
         self.classes_ = None
 
-    @cuml.internals.reflect(reset=True)
+    @cuml.internals.reflect(reset="type")
     def fit(self, y) -> "LabelBinarizer":
         """
         Fit label binarizer

python/cuml/tests/test_coordinate_descent.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2019-2025, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION.
 # SPDX-License-Identifier: Apache-2.0
 #
 
@@ -274,7 +274,7 @@ def test_lasso_predict_convert_dtype(train_dtype, test_dtype):
 
 @pytest.mark.parametrize("cls", [cuml.ElasticNet, cuml.Lasso])
 def test_set_params(cls):
-    x = np.linspace(0, 1, 50)
+    x = np.linspace(0, 1, 50)[:, None]
     y = 2 * x
 
     model = cls(alpha=0.01)

python/cuml/tests/test_dbscan.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2019-2025, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION.
 # SPDX-License-Identifier: Apache-2.0
 #
 
@@ -497,7 +497,7 @@ def test_dbscan_no_calc_core_point_indices():
 
 
 def test_dbscan_on_empty_array():
-    X = np.array([])
+    X = np.array([[]])
     cuml_dbscan = cuDBSCAN()
 
     with pytest.raises(ValueError):

python/cuml/tests/test_label_binarizer.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2020-2025, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2020-2026, NVIDIA CORPORATION.
 # SPDX-License-Identifier: Apache-2.0
 
 import cupy as cp
@@ -11,6 +11,13 @@
 from cuml.testing.utils import array_equal
 
 
+def test_label_binarizer_no_features():
+    """Ensure the features infra is never applied to LabelBinarizer"""
+    y = cp.asarray([1, 2, 1, 2, 1, 0])
+    model = LabelBinarizer().fit(y)
+    assert not hasattr(model, "n_features_in_")
+
+
 @pytest.mark.parametrize(
     "labels",
     [

python/cuml/tests/test_label_encoder.py

@@ -20,6 +20,13 @@ def _df_to_similarity_mat(df):
     return np.pad(arr, [(arr.shape[1] - 1, 0), (0, 0)], "edge")
 
 
+def test_label_encoder_no_features():
+    """Ensure the features infra is never applied to LabelEncoder"""
+    y = cp.asarray([1, 2, 1, 2, 1, 0])
+    model = LabelEncoder().fit(y)
+    assert not hasattr(model, "n_features_in_")
+
+
 @pytest.mark.parametrize("length", [10, 1000])
 @pytest.mark.parametrize("cardinality", [5, 10, 50])
 def test_labelencoder_fit_transform(length, cardinality):