[BUG] `ColumnTransformer.set_params` rejects nested transformer parameters

[apiqwe]

Describe the bug

cuml.compose.ColumnTransformer.set_params() rejects the standard nested estimator parameter syntax <transformer_name>__<parameter>.

Calling set_params(s__with_mean=False) for a named StandardScaler transformer raises:

ValueError: Invalid parameter 's__with_mean' for `ColumnTransformer`

The equivalent sklearn.compose.ColumnTransformer call accepts the nested parameter, applies it to the named transformer, and completes successfully.

Steps/Code to reproduce bug

cuML reproducer:

import numpy as np
from cuml.compose import ColumnTransformer
from cuml.preprocessing import StandardScaler

a = np.array([
    [1.0, 2.0],
    [3.0, 4.0],
    [5.0, 6.0],
])

print(
    ColumnTransformer([
        ("s", StandardScaler(), [0, 1]),
    ])
    .set_params(s__with_mean=False)
    .fit_transform(a)
)

Output:

Traceback (most recent call last):
  File "/workspace/apibughub/cuml/ColumnTransformer/error_bug1/error_bug_cuml.py", line 7, in <module>
    print(ColumnTransformer([("s", StandardScaler(), [0, 1])]).set_params(s__with_mean=False).fit_transform(a))
          ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/envs/rapids-26.08/lib/python3.14/site-packages/cuml/_thirdparty/sklearn/preprocessing/_column_transformer.py", line 637, in set_params
    self._set_params('_transformers', **kwargs)
    ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/envs/rapids-26.08/lib/python3.14/site-packages/cuml/_thirdparty/sklearn/utils/skl_dependencies.py", line 105, in _set_params
    super().set_params(**params)
    ~~~~~~~~~~~~~~~~~~^^^^^^^^^^
  File "/opt/conda/envs/rapids-26.08/lib/python3.14/site-packages/cuml/internals/base.py", line 226, in set_params
    raise ValueError(
        f"Invalid parameter {key!r} for `{type(self).__name__}`"
    )
ValueError: Invalid parameter 's__with_mean' for `ColumnTransformer`

For comparison, the equivalent scikit-learn code:

import numpy as np
from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import StandardScaler

a = np.array([
    [1.0, 2.0],
    [3.0, 4.0],
    [5.0, 6.0],
])

print(
    ColumnTransformer([
        ("s", StandardScaler(), [0, 1]),
    ])
    .set_params(s__with_mean=False)
    .fit_transform(a)
)

Output:

[[0.61237244 1.22474487]
 [1.83711731 2.44948974]
 [3.06186218 3.67423461]]

Expected behavior

cuml.compose.ColumnTransformer.set_params() should accept nested transformer parameters using the standard <transformer_name>__<parameter> syntax.

In this example, s__with_mean=False should set with_mean=False on the transformer named s, after which fit_transform(a) should complete and return the transformed matrix.

Environment details (please complete the following information):

• Environment location: Docker
• Linux Distro/Architecture: Ubuntu 24.04 / x86_64
• GPU Model/Driver: NVIDIA GeForce RTX 4090 / 595.71.05
• CUDA: 13.2
• Method of cuDF & cuML install: conda

conda list:

conda list
# packages in environment at /opt/conda/envs/rapids-26.08:
#
# Name                              Version          Build                                         Channel
# Name              Version       Build                                      Channel
python              3.14.6        h242f9ac_102_cp314                         conda-forge
numpy               2.4.6         py314h2b28147_0                            conda-forge
scipy               1.16.3        py314hf07bd8e_2                            conda-forge
scikit-learn        1.9.0         np2py314hf09ca88_0                         conda-forge
rapids              26.08.00      cuda13_260806_c2656556                     rapidsai
cuml                26.08.00      cuda13_cp311_abi3_260805_265b9da6          rapidsai
libcuml             26.08.00      cuda13_260805_265b9da6                     rapidsai
cudf                26.08.00      cuda13_cp311_abi3_260805_ff5b362d          rapidsai
libraft             26.08.00      cuda13_260805_ebf92684                     rapidsai
libraft-headers     26.08.00      cuda13_260805_ebf92684                     rapidsai
pylibraft           26.08.00      cuda13_cp311_abi3_260805_ebf92684          rapidsai
cuvs                26.08.01      cuda13_cp311_abi3_260806_25b1be43          rapidsai
libcuvs             26.08.01      cuda13_260806_25b1be43                     rapidsai
cupy                14.1.1        py314hdea9c46_0                            conda-forge
cupy-core           14.1.1        py314hcd3b49b_0                            conda-forge
numba               0.64.0        py314h8169c2f_0                            conda-forge
numba-cuda          0.30.4        py314h42812f9_0                            conda-forge
rmm                 26.08.00      cuda13_cp311_abi3_260805_42d059f1          rapidsai
librmm              26.08.00      cuda13_260805_42d059f1                     rapidsai
cuda-version        13.3           hcbadf70_3                                 conda-forge
cuda-bindings       13.3.1        py314h42812f9_1                            conda-forge
cuda-cudart         13.3.29       hecca717_0                                 conda-forge
cuda-nvrtc          13.3.33       hecca717_0                                 conda-forge
libcublas           13.6.0.2      h676940d_0                                 conda-forge
libcusolver         12.2.6.9      h676940d_0                                 conda-forge
libcusparse         12.8.2.51     hecca717_0                                 conda-forge
libcurand           10.4.3.29     h676940d_0                                 conda-forge

Additional context

The failure occurs during set_params(), before fitting or transforming any data. The named transformer is present in the ColumnTransformer, and with_mean is a valid parameter of cuml.preprocessing.StandardScaler.

The equivalent scikit-learn implementation supports this nested parameter syntax and returns:

[[0.61237244 1.22474487]
 [1.83711731 2.44948974]
 [3.06186218 3.67423461]]

Nested parameter handling is important for compatibility with scikit-learn workflows such as parameter tuning, pipelines, and cloning.