Enabling array API dispatch causes some estimators without array API support to reject valid NumPy inputs #32837
Description
Describe the bug
This might be related to #32836 (or not).
I think we need a common test to check that enabling array API support does not affect estimators (without array API support) when called on regular NumPy (or other valid non-array inputs).
This bug might be a regression.
Steps/Code to Reproduce
# %%
import os
os.environ["SCIPY_ARRAY_API"] = "1"
# %%
import sklearn
from sklearn.preprocessing import KBinsDiscretizer
from sklearn.datasets import load_iris
X_iris, y_iris = load_iris(return_X_y=True)
with sklearn.config_context(array_api_dispatch=True):
KBinsDiscretizer(
n_bins=3,
encode="onehot-dense",
strategy="uniform",
quantile_method="averaged_inverted_cdf",
).fit_transform(X_iris)Expected Results
No error.
Actual Results
Traceback (most recent call last):
Cell In[5], line 14
).fit_transform(X_iris)
File ~/code/scikit-learn/sklearn/utils/_set_output.py:316 in wrapped
data_to_wrap = f(self, X, *args, **kwargs)
File ~/code/scikit-learn/sklearn/base.py:907 in fit_transform
return self.fit(X, **fit_params).transform(X)
File ~/code/scikit-learn/sklearn/base.py:1336 in wrapper
return fit_method(estimator, *args, **kwargs)
File ~/code/scikit-learn/sklearn/preprocessing/_discretization.py:414 in fit
self._encoder.fit(np.zeros((1, len(self.n_bins_))))
File ~/code/scikit-learn/sklearn/base.py:1336 in wrapper
return fit_method(estimator, *args, **kwargs)
File ~/code/scikit-learn/sklearn/preprocessing/_encoders.py:999 in fit
self._fit(
File ~/code/scikit-learn/sklearn/preprocessing/_encoders.py:86 in _fit
X_list, n_samples, n_features = self._check_X(
File ~/code/scikit-learn/sklearn/preprocessing/_encoders.py:68 in _check_X
Xi = _safe_indexing(X, indices=i, axis=1)
File ~/code/scikit-learn/sklearn/utils/_indexing.py:357 in _safe_indexing
return _array_indexing(X, indices, indices_dtype, axis=axis)
File ~/code/scikit-learn/sklearn/utils/_indexing.py:39 in _array_indexing
key = move_to(key, xp=xp, device=device_)
File ~/code/scikit-learn/sklearn/utils/_array_api.py:508 in move_to
xp_array, _, device_array = get_namespace_and_device(array)
File ~/code/scikit-learn/sklearn/utils/_array_api.py:453 in get_namespace_and_device
xp, is_array_api = get_namespace(*array_list, **skip_remove_kwargs)
File ~/code/scikit-learn/sklearn/utils/_array_api.py:404 in get_namespace
namespace, is_array_api_compliant = array_api_compat.get_namespace(*arrays), True
File ~/code/scikit-learn/sklearn/externals/array_api_compat/common/_helpers.py:646 in array_namespace
raise TypeError("Unrecognized array input")
TypeError: Unrecognized array inputVersions
System:
python: 3.13.7 | packaged by conda-forge | (main, Sep 3 2025, 14:24:46) [Clang 19.1.7 ]
executable: /Users/ogrisel/miniforge3/envs/dev/bin/python
machine: macOS-15.6.1-arm64-arm-64bit-Mach-O
Python dependencies:
sklearn: 1.9.dev0
pip: 25.2
setuptools: 80.9.0
numpy: 2.3.3
scipy: 1.16.3
Cython: 3.1.4
pandas: 3.0.0.dev0+2566.g2bb3fef887
matplotlib: 3.10.6
joblib: 1.5.2
threadpoolctl: 3.6.0
Built with OpenMP: True
threadpoolctl info:
user_api: blas
internal_api: openblas
num_threads: 10
prefix: libopenblas
filepath: /Users/ogrisel/miniforge3/envs/dev/lib/libopenblas.0.dylib
version: 0.3.30
threading_layer: openmp
architecture: VORTEX
user_api: openmp
internal_api: openmp
num_threads: 10
prefix: libomp
filepath: /Users/ogrisel/miniforge3/envs/dev/lib/libomp.dylib
version: None