Troubleshooting¶

Common errors when working with Sklearn-Wrap, their causes, and how to fix them.

Wrapper Definition Errors¶

Missing `_estimator_name`¶

ValueError: Class should define a static '_estimator_name'.

Cause: The wrapper class does not define _estimator_name as a string class attribute.

Fix: Add a string attribute to your wrapper class:

class MyWrapper(BaseClassWrapper, RegressorMixin):
    _estimator_name = "regressor"
    _estimator_base_class = object

Missing `_estimator_base_class`¶

ValueError: Class should define a static '_estimator_base_class'.

Cause: The wrapper class does not define _estimator_base_class.

Fix: Add a class attribute. Use object to accept any class, or a specific base class for stricter validation:

_estimator_base_class = object  # accepts any class

Constructor Errors¶

Missing Required Parameter¶

TypeError: MyWrapper.__init__() missing required keyword argument: 'regressor'

Cause: The wrapped class was not passed when creating the wrapper, and no _estimator_default_class is defined.

Fix: Pass the class to wrap using the keyword matching _estimator_name:

wrapper = MyWrapper(regressor=MyEstimator, alpha=1.0)

If the wrapper should have a sensible default, define _estimator_default_class:

class MyWrapper(BaseClassWrapper, RegressorMixin):
    _estimator_name = "regressor"
    _estimator_base_class = object
    _estimator_default_class = MyDefaultEstimator

Estimator Is Not a Class¶

TypeError: regressor parameter for estimator MyWrapper is not a class. It is <MyEstimator object>.

Cause: An instance was passed instead of a class. The wrapper expects a class (not an instantiated object) because it handles instantiation itself during fit().

Fix: Pass the class, not an instance:

# Wrong
wrapper = MyWrapper(regressor=MyEstimator())

# Correct
wrapper = MyWrapper(regressor=MyEstimator)

Wrong Base Class¶

ValueError: Invalid regressor class WrongClass for estimator MyWrapper.
Valid estimator class should be derived from ExpectedBase.

Cause: The passed class does not inherit from _estimator_base_class.

Fix: Either pass a class that inherits from the expected base, or widen the constraint:

_estimator_base_class = object  # accepts any class

Parameter Errors¶

Reserved Delimiter in Parameter Name¶

ValueError: Parameter name my__param cannot contain '__' (double underscore).
This delimiter is reserved for nested parameter syntax.

Cause: The wrapped class has a constructor parameter containing __, which conflicts with Scikit-Learn's nested parameter syntax.

Fix: Rename the parameter in the wrapped class, or create an adapter class that translates the parameter name:

class Adapter:
    def __init__(self, my_param=1.0):
        self._original = OriginalClass(my__param=my_param)

Invalid Parameter Name¶

ValueError: my_param is not a valid parameter for class MyEstimator.

Cause: A keyword argument was passed to the wrapper that does not match any constructor parameter of the wrapped class.

Fix: Check the wrapped class's __init__ signature for valid parameter names.

Cannot Change Estimator Class¶

ValueError: Cannot change estimator class via set_params.
The 'regressor' parameter cannot be set. Redeclare the estimator...

Cause: set_params() was called trying to change the wrapped class itself.

Fix: Create a new wrapper instance with the different class:

# Wrong
wrapper.set_params(regressor=DifferentClass)

# Correct
new_wrapper = MyWrapper(regressor=DifferentClass, alpha=1.0)

Fit-Time Errors¶

Required Parameter Still Unset¶

ValueError: Class MyEstimator requires parameter alpha.

Cause: The wrapped class has a required constructor parameter (no default value) that was not provided when creating the wrapper.

Fix: Pass all required parameters:

wrapper = MyWrapper(regressor=MyEstimator, alpha=1.0)

Unfitted Estimator¶

sklearn.exceptions.NotFittedError: This MyWrapper instance is not fitted yet.

Cause: predict() or transform() was called before fit().

Fix: Call fit() first:

wrapper.fit(X_train, y_train)
predictions = wrapper.predict(X_test)

Nested Parameter Errors¶

Invalid Nested Parameter¶

ValueError: Invalid parameter inner_alpha for estimator MyWrapper.
Valid parameters are: ['regressor', 'alpha', 'beta']

Cause: The base key of a nested parameter (before the first __) does not match any wrapper parameter.

Fix: Check wrapper.get_params() for valid parameter names. Nested parameters use __ to separate levels:

# Set a nested parameter
wrapper.set_params(inner__alpha=2.0)

Nested Object Missing `set_params`¶

AttributeError: Cannot set nested parameters on inner.
Object of type int does not have a set_params method.

Cause: Double-underscore syntax was used to reach into a parameter that is not itself an estimator or wrapper.

Fix: Only use __ syntax for parameters that are BaseClassWrapper instances or Scikit-Learn estimators with set_params().

YAML Configuration Errors¶

Untrusted Module¶

UntrustedModuleError: Module 'xgboost' is not in the trusted modules list.

Cause: The YAML config references a class from a module not in the trusted allowlist.

Fix: Add the module to trusted modules:

from sklearn_wrap.config import set_config

set_config(trusted_modules=frozenset({"sklearn", "sklearn_wrap", "xgboost"}))