Implementation:Scikit learn Scikit learn MetadataRequests

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Machine Learning, Metadata Routing
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete utility module for metadata routing between estimators provided by scikit-learn.

Description

The _metadata_requests module implements the metadata routing framework that allows scikit-learn meta-estimators to pass metadata (such as sample_weight) to sub-estimators. It provides MetadataRequest for consumers, MetadataRouter for routers (meta-estimators), MethodMapping for caller-callee method relationships, and process_routing for resolving metadata at call time.

Usage

Use these utilities when building custom meta-estimators that need to route metadata parameters (e.g., sample_weight, groups) to their sub-estimators, or when using the set_{method}_request API to control metadata propagation.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/utils/_metadata_requests.py

Signature

class MethodMetadataRequest:
    ...

class MetadataRequest:
    ...

class MethodMapping:
    ...

class MetadataRouter:
    ...

class RequestMethod:
    ...

class _MetadataRequester:
    ...

def process_routing(_obj, _method, /, **kwargs):
    ...

Import

from sklearn.utils.metadata_routing import MetadataRouter, MethodMapping, process_routing

I/O Contract

Inputs

Name	Type	Required	Description
_obj	estimator	Yes	The router object that is processing the routing
_method	str	Yes	The method name (e.g., "fit", "score") on the router
kwargs	dict	No	The metadata key-value pairs to be routed

Outputs

Name	Type	Description
routing_info	Bunch	Dictionary-like object with routed metadata per sub-estimator and method

Usage Examples

Basic Usage

from sklearn.linear_model import LogisticRegression

# Request that sample_weight be passed to fit
lr = LogisticRegression()
lr.set_fit_request(sample_weight=True)

# In a meta-estimator, use process_routing to route metadata
from sklearn.utils.metadata_routing import MetadataRouter, MethodMapping
router = MetadataRouter(owner="MyMetaEstimator")
router.add(estimator=lr, method_mapping=MethodMapping().add(caller="fit", callee="fit"))

Related Pages

• Principle:Scikit_learn_Scikit_learn_Pipeline_Execution