dataeval.extractors¶
Feature extractors for drift detection and quality metrics.
This module provides feature extraction implementations that convert various data types into arrays suitable for drift detection and quality analysis.
The primary feature extractors are Embeddings and
Metadata, which implement the FeatureExtractor
protocol and can be used directly with drift detectors.
This submodule provides additional specialized extractors:
UncertaintyFeatureExtractor: Computes model prediction uncertainty (entropy)
Classes¶
Feature extractor that converts data to model uncertainty scores. |