Feature: Add Support for `PatchExtractor` in Sklearn Serializer

[Gnpd]

Feature: Add Support for PatchExtractor in Sklearn Serializer

Summary

Currently, the OpenModels sklearn serializer does not support the PatchExtractor transformer from scikit-learn. Attempting to serialize or deserialize a PatchExtractor instance results in a ValueError: not enough values to unpack (expected 3, got 2). This is likely due to the way the patch_size or related attributes are handled internally.

Motivation

• Completeness: PatchExtractor is a useful transformer for image data and should be supported like other scikit-learn estimators.
• User Experience: Users expect all standard scikit-learn transformers to be serializable/deserializable without errors.
• Consistency: The serializer should handle special cases or internal attribute structures that differ from typical estimators.

Error Details

• Error:
  ValueError: not enough values to unpack (expected 3, got 2)
• Context:
  This error occurs when serializing or deserializing a PatchExtractor instance, likely due to the structure of the patch_size or related attributes.

References

• scikit-learn PatchExtractor documentation
• PatchExtractor source code

Suggested Tasks

• Investigate the internal structure of PatchExtractor, especially how patch_size and related attributes are stored and restored.
• Update the serializer to correctly handle the serialization and deserialization of PatchExtractor parameters and fitted attributes.
• Add tests to ensure that PatchExtractor can be round-tripped (serialized and deserialized) without errors.
• Remove "PatchExtractor" from the NOT_SUPPORTED_ESTIMATORS list in sklearn_serializer.py once support is complete.

Acceptance Criteria

• PatchExtractor can be serialized and deserialized without errors.
• All relevant parameters and fitted attributes are preserved.
• Tests are added to cover typical usage of PatchExtractor.
• "PatchExtractor" is no longer listed in NOT_SUPPORTED_ESTIMATORS.

---

Related file: openmodels/serializers/sklearn/sklearn_serializer.py