[PatrickHenkel1]

Pull Request

Description

warmup_column_inputs und warmup_column_labels are not deterministic because of set implementation

Type of Change

• Bug fix (non-breaking change fixing an issue)
• New feature (non-breaking change adding functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Refactoring (code change that neither fixes a bug nor adds a feature)
• Documentation update

Required Checklist

Testing

• Unit tests have been created/updated for new/modified functionality
• CI/CD pipeline passes all tests (pytest, coverage, executables, installation)

Examples

• Add examples for new features and functionality

Compatibility

• Changes are backward compatible OR deprecation warnings added
• No breaking changes to public APIs
• New dependencies added to pyproject.toml (required in dependencies or optional in [project.optional-dependencies])

Documentation

• Docstrings added/updated for new/modified public methods (Google style)
• Type hints added for new functions/methods

Breaking Changes

Fix is breaking, but API is backwards compatible

Optional

GitHub Copilot Review

• Request Copilot review via GitHub UI (add 'copilot' as a reviewer)

[Copilot]

Pull request overview

This PR fixes non-deterministic warmup column ordering in PreprocessingMultiStep, improving reproducibility of the multi-step preprocessing pipeline used for sequence models (e.g., RNNs). It also expands the serialized preprocessing config to include the derived warmup column lists.

Changes:

• Replace set-based warmup column derivation with order-preserving list comprehensions based on init_features.
• Add warmup_columns_input / warmup_columns_labels to PreprocessingMultiStep.get_config().

Change Classification: Bug fix (behavioral)
Backward Compatibility: API-compatible, but behavior changes (warmup feature ordering) may affect downstream expectations.
CI Risk: Low (localized change).

Key Issues:

1. Missing an explicit unit test that asserts the deterministic warmup column ordering behavior introduced here.
2. get_config() now emits derived warmup columns that are recomputed on init and ignored on reload (passed via **kwargs), which can mislead config editing/round-tripping.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.