Skip to content

Changelog

All notable changes to this project are documented in this file.

The format is based on Keep a Changelog and this project adheres to Semantic Versioning.

v2.0.1 (2026-10-01)

Bug Fixes

  • Reintegrate release into workflow.yml (#93, ef72ce0)

  • Update job dependencies in workflow.yml for correct execution order (#93, ef72ce0)

Chores

  • Fix CI/CD workflows for release and manual PyPI publishing (#93, ef72ce0)

  • Update action versions and set timeouts for workflows (#93, ef72ce0)

Continuous Integration

  • Enhance workflow validation and release process (#93, ef72ce0)

Detailed Changes: v2.0.0...v2.0.1

v2.0.0 (2026-09-29)

Bug Fixes

  • catboost: Fix loss_function=None handling, set_params rollback, and doc example (#45, 5b477ae)

  • catboost: Harden ranker loss-function handling and GP regressor tuning (#45, 5b477ae)

  • catboost: Reject top=/mode=Classic conflicts, fix stale top/max_pairs, fix avg_ndcg_score direction (#45, 5b477ae)

  • catboost,ranking: Apply mode=NDCG fix-up to top= embedded via set_params (#45, 5b477ae)

  • catboost,ranking: Compare embedded loss-string params numerically (#45, 5b477ae)

  • catboost,ranking: Make set_params atomic and validate ranking array shapes (#45, 5b477ae)

  • catboost,ranking: Require explicit mode for embedded top=, revert avg_ndcg_score shift (#45, 5b477ae)

  • catboost,ranking: Validate malformed top/max_pairs tokens, fix avg_ndcg_score DataFrame iteration (#45, 5b477ae)

  • catboost,ranking: Validate top/max_pairs family gaps, dedupe group-indexing (#45, 5b477ae)

  • catboost,ranking,tabpfn: Harden loss-function validation, atomic set_params, uncertainty wrapper consistency (#45, 5b477ae)

  • ml: Stop GP unpickling from re-enabling disabled tuning flags (#45, 5b477ae)

  • ranking: Enable metadata routing in tutorial, document pred normalization (#45, 5b477ae)

  • ranking: Reject 2-D group_id/group_ids instead of silently flattening (#45, 5b477ae)

  • ranking: Reject missing values in avg_ndcg_score groups (#45, 5b477ae)

  • ranking: Reject zero-column ensemble arrays, document larger-is-better convention (#45, 5b477ae)

  • ranking: Validate k strictly and clean up top-k analysis docs/notebook (#45, 5b477ae)

  • ranking: Validate missing group IDs in top-k analysis (#45, 5b477ae)

  • tabpfn: Honor configured device for pre-fitted models (#45, 5b477ae)

  • tabpfn: Validate pre-fitted model before embedding extraction (#45, 5b477ae)

Chores

  • Fix example notebook formatting and nbqa check (#45, 5b477ae)

  • Remove deprecated unit tests for example notebooks (#45, 5b477ae)

  • Remove unmaintained polaris-lib dependency and fix example notebooks (still issues) (#45, 5b477ae)

  • Speedup env creation for style check (#45, 5b477ae)

  • Update actions to latest versions in workflow (#45, 5b477ae)

  • Update workflow for Python version and remove test-pypi-publish job (#45, 5b477ae)

  • Uv audit (#45, 5b477ae)

  • ranking: Remove PR status doc, not meant for the repo (#45, 5b477ae)

Documentation

  • Add CI/CD & Release Overview to documentation (#45, 5b477ae)

  • ranking: Document larger-is-better convention in avg_ndcg_score (#45, 5b477ae)

  • ranking: Update PR status doc for top=/mode fix, mother_cv note, test fixes (#45, 5b477ae)

  • tabpfn: Clarify device/precision are not enforced for pre-fitted models (#45, 5b477ae)

Features

  • ranking: Add CatBoost ranker tuning and uncertainty workflows (#45, 5b477ae)

  • ranking: Add functions to convert scores to ranks and score matrices to ranks (#45, 5b477ae)

Refactoring

  • Remove wrong _TransformOnlyValidMols usage (#45, 5b477ae)

  • catboost: Move ranking utility functions to utils module (#45, 5b477ae)

Detailed Changes: v1.2.0...v2.0.0

Resolved Issues: #25, #33, #57.

v1.2.0 (2026-09-01)

Bug Fixes

  • Potential fix for pull request finding (#72, ab22ecf)

  • Remove changelog file reference from semantic release configuration (#72, ab22ecf)

  • Update actions to latest versions in workflow and sync package version in uv.lock (#72, ab22ecf)

  • Update release preflight script to validate commits against the main branch and remove outdated report file (#72, ab22ecf)

Chores

Continuous Integration

  • Enhance release workflow with preflight checks and reporting (#72, ab22ecf)

Documentation

  • Add changelog file configuration to pyproject.toml (#72, ab22ecf)

  • Add initial configuration for Zensical documentation site (#72, ab22ecf)

  • Add markdown extensions and update mkdocstrings configuration (#72, ab22ecf)

  • Add missing changelog entries for v1.1.1, v1.1.2, and v1.1.3 (#72, ab22ecf)

  • Update README to include 'Why Mother?" section (#72, ab22ecf)

Features

  • Add release preflight workflow and reporting script (#72, ab22ecf)

Detailed Changes: v1.1.3...v1.2.0

v1.1.3 (2026-08-31)

Bug Fixes

  • Fix use count based fingerprints as default, including TabPFN precision handling and test updates (#75, 74cac2ec)

  • (tabpfn) Restore upstream TabPFN precision handling with float32 for all operations

v1.1.2 (2026-08-25)

Bug Fixes

  • Solving the max_features parameter not being correctly passed to the superclass (#40, 4fd6869c)

v1.1.1 (2026-08-21)

Chores

  • Update changelog for version 1.1.0 with bug fixes and new features (8a69fa24)

v1.1.0 (2026-08-20)

Bug Fixes

  • (tabpfn) Corrected a bug concerning unsupported BFloat16 for tabpfn. Now force np.float32 for all operations. (#21, ca45d17)

Features

  • (tabicl) add MotherML wrappers, embeddings and uncertainty support (#21, ca45d17)

v1.0.6 (2026-08-18)

Bug Fixes

  • Declare scipy explicitly in base dependencies (#66, 5436111)

  • Local TabPFN import drift and declare SciPy explicitly (#66, 5436111)

v1.0.5 (2026-08-17)

Bug Fixes

  • Make MVS default boostrap for catboost tuning (#61, 7141902)

  • Make MVS the default bootstrap for catboost tuning (#61, 7141902)

v1.0.4 (2026-07-23)

Bug Fixes

  • Correct typo in workflow configuration for error handling (1a54319)

v1.0.3 (2026-07-21)

Bug Fixes

  • Add missing readme field in project metadata (3caa87b)

  • Update Python version and actions in workflow configuration (f14ef68)

Chores

v1.0.2 (2026-07-21)

Bug Fixes

  • Update entry for docs-python-fences hook to use 'uv run poe' (#55, f09baf4)

Documentation

  • Add CLAUDE.md for project guidance and command usage (#55, f09baf4)

  • Enhance documentation and add python code validation for markdown files (#55, f09baf4)

  • Enhance SKILL.md for changelog management and update init.py for version handling (#55, f09baf4)

  • Remove single-source dependency from pyproject.toml and uv.lock (#55, f09baf4)

  • Update README and examples for improved clarity and navigation (#55, f09baf4)

  • Update SKILL.md and CHANGELOG.md for improved documentation workflow and versioning (#55, f09baf4)

v1.0.1 (2026-06-15)

Bug Fixes

  • ml: Unify predict_uncertainty output schema across all model backends and pipelines (#16, 2931c6e)

  • models: Add data and total uncertainty outputs to CatboostRegressorMother predictions (#16, 2931c6e)

  • models: Add knowledge and data uncertainty outputs to RandomForestRegressorMother predictions (#16, 2931c6e)

  • models: Enforce single target type in CatboostGaussianProcessRegressorMother, update predict() and predict_uncertainty() to accept **kwargs (#16, 2931c6e)

  • models: Harmonize prediction outputs across model backends (#16, 2931c6e)

  • models: Improve error message for quantile validation in CatboostRegressorMother (#16, 2931c6e)

  • models: Initialize quantile_array in CatboostRegressorMother predict method (#16, 2931c6e)

  • notebook: Correct minor text and formatting issues in prediction interface guide (#16, 2931c6e)

  • notebook: Update predict_uncertainty interface examples and add CatboostGaussianProcessRegressorMother (#16, 2931c6e)

  • pipeline: Add overloads for mother_cv function to enhance type hints (#16, 2931c6e)

  • pipeline: Improve mother_cv function to conditionally return estimators (#16, 2931c6e)

  • pipeline: Update multi-target prediction fallback to clarify uncertainty handling (#16, 2931c6e)

  • tests: Enhance uncertainty predictions in CatboostRegressorMother and GaussianProcessRegressorMother tests (#16, 2931c6e)

Chores

Testing

  • models: Add regression and classification tests for unified predict behavior (#16, 2931c6e)

v1.0.0 (2026-04-17)

  • Initial Release