MPSBoost 1.0.0 Release Audit¶
1.0.0 is the stable customer-commitment release for the native CPU/MPS MPSBoost scope.
Scope¶
- Native CPU and MPS tree-learning backends.
- sklearn-style gradient boosting, decision tree, random forest, ExtraTrees, and CatBoost-like numeric estimators.
- Binary and multiclass classification with native CPU softmax and explicit OvR compatibility.
- Quantile, Poisson, Tweedie, feature importance, permutation importance, and approximate SHAP-like explanations.
- CPU-suitable anomaly detection and pointwise ranking with warning-based MPS routing.
- Explicit optional SHAP and portable backend diagnostics.
- Project-local real-world dataset cache policy and documented HIGGS local-file boundary.
- Customer-facing setup, skip, warning, and external-dependency attribution paths.
Validation¶
PYTHONPATH=src pytest -q tests/real_world -rs: 21 passed, 1 HIGGS local-file boundary skip.PYTHONPATH=src pytest -q tests/unit/test_portable_backend.py tests/unit/test_diagnostics.py tests/packaging/test_public_api.py::test_completed_estimators_are_public tests/packaging/test_public_api.py::test_estimator_capability_registry_reports_available_and_planned_models -rs: 12 passed.- Fresh Python 3.13 wheel environment,
/tmp/mpsboost-release-check-100-py313/bin/python -m pytest -q tests/unit tests/integration tests/packaging -m "not gpu" -rs: 220 passed. - GitHub CI and Docs passed on the 0.5.0 publication record.
mkdocs build --strict: passed after the 1.0.0 documentation updates and zh-Hans navigation translations.python -m twine check dist/mpsboost-1.0.0-*.whl: passed.- Fresh local wheel smoke passed on Python 3.10 and Python 3.13: import, version
1.0.0, multiclassGradientBoostingClassifier.fit,predict_proba, andsystem_info().
Artifacts¶
dist/mpsboost-1.0.0-cp310-cp310-macosx_13_0_arm64.whl: 280K, SHA2560ef03786ba5c1511cca88b126d5632fca8ee5f49cf8cc80dafd9499f39b6e1fa.dist/mpsboost-1.0.0-cp313-cp313-macosx_13_0_arm64.whl: 284K, SHA25661df4018d7f936a7b4814485b22a0ed755ef529bb359117aafc15e8cd972c4b5.- Wheel tags are
macosx_13_0_arm64for Apple Silicon compatibility across supported modern macOS releases.
Artifact Rules¶
The wheel must not contain raw datasets, test caches, specs/, tests/, benchmarks/, .github/,
build directories, credentials, or runner files.
PyPI¶
- PyPI URL: https://pypi.org/project/mpsboost/1.0.0/
- Upload command:
python -m twine upload dist/mpsboost-1.0.0-cp310-cp310-macosx_13_0_arm64.whl dist/mpsboost-1.0.0-cp313-cp313-macosx_13_0_arm64.whl - Fresh formal PyPI install smoke passed on Python 3.10 and Python 3.13:
python -m pip install --no-cache-dir mpsboost==1.0.0, import, version1.0.0, multiclassGradientBoostingClassifier.fit,predict_proba, andsystem_info().