v0.9.0

This release drops support for Python 3.8 and 3.9 (both end-of-life) and updates dependencies for modern Python versions.

SHAP 0.45 Compatibility

• Adapted codebase to breaking changes in SHAP 0.45 for multi-output models, updating code and tests accordingly.
• Updated the SHAP MLP explainer to use ProcessPoolExecutor for parallel SHAP value computation.
• Removed SHAP_Analyzer tests, as SHAP's check_additivity now handles validation internally.
• Made no-inplace operations explicit for SHAP compatibility.
• Note: scripts in epiclass/utils/shap may still rely on the old SHAP output shape and need review.
• Python 3.13 support temporarily removed due to compatibility issues.

New Install Script

• Added install.py as a flexible alternative to pip install . for local development.
• Auto-detects CPU/GPU and installs the appropriate PyTorch variant.
• Uses uv if available, falls back to pip.
• Supports optional extras, --freeze to save installed packages, and --torch-version to specify a PyTorch version.
• Exits if not in a virtual environment.

Test Improvements

• Tests now run in parallel via just test-all, with each Python version in a separate background process.
• Made tests robust to parallel execution by using temporary directories for mock data, resolving HDF5 read/write conflicts.
• Enabled pytest -k for filtering tests.
• Improved test log naming for easier navigation.
• Skipped non-crucial confusion matrix creation tests.

Dependency & Build Changes

• Removed legacy per-version requirement files in favor of req_core.in with version specifiers.
• Separated Compute Canada Python indexes into a dedicated file to avoid uv pip install failures.
• Added uv virtual environment setup script.
• Updated requirements for modern SHAP, pandas, and PyTorch compatibility.

Cleanup

• Removed the deprecated TensorFlow module.
• Fixed get_cmap deprecation warning.
• Clarified README.