Feature/multiclass cli classify

[gbeane]

This pull request significantly refactors and extends the src/jabs/scripts/classify.py script to add robust support for multi-class classification in addition to binary classification. It introduces new helper functions for classifier loading and training, updates the main classification logic to handle both classifier types, and improves code clarity and type safety throughout. The changes also enhance user feedback and error handling, and update command-line argument parsing and documentation.

Key changes include:

Multi-class Classification Support:

• Added support for multi-class classifiers throughout the script, including new helper functions (_is_multiclass_training_file, train_multiclass, _load_classifier_from_pickle) and updates to the main classification and training logic to handle both binary and multi-class cases. [1] [2] [3]

Refactoring and Type Improvements:

• Updated function signatures and variable types for better type safety and clarity, including explicit return types and clearer argument documentation. [1] [2] [3] [4] [5] [6]

Enhanced Classifier Loading and Error Handling:

• Improved classifier loading logic to auto-detect classifier type from pickle files, with robust error handling for version mismatches and unrecognized types. [1] [2]

Command-line Interface and Documentation:

• Updated command-line argument help messages, improved docstrings, and clarified usage instructions for both binary and multi-class workflows. [1] [2] [3] [4]

Prediction Output and File Handling:

• Modified prediction output logic to handle multi-class probability matrices and updated output file naming and directory creation for consistency. [1] [2]

These changes collectively make the script more robust, extensible, and user-friendly for both binary and multi-class behavior classification tasks.

[Copilot]

Pull request overview

This PR refactors src/jabs/scripts/classify.py to support both binary and multi-class behavior classification from the CLI, including training from exported HDF5 and classification using either training exports or serialized classifier pickles.

Changes:

• Added multi-class training detection and training path (_is_multiclass_training_file, train_multiclass), and updated classification flow to accept Classifier | MultiClassClassifier.
• Updated prediction writing to support multi-class probability tensors and to use a dedicated multi-class prediction key.
• Improved CLI help text and type annotations across the script, plus added a new classifier pickle loading helper.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.