What problem does it solve? When extending a machine learning codebase, new datasets, models, and augmentations must follow consistent factory and registry patterns or the project becomes fragmented and hard to maintain. This Skill enforces a standardized architecture so every new component registers correctly and stays config-driven. ## Core Features & Use Cases - Registry Pattern Guidance: Shows how to add new Dataset, Model, and Augmentation classes using @register_dataset, @register_model, and @register_augmentation decorators. - Config-Driven Model Design: Enforces models whose __init__ accepts only a Hydra cfg object and whose forward() returns a dict with loss, labels, and logits. - Auto-Import Discovery: Explains how module __init__.py files auto-discover submodules so new files register without manual imports. - Use Case: When adding a new brain-decoder model to an ML research project, follow the Skill to create the file in src/model_module/, decorate it with @register_model('MyModel'), wire all hyperparameters through the Hydra config, and return the standard output dict. ## Quick Start Ask the AI to create a new dataset class for the project following the architecture-design patterns with the proper registration decorator and config-driven structure.
