confection
Confection is a lightweight library providing a configuration system for hyperparameter management in machine learning. It facilitates the definition of complex object trees and the registration of function calls with an easy decorator syntax. The system is advantageous for making configuration changes without affecting backward compatibility, thanks to versioning support. It processes .cfg files and converts them into dictionaries, supporting JSON-like values and structured sections. Function registries allow for the implementation of complex configurations, including custom training pipelines. Confection offers a straightforward solution for managing configurations in Python.
