The KX ML Registry Library contains functionality to create centralized registry locations for the storage of versioned machine learning models, workflows and advanced analytics, alongside parameters, metrics and other important artefacts.
The ML Registry functionality, provided within the
.ml.registry namespace in q, is intended to provide a key component in any MLOps stack built upon KX technology. Registries provide a location to which information required for model monitoring can be stored, retrained pipelines can be pushed and models for deployment can be retrieved.
The functionality aims to enhance our offering and provide users of kdb Insights with:
- A method of introducing a users own models generated outside of kdb Insights to the platform, with wrapped functionality allowing these models to be integrated seamlessly with specified limitations.
- A method to understand stored models.
- A single storage location for all
q/python modelsthat can wrap these models in a way that, upon retrieval, complies with the requirements of kdb Insights for deployment.
Documentation is broken into the following sections: