Model Registry

Introduction

The Model Registry module is a centralized solution for managing machine learning models within our platform. It provides a structured approach to organize, track, and deploy different versions of your models with their associated configurations.

Key Concepts

Registered Models

A registered model serves as a container for different versions of a specific machine learning model. It allows you to:

• Organize related model versions under a single collection
• Manage access controls and permissions at the model level
• Link the registry to deployments and maintain the visibility of versions in use

Model Versions

Each registered model can contain multiple versions, representing iterations or variants of the same model. Each version includes:

• Source information (where the model artifacts are stored)
• Metadata describing the version's characteristics (name, comments, tags)
• Access credentials when required