We've observed hundreds of model deployments, and based on this experience, Superwise is built on a few important concepts and abstractions to monitor models effectively. Understanding Superwise concepts are important to understand the platform's use and capabilities fully. The following set of guides will guide you through the following concepts:
- Models - A machine learning model observed and monitored.
- Versions - A specific model version.
- Baselines - A baseline of a version. For example, training dataset.
- Segments - Subpopulations that require observability.
- Metrics - Model metrics that measure the entire ML process.
- Policies - Monitoring policies that scan and detect potential issues.
- Incidents - Grouped and correlated issues for certain populations in a given timeframe.
- Integrations - The set of potential upstream app integrations available.
- Transactions - The basic atomic production data transfer sent to Superwise.
Updated 5 months ago