Alibi Detect
An open-source library for drift and outlier detection in model inputs and related signals.
Published by Seldon
How to use it
Fit a detector on known reference inputs, then compare it against realistic production-like samples. Choose a detector and preprocessing steps appropriate to the feature representation.
Keep detector state and preprocessing settings with the model version. Review alerts with data and model checks before changing the serving policy.
- Compatibility
- Python, Tabular data, Embeddings
- Monitoring areas
- Drift detection, Outlier detection, Model input monitoring
- Access
- Local Python library. Dependencies vary by detector and backend.