Managing data in feature stores.
Feature stores allow data teams to serve data via an offline store and an online low-latency store where data is kept in sync between the two. It also offers a centralized registry where features (and feature schemas) are stored for use within a team or wider organization.
As a data scientist working on training your model, your requirements for how you access your batch / 'offline' data will almost certainly be different from how you access that data as part of a real-time or online inference setting. Feast solves the problem of developing train-serve skew where those two sources of data diverge from each other.
Feature stores are a relatively recent addition to commonly-used machine learning stacks.
The feature store is an optional stack component in the ZenML Stack. The feature store as a technology should be used to store the features and inject them into the process on the server side. This includes
- Productionalize new features
- Reuse existing features across multiple pipelines and models
- Achieve consistency between training and serving data (Training Serving Skew)
- Provide a central registry of features and feature schemas
For production use cases, some more flavors can be found in specific
integrationsmodules. In terms of features stores, ZenML features an integration of
Connect ZenML with already existing Feast
Extend the feature store abstraction and provide your own implementation
If you would like to see the available flavors for feature stores, you can use the command:
zenml feature-store flavor list
The available implementation of the feature store is built on top of the feast integration, which means that using a feature store is no different from what's described on the feast page: How to use it?.