You can view & comment the following doc
Is your feature request related to a problem? Please describe.
Feature Tables and Views allow for storage level grouping of features based on how they are produced. However, there is no concept of a retrieval-level grouping of features in Feast. Without a retrieval-level grouping of features that maps to models
- There is no way for users to track which features were used in the training of a model, or which features are being used for the serving of a specific model.
- Retrieving features during training requires a complete list of features to be provided, and this complete list of features needs to be persisted and loaded for online serving. This persistence and loading is a low-level implementation that may be error prone (if not all features are references, or provided in the wrong order).
- There is no way to ensure that consumers won't have breaking changes when feature views and tables change since there is no awareness of features that are in use.
Describe the solution you'd like
See doc #design
Describe alternatives you've considered
See doc #alternatives
Additional context
Add any other context or screenshots about the feature request here.
You can view & comment the following doc
Is your feature request related to a problem? Please describe.
Feature Tables and Views allow for storage level grouping of features based on how they are produced. However, there is no concept of a retrieval-level grouping of features in Feast. Without a retrieval-level grouping of features that maps to models
Describe the solution you'd like
See doc #design
Describe alternatives you've considered
See doc #alternatives
Additional context
Add any other context or screenshots about the feature request here.