You can use the Feature Store UI to:
- Search for feature tables by feature table name, feature, or data source.
- Identify the data sources used to create a feature table.
- Identify models that use a particular feature.
- Check feature freshness.
- Delete feature tables.
- Control access to feature tables.
To use the Feature Store UI, click Feature Store in the sidebar. This icon appears only when you are in the machine learning persona. The Feature Store UI displays. The Feature Store UI lists all of the available feature tables, along with the following metadata:
- Who created the feature table.
- Data sources used to compute the feature table.
- Online stores where the feature table has been published.
- Scheduled jobs that compute the features in the feature table.
- The last time a notebook or job wrote to the feature table.
Use the search box to search for feature tables. You can enter all or part of the name of a feature table, a feature, or a data source used for feature computation. Search text is case-insensitive.
The following screen shot shows an example. The results from searching on “age” include a table with a feature
age, a table with a feature
HouseAge, and a table based on the data source
In the UI you can track both how a feature was created and where it is used. For example, you can track the raw data sources, notebooks, and jobs that were used to compute the features. You can also track the models trained with it and the notebooks and jobs that read it.
In the Feature Store UI, click the name of any feature table to display the feature table page.
On the feature table page, the Producers table provides information about all of the notebooks and jobs that write to this feature table so you can easily confirm the status of scheduled jobs and the freshness of the feature table.
The Features table lists all of the features in the table and provides links to the models, endpoints, jobs, and notebooks that use the feature.
To return to the main Feature Store UI page, click Feature Store near the top of the page.
Deleting a feature table can lead to unexpected failures in upstream producers and downstream consumers (models, endpoints, and scheduled jobs). You must delete published online stores and the underlying Delta table separately.
On the feature table page, click at the right of the feature table name and select Delete. If you do not have Can Manage permission for the feature table, you will not see this option.
In the Delete Feature Table dialog, click Delete to confirm.
If you want to also drop the underlying Delta table, run the following command in a notebook.
%sql DROP TABLE IF EXISTS <feature_table_name>;