Predictions are stored as data row attributes. To view them, open data rows in Catalog.
You can use Catalog filters to view individual data rows or to find data rows in a model run.
These features are currently in beta preview. Some behavior may change before release and some recent improvements may not yet be documented.
To use a Catalog filter to find a specific data row, you need to know its external ID, data row ID, or global key. To find a data row using one of these values:
From Catalog, select Data row from the Search menu.
From the Attribute menu, select the attribute you want to search.
Enter your attribute value in the search box.
Select the data row to view its predictions.
To view predictions for data rows in a model run, use a Catalog filter to search for the model run data rows:
From the Catalog Search menu, select Model.
From the Model menu, select the model used in your model run.
Use the Model run menu to select your model run.
When viewing multiple data rows, you can use the banner controls to scroll through the list.
You can use Model to locate specific model runs.
Start by selecting the model from Model gallery. This opens the model details on an Overview tab:
Model cards display recent model runs in the right column, just below the model pricing details. If you find the model run here, you can select it to view its data rows.
Alternatively, select the Runs tab and then select your model run from the list.
To view model run details, use Model to open the model run and then select Model run details from the Settings menu.
The details are displayed in an overlay panel.
The details panel shows the cost of the model run and displays configuration as a Python dictionary. You can copy the configuration to the Clipboard or download it.
You can also use the Notification Center to find model run predictions:
To begin, select Notifications from the Labelbox app menu.
From the Task Type menu, select Generate model predictions.
To view the data rows for a given model run, select View model run in Catalog.
If your model run encounters a problem, you can use the Notification Center to download an error log.
To do so, use Notifications to locate the model run and the select Download error log.
Updated 18 days ago