How to Delete Old Predictions and Fetch New Ones

Summary

This article provides a step-by-step guide to delete old predictions and fetch new ones in Label Studio. If you encounter issues with outdated predictions or need to refresh your predictions, follow these troubleshooting steps to ensure your project displays the most recent predictions generated by your machine learning model.

Troubleshooting Steps

1. Navigate to the Data Manager:

   • Open your project in Label Studio.
   • Go to the Data Manager by clicking on the Data tab.

2. Select Tasks:

   • In the Data Manager, select the tasks for which you want to delete old predictions. You can select multiple tasks by using the checkboxes next to each task.

3. Delete Old Predictions:

   • With the tasks selected, click on the Actions drop-down menu.
   • Select Delete Predictions. This will remove all existing predictions for the selected tasks.

4. Fetch New Predictions:

   • After deleting the old predictions, you need to fetch new ones.
   • Select the tasks again if they are not already selected.
   • From the Actions drop-down menu, select Retrieve Predictions. This will send the selected tasks to your connected ML backend to generate new predictions.

5. Verify New Predictions:

   • Once the new predictions are fetched, verify that they appear correctly in the Data Manager.
   • You can check the Total predictions per task column to confirm that each task has at least one prediction.

More Resources

• Set up machine learning with Label Studio
• Generate predictions from a prompt
• Active learning with Label Studio
• Import pre-annotated data into Label Studio
• Label Studio Documentation

By following these steps, you can ensure that your project always displays the most up-to-date predictions, improving the accuracy and efficiency of your data labeling process.