Because the ML backend and Label Studio are different services, the assets (images, audio, etc.) that you label must be hosted and be accessible with URLs by the machine learning backend, otherwise it might fail to create predictions.
ML backend is unable to access tasks
