Lr find for multitask learning

Hi All,
How to perform lr find on multitask network, for example, a YOLO network outputs a. confidence for a pixel being some object and b. bounding box values for the object?