I try to implement a model with an image as input and 4 outputs. The data is labeled as [path, y1, y2, y3, y4]. y1 is categorical classification and y2-4 is binary classificatio
