How to use class_weights in keras for multioutput network?

Question

I have one input, and multiple outputs, like a multilabel classification, but I chose to try another approach to see if I have any improvements.

I tried passing a lis