Problem Using py_function to write a Loss function that works on numpy array

Question

i am currently working on a neural network on google colab with tensorflow and keras and i am trying to implement my own loss function, my problem is that i have to call a v