The inputs given when calling all loss_functions in both CGAN and PCGAN were written backwards, i.e. loss_function(y_pred, y_true) instead of the correct loss_function(y_true, y_pred). This lead to negative loss values when using BinaryCrossentropyLoss.
I also updated super(PCGAN).__init__() to super().__init__()
The inputs given when calling all loss_functions in both CGAN and PCGAN were written backwards, i.e.
loss_function(y_pred, y_true)instead of the correctloss_function(y_true, y_pred). This lead to negative loss values when using BinaryCrossentropyLoss.I also updated
super(PCGAN).__init__()tosuper().__init__()