Does Keras use class_weights in a custom loss function?

Simple question, cannot find anywhere:

An example Custom Loss Function:

def player_loss(self, y_true, y_pred): 
     return K.categorical_crossentropy(y_true,y_pred)

Will this loss function use the class_weights if provided to the fit_gen method when called automatically? Or if you use a custom loss function, you must write a loss function which uses the weights manually?