How to use a trained TensorFlow model inside a custom loss function in Keras

Question

Problem Statement

I want to use a pretrained TensorFlow model inside a custom loss function to train another model, but get an error message. I am using TensorFlow 2.3 Keras with this code, where the pretrained model named 'discriminator_model' is handed over to the custom_loss:

Code snippet

def custom_loss(discriminator_model):

    def loss_function(y_actual, y_pred):
        prediction_actual = discriminator_model.predict(y_actual)
        prediction_pred = discriminator_model.predict(y_pred)
        custom_loss = (prediction_pred - prediction_actual)**2
        return custom_loss

    return loss_function

...

    model.compile(optimizer=tf.keras.optimizers.Adam(lr_schedule), loss=custom_loss(discriminator_model), metrics='mse')
    generator_model.fit(DG_train, batch_size=batch_size, epochs=epochs)

Error message

RuntimeError: Detected a call to `Model.predict` inside a `tf.function`. `Model.predict is a high-level endpoint that manages its own `tf.function`. Please move the call to `Model.predict` outside of all enclosing `tf.function`s. Note that you can call a `Model` directly on `Tensor`s inside a `tf.function` like: `model(x)`.

Is there a way to use a pretrained model inside a custom loss function? I think there should be, as this is also the underlying idea behind GANs, right? But how?