Transforming keras model output during training and use multiple losses

Question

       if self.use_embedding:
            input_x = Input(shape=(self.input_length,), name=\'embedding_input\')
            x = Embedding(self.input_dim, self.embedd