When label dimension is less then number of last layer

Question

If I have one label, one value to predict at each time, but If I write the following way the last layer tf.Keras.layers.Dense(3), the number of prediction value