def gauss_ent(mu, logstd): h = tf.reduce_sum(logstd + tf.constant(0.5*np.log(2*np.pi*np.e), tf.float32)) return h