global_step in momemtum_optimezer

[agarwl]

Is setting trainable = False not necessary in the variable batch in the function momentum_optimzer? ie. batch = tf.Variable(0,trainable=False,name=global_step) instead of just batch = tf.Variable(0).