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).