TensorFlow Caculate Average Value with Keeping Demension

import tensorflow as tf
import numpy as np

l = np.array([[[0.4,0.6],[0.1,0.9]],[[0.7,0.3],[0.8,0.2]]])
ll = tf.convert_to_tensor(l,dtype=tf.float32)
ppt = tf.reduce_mean(ll,2,keep_dims=True)

with tf.Session() as sess:
    sess.run([init, init_local])
    print(sess.run(ppt))

The result is:

[[[0.5]
  [0.5]]

 [[0.5]
  [0.5]]]

,