def test_embedding_look_up():
    input_ids = tf.placeholder(dtype=tf.int32, shape=[3,2])

    #embedding = tf.get_variable('test', shape=[5,5])
    embedding = tf.Variable(np.identity(5, dtype=np.int32))
    input_embedding = tf.nn.embedding_lookup(embedding, input_ids)

    sess = tf.InteractiveSession()
    sess.run(tf.global_variables_initializer())
    print("embeding:\n", embedding.eval())
    result = sess.run(input_embedding, feed_dict={input_ids:[[1, 2], [2, 1], [3, 3]]})
    print("结果：\n", result)
    print(result.shape)
test_embedding_look_up()

outputs:
embeding:
 [[1 0 0 0 0]
 [0 1 0 0 0]
 [0 0 1 0 0]
 [0 0 0 1 0]
 [0 0 0 0 1]]
结果：
 [[[0 1 0 0 0]
  [0 0 1 0 0]]

 [[0 0 1 0 0]
  [0 1 0 0 0]]

 [[0 0 0 1 0]
  [0 0 0 1 0]]]
(3, 2, 5)