动态确定在TensorFlow中初始化变量的张量

我发现了一个有点丑陋的方法来做到这一点,我发布了一个帖子,以防其他人会感兴趣:

import tensorflow as tf
import re

var1 = tf.Variable(tf.random_normal(shape=(2, 2)))

shape = var1.initializer.outputs[0].shape
input_ = var1.initializer.node_def.input[1]
tensor_name = (input_ if re.search(r'\d+$', input_) is None
               else '_'.join(input_.split('_')[:-1]))

var2 = tf.Variable(tf.__dict__[tensor_name](shape=shape))

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    print(var1.eval())
    print(var2.eval())
# [[ 0.23845878 -1.1440094 ]
#  [ 0.593299   -1.1108586 ]]
# [[ 1.1235769   1.1481414 ]
#  [ 1.8934027  -0.33171055]]

无论如何,如果有人能提供一种更干净的方法,我会很高兴的。

编辑

这仅在初始化张量具有默认名称时有效。

更新

import tensorflow as tf
var1 = tf.Variable(tf.random_normal(shape=(2, 2), mean=1, stddev=3))
var2 = tf.contrib.copy_graph.copy_variable_to_graph(var1, var1.graph)

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    print(var1.eval())
    print(var2.eval())
# [[1.7009485 0.9412894]
#  [1.0769905 1.3085879]]
# [[ 5.8595214  8.652523 ]
#  [ 1.86671   -3.170361 ]]