module ‘tensorflow.python.pywrap_tensorflow‘ has no attribute ‘NewCheckpointReader‘的解决方法

  

博客首页：knighthood2001

欢迎点赞评论️

❤️ 热爱python，期待与大家一同进步成长！！❤️

给大家推荐一款很火爆的刷题、面试求职网站

利用以下代码获取输出节点的检查点模型名称。

from tensorflow.python import pywrap_tensorflow

checkpoint_path = r'E:\crack\models_trainning\my_model\ckpt-2'
reader = pywrap_tensorflow.NewCheckpointReader(checkpoint_path)
var_to_shape_map = reader.get_variable_to_shape_map()

for key in var_to_shape_map:
    print("tensor_name: ", key)

不过出现了以下错误

Traceback (most recent call last):
  File "C:/Users/knighthood/OneDrive/桌面/tensorflow_ckpt_2_pb-master/tensorflow_ckpt_2_pb-master/view_ckpt_graph.py", line 6, in 
    reader = pywrap_tensorflow.NewCheckpointReader(checkpoint_path)
AttributeError: module 'tensorflow.python.pywrap_tensorflow' has no attribute 'NewCheckpointReader'

即

module 'tensorflow.python.pywrap_tensorflow' has no attribute 'NewCheckpointReader'

原因：由于tensorflow版本有1.和2.的区别，以上只适用于tensorflow 1.的版本，争对tensorflow 2.的版本，可以使用如下代码

import tensorflow.compat.v1 as tf1

tf1.disable_v2_behavior()
checkpoint_path = r'E:\crack\models_trainning\my_model\ckpt-2'
# Read data from checkpoint file
reader = tf1.train.NewCheckpointReader(checkpoint_path)
var_to_shape_map = reader.get_variable_to_shape_map()

for key in var_to_shape_map:
    print("tensor_name: ", key)

运行结果如下：