tensorflow兼容处理 tensorflow.compat.v1及module ‘tensorflow‘ has no attribute ‘reset_default_graph‘

因为tensorflow的版本问题，所以引用其函数时总是容易报错。

本来用的以下代码：

import tensorflow.compat.v1 as tf1

tf1.disable_v2_behavior()
checkpoint_path = r'D:/INCREASE-main/Beijing/data/INCREASE_SP_pretrained'
# 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)

想要查看一下训练的节点。但是报错。

 改为：

import tensorflow as tf1

tf1.compat.v1.disable_v2_behavior()
checkpoint_path = r'D:/INCREASE-main/Beijing/data/INCREASE_SP_pretrained'
# Read data from checkpoint file
reader = tf1.compat.v1.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)

可以运行了，结果如下：