Tensorflow共享变量

tf.name_scope() vs tf.variable_scope()

1. tf.name_scope creates namespace for operators in the default graph
2. tf.variable_scope creates namespace for both variables and operators in defaul graph
   3 tf.get_variable will ignore “name scope” but use “variable scope)

for example :

In [9]: with tf.name_scope("scope1"):
   ...:     v1=tf.get_variable("var",[1],dtype=tf.float32)
   ...:     v2=tf.Variable(1,name="var2",dtype=tf.float32)
   ...:     a=tf.add(v1,v2)
   ...:     

In [10]: print(v1.name)
var:0

In [11]: print(v2.name)
scope1/var2:0

In [12]: print(a.name)
scope1/Add:0

In [13]:  with tf.variable_scope("scope2"):
    ...:    v1=tf.get_variable("var",[1],dtype=tf.float32)
    ...:    v2=tf.Variable(1,name="var2",dtype=tf.float32)
    ...:    a=tf.add(v1,v2)
    ...: 

In [14]: print(v1.name)
scope2/var:0

In [15]: print(v2.name)
scope2/var2:0

In [16]: print(a.name)
scope2/Add:0

Use cases:

• tf.name_scope(name) for (name scope)
• tf.variable_scope(name_or_scope) for (variable scope )
• -tf.op_scope(values,name,default_name=None) for (name scope)

• tf.variable_op_scope(values,name_or_scope,Default_name=None) for (variable scope)

• -