文章均迁移到我的主页 http://zhenlianghe.com

my github: https://github.com/LynnHo

tf.name_scope(name, default_name=None, values=None)

1. 域重名
   • 当命名域重名的时候，tf.name_scope会自动对重名的域打上序号

     with tf.name_scope('s'):
         a = tf.placeholder(dtype=tf.int64, name='a')
     with tf.name_scope('s'):
         b = tf.placeholder(dtype=tf.int64, name='b')
     print(a.name)
     print(b.name)
     
     [out]
     s/a:0
     s_1/b:0
     

tf.variable_scope(name_or_scope, default_name=None, values=None, initializer=None, regularizer=None, caching_device=None, partitioner=None, custom_getter=None, reuse=None, dtype=None, use_resource=None)

1. 这个函数会同时创建一个name_scope，即相当于先调用了一次tf.name_scope
2. 域重名
   • 当变量域重名的时候，官方文档 -> If name_or_scope is not None, it is used as is. If scope is None, then default_name is used. In that case, if the same name has been previously used in the same scope, it will made unique be appending _N to it.
     • 如果给定了name_or_scope，那么两个重名的变量域对tf.get_variable生成的变量来说实际上是同一个域，而对于ops来说则是不同的命名域。可以理解为除了tf.get_variable生成的变量，其他ops均受到这个函数创建的tf.name_scope影响

       with tf.variable_scope('s'):
           a = tf.get_variable(name='a', shape=(2))
           c = tf.placeholder(dtype=tf.int64, name='c')
       with tf.variable_scope('s'):
           b = tf.get_variable(name='b', shape=(2))
           d = tf.placeholder(dtype=tf.int64, name='d')
       print(a.name)
       print(b.name)
       print(c.name)
       print(d.name)
       
       [out]        
       s/a:0
       s/b:0
       s/c:0
       s_1/d:0
       

     • 如果name_or_scope为None，那么tf将对重复的命名域打上序号，这个对一些函数式的layer很重要

       
       with tf.variable_scope(None, 's'):
           a = tf.get_variable(name='a', shape=(2))
           c = tf.placeholder(dtype=tf.int64, name='c')
       with tf.variable_scope(None, 's'):
           b = tf.get_variable(name='b', shape=(2))
           d = tf.placeholder(dtype=tf.int64, name='d')
       print(a.name)
       print(b.name)
       print(c.name)
       print(d.name)
       
       [out]
       s/a:0
       s_1/b:0
       s/c:0
       s_1/d:0