TensorFlow中，variable_scope和name_scope的不同之处

之前一直很困惑，tf.variable_scope和tf.name_scope都是管理上下文环境的，它们有什么不同？

查阅资料时，发现了一段有意思的测试代码

import tensorflow as tfdef scoping(fn, scope1, scope2, vals):    with fn(scope1):        a = tf.Variable(vals[0], name='a')        # 此处注意 b是get_variable        b = tf.get_variable('b', initializer=vals[1])        c = tf.constant(vals[2], name='c')        with fn(scope2):            d = tf.add(a * b, c, name='res')        print('\n  '.join([scope1, a.name, b.name, c.name, d.name]), '\n')    return dd1 = scoping(tf.variable_scope, 'scope_vars', 'res', [1, 2, 3])d2 = scoping(tf.name_scope,     'scope_name', 'res', [1, 2, 3])# 如果加上这一行，就会报错，因为d3的变量b会和d2的变量b冲突# d3 = scoping(tf.name_scope,     'scope_name2', 'res', [1, 2, 3])# 但这一行就不会冲突，因为d3和d1的变量b各自有作用域# d3 = scoping(tf.variable_scope,     'scope_vars2', 'res', [1, 2, 3])with tf.Session() as sess:    writer = tf.summary.FileWriter('logs', sess.graph)    sess.run(tf.global_variables_initializer())    print(sess.run([d1, d2]))    writer.close()

运行后，得到如下输出

scope_vars  scope_vars/a:0  scope_vars/b:0  scope_vars/c:0  scope_vars/res/res:0 scope_name  scope_name/a:0  b:0  scope_name/c:0  scope_name/res/res:0 

TensorBoard显示

总而言之，tf.name_scope仅为非tf.get_variable创建的tensor添加前缀；而tf.variable_scope为所有tensor添加前缀