tensorflow中的tf.group()和tf.tuple()

一、tensorflow中的tf.group()

1.用处

  tf.group()常用于组合一些训练节点，如在Cycle GAN中的多个训练节点，例子如：

generator_train_op = tf.train.AdamOptimizer(g_loss, ...)
discriminator_train_op = tf.train.AdamOptimizer(d_loss,...)
train_ops = tf.groups(generator_train_op ,discriminator_train_op)

with tf.Session() as sess:
  sess.run(train_ops) 
  # 一旦运行了train_ops,那么里面的generator_train_op和discriminator_train_op都将被调用

2.演示

x = tf.constant(5, dtype=tf.float32)
y = tf.constant(7, dtype=tf.float32)
multi = x * y 
add = x + y 
sub = x - y
train_op = tf.group(multi, add, sub)

init = tf.global_variables_initializer()
with tf.Session() as sess:
    sess.run(init)
    print(sess.run(train_op))

  tf.group(multi, add, sub, 这个可以一直写下去的)因为源码定义是*args。
  sess.run(tf.group())结果是None

二、tf.tuple()

1.和tf.group()的区别

  tuple返回的是tensor，可以查看结果。

2.演示

x = tf.constant(5, dtype=tf.float32)
y = tf.constant(7, dtype=tf.float32)
multi = x * y 
add = x + y 
sub = x - y

# 这里传的是 list[op]
train_op = tf.tuple([multi, add, sub])

init = tf.global_variables_initializer()
with tf.Session() as sess:
    sess.run(init)
    print(sess.run(train_op))

结果是 [35.0, 12.0, -2.0]

参考自 https://blog.csdn.net/LoseInVain/article/details/81703786