在tensorflow的计算图中,我们可以利用函数def来封装一些tf操作,但是我们需要使用return语句去规避一些错误,看如下分析:
看如下程序:
import tensorflow as tf
a = tf.Variable(5.0)
def add():
a_change = tf.assign(a, 10.0)
sess = tf.Session()
sess.run(tf.global_variables_initializer())
print(sess.run(add()))
print(sess.run(a))
运行之后,显示如下错误:
Traceback (most recent call last):
File "E:/PythonProject/cnn_cifar10.py", line 12, in <module>
print(sess.run(add()))
File "E:\Python\lib\site-packages\tensorflow\python\client\session.py", line 889, in run
run_metadata_ptr)
File "E:\Python\lib\site-packages\tensorflow\python\client\session.py", line 1105, in _run
self._graph, fetches, feed_dict_tensor, feed_handles=feed_handles)
File "E:\Python\lib\site-packages\tensorflow\python\client\session.py", line 414, in __init__
self._fetch_mapper = _FetchMapper.for_fetch(fetches)
File "E:\Python\lib\site-packages\tensorflow\python\client\session.py", line 231, in for_fetch
(fetch, type(fetch)))
TypeError: Fetch argument None has invalid type <class 'NoneType'>
此时表明add()没有fetch到参数,因此我们需要对函数里的操作添加return操作:
import tensorflow as tf
a = tf.Variable(5.0)
def add():
a_change = tf.assign(a, 10.0)
return a_change
sess = tf.Session()
sess.run(tf.global_variables_initializer())
print(sess.run(add()))
print(sess.run(a))
输出:
10.0
10.0
如果函数中封装了多个操作,就要分情况了
如下程序:
import tensorflow as tf
a = tf.Variable(5.0)
def add():
a_change = tf.assign(a, 10.0)
a_plus_1 = tf.assign_add(a, 1.0)
return a_change
sess = tf.Session()
sess.run(tf.global_variables_initializer())
print(sess.run(add()))
print(sess.run(a))
输出:
10.0
10.0
此时def中有2个操作,但是只return了a_change,所以会获取到参数,不会出现之前的错误,但是由于没有return a_plus_1操作,所以只运行了a_change操作,所以需要同时return 这2个操作,return的顺序无所谓,修改如下:
import tensorflow as tf
def add():
a_change = tf.assign(a, 10.0)
a_plus_1 = tf.assign_add(a, 1.0)
return a_plus_1, a_change
sess = tf.Session()
sess.run(tf.global_variables_initializer())
print(sess.run(add()))
print(sess.run(a))
输出:
(10.0, 11.0)
11.0
a = tf.Variable(5.0)
import tensorflow as tf
a = tf.Variable(5.0)
def add():
a_change = tf.assign(a, 10.0)
y = a_change + 1
return y
sess = tf.Session()
sess.run(tf.global_variables_initializer())
print(sess.run(add()))
print(sess.run(a))
输出:
11.0
10.0
关于用def函数进行封装的操作是否在tf计算图上的说明:
import tensorflow as tf a = tf.Variable(5.0) def add(): a_change = tf.assign(a, 10.0) return a_change
sess = tf.Session() sess.run(tf.global_variables_initializer()) print(sess.run(add())) print(sess.run(a))
上述的tf操作被封装在add()函数中,没有被放在计算图中,若要放在计算图中,则需要调用函数:
import tensorflow as tf a = tf.Variable(5.0) def add(): a_change = tf.assign(a, 10.0) return a_change add() sess = tf.Session() sess.run(tf.global_variables_initializer()) print(sess.run(add())) print(sess.run(a))