出错Fetch argument None has invalid type

Question

Step=0,Train loss=2.3095627,[Tset accuracy=0.21733333]

Traceback (most recent call last):

 File "f:\machine learning\practice\cnn_mnist.py", line 102, in <module>

   train_loss,train_op = sess.run([loss,train_op],{input_x:batch[0],output_y:batch[1]})

 File "G:\python\lib\site-packages\tensorflow\python\client\session.py", line 895, in run

   run_metadata_ptr)

 File "G:\python\lib\site-packages\tensorflow\python\client\session.py", line 1113, in _run

   self._graph, fetches, feed_dict_tensor, feed_handles=feed_handles)

 File "G:\python\lib\site-packages\tensorflow\python\client\session.py", line 420, in __init__

   self._fetch_mapper = _FetchMapper.for_fetch(fetches)

 File "G:\python\lib\site-packages\tensorflow\python\client\session.py", line 240, in for_fetch

   return _ListFetchMapper(fetch)

 File "G:\python\lib\site-packages\tensorflow\python\client\session.py", line 347, in __init__

   self._mappers = [_FetchMapper.for_fetch(fetch) for fetch in fetches]

 File "G:\python\lib\site-packages\tensorflow\python\client\session.py", line 347, in <listcomp>

   self._mappers = [_FetchMapper.for_fetch(fetch) for fetch in fetches]

 File "G:\python\lib\site-packages\tensorflow\python\client\session.py", line 237, in for_fetch

   (fetch, type(fetch)))

TypeError: Fetch argument None has invalid type <class 'NoneType'>

就是只训练了一次，错误在这部

train_loss,train_op = sess.run([loss,train_op],{input_x:batch[0],output_y:batch[1]})

Answer 1

Oscar 2018-06-30 15:06:44

问题是在 46 行的地方，那个应该是 train_op_， 和之前我们在 34 行定义的 train_op 不是一个哦：   

train_op = tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss)

它多一个下划线（_）。 

所以只要将 46 行这句

train_loss, train_op = sess.run([loss, train_op], {input_x: batch[0], output_y: batch[1]})

改为

train_loss, train_op_ = sess.run([loss, train_op], {input_x: batch[0], output_y: batch[1]})

即可。当然了，这个 train_op_ 我们后面没有用到，所以你随便取什么名字都行。

Answer 2

Oscar 2018-02-26 21:14:04

能否贴出你的代码呢？谢谢