Python2.7+spyder+keras+GPU加速，训练过程中老出 I/O operation on closed file的错误

[Hello TiMe]

训练CNN，下面是在迭代，训练过程中经常报以下错误，但有时候不会，能够训练完，请教大家原因？
2717/2717 [==============================] - 0s - loss: 2.9779 - val_loss: 5.7410
Epoch 80/500
2717/2717 [==============================] - 0s - loss: 2.9737 - val_loss: 5.7502
Epoch 81/500
2717/2717 [==============================] - 0s - loss: 2.9646 - val_loss: 5.8087
Epoch 82/500
2717/2717 [==============================] - 0s - loss: 2.9721 - val_loss: 5.7605
Epoch 83/500
2717/2717 [==============================] - 0s - loss: 2.9439 - val_loss: 5.8233
Epoch 84/500
2717/2717 [==============================] - 0s - loss: 2.9256 - val_loss: 5.8236
Epoch 85/500
2717/2717 [==============================] - 0s - loss: 2.9157 - val_loss: 5.8299
Epoch 86/500
2717/2717 [==============================] - 0s - loss: 2.9014 - val_loss: 5.8460
Epoch 87/500
2717/2717 [==============================] - 0s - loss: 2.8909 - val_loss: 5.8696
Epoch 88/500Traceback (most recent call last):
File "<ipython-input-48-66e169292e9b>", line 1, in <module>
runfile('F:/Python/workspace/mnist2/main.py', wdir='F:/Python/workspace/mnist2')

File "H:\Anaconda\lib\site-packages\spyderlib\widgets\externalshell\sitecustomize.py", line 714, in runfile
execfile(filename, namespace)

File "H:\Anaconda\lib\site-packages\spyderlib\widgets\externalshell\sitecustomize.py", line 74, in execfile
exec(compile(scripttext, filename, 'exec'), glob, loc)

File "F:/Python/workspace/mnist2/main.py", line 125, in <module>
train_model(model,X_train,Y_train)

File "F:/Python/workspace/mnist2/main.py", line 96, in train_model
show_accuracy=True, verbose=1, validation_split=0.05)

File "H:\Anaconda\lib\site-packages\keras\models.py", line 597, in fit
sample_weight=sample_weight)

File "H:\Anaconda\lib\site-packages\keras\engine\training.py", line 1107, in fit
callback_metrics=callback_metrics)

File "H:\Anaconda\lib\site-packages\keras\engine\training.py", line 801, in _fit_loop
callbacks.on_epoch_begin(epoch)

File "H:\Anaconda\lib\site-packages\keras\callbacks.py", line 33, in on_epoch_begin
callback.on_epoch_begin(epoch, logs)

File "H:\Anaconda\lib\site-packages\keras\callbacks.py", line 169, in on_epoch_begin
print('Epoch %d/%d' % (epoch + 1, self.nb_epoch))

File "H:\Anaconda\lib\site-packages\ipykernel\iostream.py", line 317, in write
self._buffer.write(string)[/color]

ValueError: I/O operation on closed file

[blackarrow3542]

https://github.com/fchollet/keras/issues/2110 你看下这个链接 This looks like an IO bug with IPython, you might want to file a bug with the devs. It seems to be triggered by the Keras progbar's use of sys.stdout. I'm experiencing the same with Spyder (because it uses IPython kernel too). With each epoch completes at 2s, it results on I/O operation error too. verbose = 0 does the trick. Thanks. 总结下解决方案： 设置verbose=0 或者在model.fit之后加一行time.sleep(0.1)

[Peng学不玩了]

有代码可以一看？

[tju_zhuangshuo]

解决了吗，同样问题