hi，when i train segmentation with two classes

[zhangyunming]

hi, when i run the segmentation with two classes ,i get the error as below:

[sacmehta]

It seems like you afe passing one argument as float to Conv layer. Could you print the values and check?

[tokyokuma]

I also faced the same problem.
This is caused by executing the code written in python3 with python2.
I do not know if this solution is correct, but for now, adding this one line to Model.py should resolve it.

Model.py
#print('Config: ', config) #ln180
config = map(int, config) #ln181 adding this line

I hope this method will work.

[zhangyunming]

hi，tokyokuma and sacmehta：

thank you very much for your reply, it is the reason of python2, and i can train the data of mine
but when i change my data , and trained again, there is no model saved,and no information print

[tokyokuma]

The training seems to be able to be executed without problems.
The model has been updated every 1 epoch end.
Perhaps, have not you finished one epoch yet?
If you can train without problems
I think the model is saved in the following file.
segmentation / results_espnetv2_1.0 / model_best.pth

[zhangyunming]

the model is saved in the following

yes,i have train several epoches, but there is no model saved in /results_espnetv2_1.0

[tokyokuma]

Sorry, I have no idea at the moment.

[zhangyunming]

Sorry, I have no idea at the moment.

thank you all the same

but when i use 6 images to train , there is the model saved and the information print

[zhangyunming]

just a whole epoch need train several times? like this:

[zhangyunming]

ok
my falut.
thank you