DL4J LeNet error "NaNs in model output, likely caused by arithmetic underflow"

When I selected and conducted DL4JLeNet analysis using WEKA DL4J, 
The following error msg came out.
=============================================================
org.deeplearning4j.exception.DL4JException: NaNs in model output, likely caused by arithmetic underflow
        weka.classifiers.functions.Dl4jMlpClassifier.distributionsForInstances(Dl4jMlpClassifier.java:1916)
        weka.gui.explorer.ClassifierPanel$20.run(ClassifierPanel.java:1779)

        at weka.classifiers.functions.Dl4jMlpClassifier.distributionsForInstances(Dl4jMlpClassifier.java:1916)
        at weka.gui.explorer.ClassifierPanel$20.run(ClassifierPanel.java:1779)
=============================================================

I used MNIST csv file as input data.
(Please refer to the link below)

https://bigfile.mail.naver.com/download?fid=3YbwKzJRWXKqa3YwHq2ZKqvda3YZKAUZKxUmKxUqKogdaIYwFobZFAv/FqvlHqF0MquwF6Mwp4EwFo0vKquwMotZaAivFz3vpokoMo2l

Convolutioninstanceiterator is used as iterator. Other settings are as shown below.
- Error appears only in WekaDL4J 1.6.x or higher version(1.7.2 ~ 1.6.0.)
- In WekaDL4J version 1.5.x or lower, even if LeNet is selected, no error appears and the analysis works well.
- Even in WekaDL4J 1.6.x or higher version, if you create Lenet with customNet, analysis is performed without errors.

https://bigfile.mail.naver.com/download?fid=3YnwKzJRWXKqa3YZHq2ZKqvda3YZKAUZKxUmKxUqKogdaIYwFobZFAv/FqvlHqbdMotwMxtZKzk0FxgwFo29axuwpAU/KrJCp6UqpoUd

https://bigfile.mail.naver.com/download?fid=3YYwKzJRWXKqa3YqHq2ZKqvda3YZKAUZKxUmKxUqKogdaIYwFobZFAv/FqvlHqMrFqJ4Kqu/M4uqp6J0Kovrp63CpoKwFA2dpxK/FAgZ

https://bigfile.mail.naver.com/download?fid=3eKwKzJRWXKqa3YdHq2ZKqvda3YZKAUZKxUmKxUqKogdaIYwFobZFAv/FqvlHqJCKzKXpztXFxiCaA2/MrtqpAMqMopCK4uZpzKwaAtX
_______________________________________________
Wekalist mailing list -- [email protected]
Send posts to [email protected]
To unsubscribe send an email to [email protected]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html