Having trouble running inference on models that contain Japanese characters

[tmbrye]

I have created a model using Japanese characters and am unable to run the inference correctly as it is unable to match up my input values with the model values. This wouldn't be an issue if I had control of the JVM as I'd just add -Dfile.encoding=UTF-8 to the startup, but in this case the JVM is always up and running and unfortunately has its file encoding set by default to Ascii which causes the column names in the model.ini file to not process correctly.

I tried looking to see if there is a way to override the backendreader encoding but have yet to find one. I could possibly rewrite the H2O code that hands this, but was hoping there is another possible way to somehow update the file encoding of the reader before it processes loading the model.

Any ideas are appreciated. Thanks.