Deploying a Keras model in Windows?

[Jacob Williams]

I'm looking for some advice on making an executable that can run a Keras model in Windows. Well, more specifically, will host a Flask server that will report the results of Keras models classification locally so it can interface with an existing code base that's written in C#. I have the server running and communicating with the C# well enough (I'm just at the proof of concept stage and not too concerned with performance/security yet), but now I need to be able to deploy it to some other machines that don't have Theano/Keras installed and it needs to be self-contained.

I've tried py2exe and cx_freeze with no luck so far. py2exe misses a large number of Theano dependencies and cx_freeze crashes with a "Type Error: can only concatenate list (not "NoneType") to list". Does anyone have advice on getting this sort of thing to work? I'm willing to try another approach than making an exe, as long as I can port the server to another machine without having to require a prior Theano installation.