Refactor for easier experimentation

[jcreinhold]

The current state of the code is bad. It's hard to understand and essentially hard-coded for the experiments I ran for the paper. But the core functionality could be broken out into class/module which could be extended for other experiments and other types of data.

[HarshaSatyavardhan]

Hey is this still progress. I was trying to extend to different datasets but finding hard to understand the code

[jcreinhold]

I don't have plans to refactor this code anytime soon.

[jcreinhold]

In the code's current state, you'd need to write a new module/model which is located in experiment to extend this framework to other datasets/diseases. Easier said than done, but it's not impossible. Doing so would likely require fairly deep knowledge of Pyro internals, if I recall correctly. And I'm not even sure if this framework is compatible with the current implementation of Pyro/PyTorch.

[HarshaSatyavardhan]

Is there an alternative to this ? As I am new to pyro, are there any implementations of this or deepscm paper with proper documentation available that are compatible with the latest version of Pyro, to the best of your knowledge ?.