Using PythonKit with complex python modules

I've only been leaning Swift for about six months, but thanks to some amazing help from @SDGGiesbrecht, I've come a long way.

I've been looking for a package or framework to fulfill a particular purpose I have, and haven't had a lot of luck finding anything that will work for me in Obj-C or even C++, but there seems to be a couple options in Python, and thanks to PythonKit, it looks like they might actually be usable in my Swift app.

I don't know Python at all, but looking at the documentation for PythonKit, it looks similar enough to Swift that I think I can work out what I need to do. But I'm missing a step in getting the module I want to work with set up, I think?

First of all, I assume ultimately I need these files in my Bundle instead of in the repository I downloaded, correct? Is there anything specific I should do to compile them before adding them to the Bundle?

Second, I know how to use a specific .py file:

    func testImport() {
        let sys = Python.import("sys")
        sys.path.append("/path/to/mymodule/")
        let myModule = Python.import("mymodule")   // where mymodule is mymodule.py
    }

but the module I want to work with is comprised of many .py files. __init__.py seems a likely suspect, but it doesn't seem to have the functions I want, nor does it import them from the other .py files, and since I don't know Python, I'm not sure how self-contained each .py file is. Will it have functions that will call functions contained in other .py files? If so, do I need to import or load those as well?

I haven't had any luck finding a tutorial for this that assumes no real starting knowledge of Python. Any pointers would be appreciated.

I'm wondering a similar thing about module dependencies. They don't seem to be loading successfully with my custom module. For example, I'm getting an error loading torchvision, from within one of my modules, which seems strange since it's definitely installed in my running virtualenv. But for some reason torchvision also fails when loading directly from Python.import("torchvision").