UX: Test if models are platform dependent

[thao-do]

would a provided model from us (trained from working on a tutorial) work on a user's system (of course within the context of the napari plugin)?

Action: will test once we have Training working, we'll pass a model trained on one of our machines to other machines and use it on a same set of images with Prediction to see if it provides the same results. V1 won't allow retraining a (provided) model anyway, so we can test that later for V2.

[benjijamorris]

They should not be! Pytorch, pytorch-lightning, and cuda versions aremore likely to mess things up than the model checkpoints themselves. This is definitely worth testing though!

[benjijamorris]

The "model" (configs + checkpoint) should be os-agnostic. However, different versions of packages might be installed based on the users OS. Some of these are compatible (for example, pytorch lightning versions are largely compatible), but it's possible that users on different operating systems might not be able to share models because of different package versions. CUDA operations in pytorch are often stochastic, so changing CUDA version will slightly change results (I don't think this is a huge effect).
I don't think this will be a problem, but it's worth testing.