feat: Update model architecture and multiprocessing

[mihirkatare]

• Add multiprocessing support for the dataloader (cache-based multiprocessing)
• Add residual network architectures
• Separate model and networks for scalability
• Add draft of API arguments to README

[matthewfeickert]

@mihirkatare Do we need to keep the contents of post/ in version control? They're all binaries, so unless there is a necessity to have them be versioned I'd suggest instead attaching them as artifacts to a GitHub release (which for the time being would mean copying out elsewhere to keep).

[mihirkatare]

@mihirkatare Do we need to keep the contents of post/ in version control? They're all binaries, so unless there is a necessity to have them be versioned I'd suggest instead attaching them as artifacts to a GitHub release (which for the time being would mean copying out elsewhere to keep).

We mention that the library will be pip-installable. Does that include the pre-trained models? i.e the .pth binaries. Otherwise i'll get rid of the post/directory

[mihirkatare]

Okay the binaries have been removed from main

[matthewfeickert]

Does that include the pre-trained models? i.e the .pth binaries. Otherwise i'll get rid of the post/directory

Yeah, I'd distribute those separately, or rather, make them available separately. I'll go back into the commit history and sweep these into another branch that we can just keep around as a vestigial holder for the time being.

Once I do that I'll merge this in.

[matthewfeickert]

@mihirkatare for reference those binaries now have been cherry-picked to the branch backup/post-model-binaries.