CUDA 11 could be supported by replacing .gitmodules url by different n3mr repository

[kimjson]

I got this repo working in following library versions:

• Ubuntu 20
• RTX 3090
• Python 3.7
• PyTorch 1.7.0
• CUDA 11.0
• OpenCV 4.1
• Scipy 1.3
• Scikit-Image 0.15

by replacing n3mr submodule url in .gitmodule to https://github.com/adambielski/neural_renderer from https://github.com/daniilidis-group/neural_renderer/issues/106#issuecomment-835361623 like this:

[submodule "external/neural_renderer"]
	path = external/neural_renderer
	url = https://github.com/adambielski/neural_renderer

so want to make contribution in this repository by creating PR.

By the way, I also requested daniilidis-group/neural_renderer maintainer apply changes of that fork to support cuda 11, so I ll add comment in .gitmodules to change the repo url back to original after the request accepted and conducted.

Questions

1. PR to master / PR to cuda-11(new branch) which do you prefer? I don't know which cuda version is more popular these days but I guess 11 is better so I think the default main branch should be cuda 11 compatible.
2. Is there any other contribution guideline?

More suggestions

1. Besides stating library versions in README, how about adding dependency file such as requirements.txt, conda.yaml, or even installation script in order to make the installlation process reproducible? ← which I also want to work on
2. Stating NVIDIA driver version and graphic card model name as experiment setting in README would also be good because I guess they affect compatibility.

[ultmaster]

PR to master / PR to cuda-11(new branch) which do you prefer?

Please PR to master branch.

Is there any other contribution guideline?

If it's not a huge pull request, I can always review and make ad-hoc comments.

adding dependency file

Sounds a good idea.

Stating NVIDIA driver version

I'm not sure whether it's necessary, but good to have.

This repository has been out of maintence for a while. I would appreciate it if you can keep it refreshed with up-to-date libraries. We can also add you as a collaborator if you are willing to contribute in the long run.