Create a Contributing.md file

🚀 Feature

Just a simple .md or .rst file which will help people to get going while contributing.

There are a lot of stuff which people keep stumbling, e.g. linting, docs, tests, CI, etc. which are not actually known at the time of PR for newcomers.

Also, we should have clarification about models since this repo is for models mainly

As discussed, what models should go in pl-bolts is also a major question that we can answer in this file.

1. Many papers claim SOTA and to really reproduce results is terribly hard.
2. Less SOTA but more used features are too necessary, e.g. mobilenet backbones in detection models.
3. API should remain consistent, we cannot keep say Pascal VOC format support for FRCNN and then COCO for Detr, it will create inconsistency for the user.
4. Open source licensing, not all are highly permissive like MIT, some are under Apache or GPL.
5. Support for old models, E.g. we know that Deep Lab does better than U-Net or LinkNet, do we still add them as they were popular and used before? E.g. Efficient-Det does better than FRCNN but it might be wise to support both as both have some caveats at few places. E.g. If we add Retina Net (probably with torchvision very soon), It is a very good single-stage detector, we might not need to use SSD detectors, they aren’t so good.

New SOTA models keep popping in every field, (every year rather) how to keep supporting and deprecating.

1. Production support. Pretrained weights are sometimes hard to train and find. Models and implementations are SOTA but weights aren’t, what do we do then ?

Feel free to add points further too.

cc @Borda @williamFalcon @ananyahjha93 @nateraw

[Borda]

@nateraw I cannot find contributing.md in the lightning repo! Maybe I’m missing it somewhere.

https://github.com/PyTorchLightning/pytorch-lightning/blob/master/.github/CONTRIBUTING.md

[nateraw]

@oke-aditya you want to do it? can assign you. Should be nearly a copy paste from lightning’s contributing.md file (unless you want to rewrite it better here 😉 ).