[Feature] State of the Open-MMLab project

[JohannesTheo]

What is the feature?

Hello, I noticed that you have recently resumed development with a few commits? 👀

Could you tell us something about the general status and plans of the Open-MMLab project and the various repositories? I don't know where else to ask this and since I've seen your activity @MGAMZ and @HAOCHENYE, here might be as good as anywhere :D

Any information would be very welcome since some people are starting to claim your project, e.g. open-mmlab/mmcv#3302 (comment)

Any other context?

No response

[JohannesTheo]

Best to follow progress here #1665

[MGAMZ]

I am contributing commits from my mmengine usage experiments across these years.

mmengine has a great framework from my view point. Only a few modifications is enough to make it happy with the latest deep learning dependencies.

It still needs everyone's high quality contributions now, as former major developpers have new jobs to do.

And, some developpers have redistributed their own version of OpenMMLab, e.g. ONEDL. It may have a more stable release plan. I'm also keeping an eye on these projects.

[JohannesTheo]

I'd love to help, but aside from mmengine, the repos do not appear to be maintained anymore. Most have a huge backlog of pending PRs already.

@HAOCHENYE do you have info regarding the maintainers of the other open-mmlab repos? I'd start contributing but only if it makes sense :)

[Petros626]

https://github.com/VBTI-development/onedl-mmengine

[lauriebax]

@JohannesTheo we are a small company in the Netherlands who aim to support a subset of mmlabs repos. We forked them and made them work with pytorch 2.8 and python 3.10 and 3.12. Check out the work on our organization page.

We are committed to review all PRs that are made to our repos and greatly appreciate any contributions!

[JohannesTheo]

Hey @lauriebax , thank you for the explanation! I've seen your project, it's hard to miss ;) I'm sure many people will appreciate the easier install. Your effort is clearly honorable but I'm not yet sure if refactoring is enough moving forward. The strength of openMM-lab was that many researchers used it to implement and publish new methods (in external 'projects'). Maybe it is time for something new and more lightweight, idk. Will certainly watch the progress of your project, nice work so far.