Hey @saudet , any plan for releasing this fix soon? Thanks

@karllessard That only affects compilation, not runtime

Yes but we were in the process of upgrading to TensorFlow 2.18 and this is blocking us right now, since I need to compile locally once for getting the ops generated.

We can easily use the snapshots:
http://bytedeco.org/builds/

Hi Samuel,
it would be great if we could a new release for this out.
Thanks,
Jiri

Yeah, well, NVIDIA seems to think otherwise, see bytedeco/javacpp-presets#1620 (comment)

Is it technically possible to release 1.5.12 for everything except Pytorch and OpenCV, and then release these two (as 1.5.12, or 1.5.13) as soon as the issue is fixed?

Sure, anything is possible, but it's not something I will personally take the time to deal with. It's not like NVIDIA doesn't have enough money to help us out now do they. 😮‍💨

Sure, but they don't care about us. Maybe they'd even care about this issue if we could reach someone in the trenches and let them know that the issue exists, but I unfortunately don't have any contact...

As for the help to this project, I feel your frustration. But I do think that you're too hesitant of explicitly asking for financial support by companies. Yes, GitHub offers easy way to sponsor a project on a small scale, but companies especially don't care about something lke that unless someone pushes the people in seeing this as an issue. Maybe there should be an explicit plea for funding, with explicit reasoning, ways to contribute, budget needed, expected amounts, etc. I'm sure that wouldn't send tons of money towards your efforts with JavaCPP, but I'm pretty sure there would be some funds, and, who knows, maybe a decent support after some time. I understand that you may be frustrated enough with the cruel corporate world that you don't want to go that route, and that you've tried many times in the past with insufficient success, but times have changed in the past decade I believe. Anyway, currently I don't see any prominent call to action related to that on the JavaCPP website, so even people who could help, just don't think about that and don't even know that you need it.

It's all politics at this point, and I don't do politics. If the Java community had a need of JavaCPP, there would be demand for that kind of tool out there, but there isn't, not anymore, Oracle made sure of that. Anyway, I've been able to fix the builds in a satisfactory fashion, so if there are no other major issues, we should be able to make a release by the end of the month.