2025-03-27 16:30:08 <@tflink:fedora.im> !startmeeting fedora-ai-ml-sig 2025-03-27 16:30:09 <@meetbot:fedora.im> Meeting started at 2025-03-27 16:30:08 UTC 2025-03-27 16:30:09 <@meetbot:fedora.im> The Meeting name is 'fedora-ai-ml-sig' 2025-03-27 16:30:15 <@tflink:fedora.im> !hi 2025-03-27 16:30:16 <@zodbot:fedora.im> Tim Flink (tflink) 2025-03-27 16:30:32 <@dherrera:fedora.im> !hi 2025-03-27 16:30:33 <@zodbot:fedora.im> Diego Herrera (dherrera) - he / him / his 2025-03-27 16:31:51 <@mystro256:fedora.im> !hi 2025-03-27 16:31:52 <@zodbot:fedora.im> None (mystro256) 2025-03-27 16:32:41 <@xanderlent:fedora.im> !hi 2025-03-27 16:32:42 <@zodbot:fedora.im> Alexander Lent (xanderlent) 2025-03-27 16:33:24 <@tflink:fedora.im> ok, let's get started 2025-03-27 16:33:48 <@tflink:fedora.im> !topic ROCm 6.4 Release 2025-03-27 16:33:56 <@tflink:fedora.im> @trix I think this is you? 2025-03-27 16:34:35 <@trix:fedora.im> yes. this is more informational, that 6.4 is coming and while it is happening rawhide rocm bits will be in flux. 2025-03-27 16:35:28 <@tflink:fedora.im> !info ROCm 6.4 will be released soon. While the Fedora packages are being updated, there may be some instability until those updates are complete 2025-03-27 16:35:36 <@mystro256:fedora.im> Yeah I'll be doing most of the work, Tom helping if need be 2025-03-27 16:35:44 <@trix:fedora.im> yup 2025-03-27 16:35:49 <@mystro256:fedora.im> upstream says release is "soon" 2025-03-27 16:36:02 <@mystro256:fedora.im> so I'll keep my eye on it 2025-03-27 16:36:27 <@tflink:fedora.im> cool, sounds good 2025-03-27 16:36:56 <@tflink:fedora.im> ok, moving on to the next (I think) quick one 2025-03-27 16:37:04 <@tflink:fedora.im> !topic Intel/AMD NPU Drivers 2025-03-27 16:37:16 <@tflink:fedora.im> Alexander Lent: I think this is yours? 2025-03-27 16:37:44 <@xanderlent:fedora.im> Yep, not much has changed. Biggest win is we have firmware for both in Fedora proper. 2025-03-27 16:38:08 <@xanderlent:fedora.im> Update your system and the kernel drivers will Just Workโ„ข. 2025-03-27 16:38:11 <@tflink:fedora.im> !info Firmware for both Intel and AMD NPUs is shipping for all Fedora branches 2025-03-27 16:38:46 <@tflink:fedora.im> it sounds like there is still work to be done for both the NPUs, though? 2025-03-27 16:38:56 <@xanderlent:fedora.im> Correct. I'm still working on both user-space drivers otherwise, relatively little progress since last meeting. 2025-03-27 16:39:19 <@xanderlent:fedora.im> On the AMD side they are seeking patches upstream so I'm going to get involved with their roadmap. 2025-03-27 16:39:19 <@tflink:fedora.im> !info work is still needed for the user-space drivers for both NPU stacks 2025-03-27 16:39:59 <@xanderlent:fedora.im> On the Intel side I need to sit down and think it out a bit. 2025-03-27 16:40:22 <@xanderlent:fedora.im> That's is everything for the NPU. 2025-03-27 16:40:38 <@xanderlent:fedora.im> That's everything for the NPU. 2025-03-27 16:40:42 <@tflink:fedora.im> cool, thanks for the update 2025-03-27 16:41:00 <@tflink:fedora.im> !topic Huggingface AI Libraries 2025-03-27 16:41:08 <@tflink:fedora.im> Alexander Lent: I think this is also yours 2025-03-27 16:41:33 <@xanderlent:fedora.im> Yep. First thing was a status update. 2025-03-27 16:41:59 <@tflink:fedora.im> not sure I understand what you mean by needing another review round 2025-03-27 16:42:22 <@tflink:fedora.im> does that mean that you've fixed some things and you want it to be reviewed again? 2025-03-27 16:42:45 <@xanderlent:fedora.im> Yes, I will work with @trix who is the reviewer on that. 2025-03-27 16:42:50 <@tflink:fedora.im> ah, ok 2025-03-27 16:43:10 <@tflink:fedora.im> !info python-safetensors has been updated and is ready for review again 2025-03-27 16:43:27 <@trix:fedora.im> ok, sorry i need get on that stat! 2025-03-27 16:43:54 <@xanderlent:fedora.im> No problem; we've both been busy recently. ๐Ÿ™‚ 2025-03-27 16:44:07 <@tflink:fedora.im> !info python-accelerate packaging will start soon 2025-03-27 16:44:28 <@xanderlent:fedora.im> 2025-03-27 16:44:28 <@xanderlent:fedora.im> That's the whole bullet point. 2025-03-27 16:44:28 <@xanderlent:fedora.im> My plan is to submit python-accerlerate next, which is a pure-python package for automatically using accelerator hardware in PyTorch. 2025-03-27 16:45:20 <@tflink:fedora.im> Any thoughts on whether we want to push for a self-contained change about HF packages for F43? 2025-03-27 16:45:24 <@xanderlent:fedora.im> The final thing was that I've been thinking about listing the remaining Huggingface packages - there are about four or five main ones still TODO. 2025-03-27 16:45:38 <@trix:fedora.im> python-accelerate has a lot of dependencies, i do not think are met. 2025-03-27 16:45:54 <@tflink:fedora.im> Alexander Lent: do you think that they'll be done for F43, it sounds like there is quite a bit of work left 2025-03-27 16:46:32 <@xanderlent:fedora.im> 2025-03-27 16:46:32 <@xanderlent:fedora.im> I think a formal change might help with marketing and/or coordinating all the work. 2025-03-27 16:46:32 <@xanderlent:fedora.im> I am not certain I'll have time for F43 given the amount of work. Perhaps I can write a draft and we can do it in the F44 cycle? 2025-03-27 16:47:06 <@tflink:fedora.im> sure, that can work. I think it's up to you whether you want to start drafting the change now or once the work is done 2025-03-27 16:47:10 <@tflink:fedora.im> I don't have a strong opinion on it 2025-03-27 16:47:58 <@xanderlent:fedora.im> OK, I wasn't sure what the process was. Since it's something we're doing incrementally, I think writing it up afterwards makes sense. 2025-03-27 16:48:33 <@tflink:fedora.im> there's not really a formal process that I know of until you get to the point of proposing the feature 2025-03-27 16:48:48 <@xanderlent:fedora.im> I think a lot of the dependencies are from extras that would not be packaged in Fedora, but I'll have to look into it. Thanks for the heads-up. 2025-03-27 16:49:30 <@xanderlent:fedora.im> Ok. I think that's all I have for today's meeting. 2025-03-27 16:49:51 <@tflink:fedora.im> ok, thanks for participating on top of all the work you're doing 2025-03-27 16:50:02 <@tflink:fedora.im> moving on to the last topic on the agenda for today 2025-03-27 16:50:13 <@tflink:fedora.im> !topic onnx-runtime backend 2025-03-27 16:50:21 <@tflink:fedora.im> Tom Rix: I think this is yours 2025-03-27 16:51:32 <@trix:fedora.im> so i have a PR that enable rocm backend and delivers as a seperate subpackage. does this seem like a good way to go for general adding backends to onnx-runtime ? and others more generally ? 2025-03-27 16:52:19 <@tflink:fedora.im> I don't pretend to know much about onnx but it makes sense to me 2025-03-27 16:52:41 <@dherrera:fedora.im> I'm a comaintainer of that package, I merged that PR today btw :) 2025-03-27 16:52:57 <@trix:fedora.im> oh.. question #1 answered. 2025-03-27 16:52:59 <@tflink:fedora.im> pulling in chunks of the rocm stack for folks who aren't interested in those accelerators seems like it would annoy people 2025-03-27 16:53:14 <@trix:fedora.im> yes, it does. 2025-03-27 16:53:35 <@dherrera:fedora.im> I had some ideas on alternatives on how to deliver that, but after researching they ended being inconvenient and had problems in their implementation, so I just went to merge yours :) 2025-03-27 16:54:24 <@trix:fedora.im> so next step is for me to see how connecting it to sourceextractor++ works. 2025-03-27 16:55:08 <@dherrera:fedora.im> honestly, idk how that would work either, and i'm interested since I also have a couple of packages that depend on that in COPR :) 2025-03-27 16:55:08 <@trix:fedora.im> goal would be something like a sourceextratorc++-rocm that installs to the default location 2025-03-27 16:55:36 <@trix:fedora.im> and doesn't step on anyone else. 2025-03-27 16:56:00 <@tflink:fedora.im> would alternatives be an option? 2025-03-27 16:56:10 <@trix:fedora.im> yes. 2025-03-27 16:56:15 <@man2dev:fedora.im> Rpm has too many ways to make spec work with mutiple code bases and not enough examples i think the only people that know rpm well enough about such deails are rpm devs 2025-03-27 16:56:34 <@trix:fedora.im> ๐Ÿ˜‚ 2025-03-27 16:56:46 <@trix:fedora.im> true. 2025-03-27 16:57:01 <@man2dev:fedora.im> I trired and faild at mutiple times to find way to add backends to specs thats scales 2025-03-27 16:57:26 <@trix:fedora.im> it's not an easy problem. 2025-03-27 16:58:30 <@tflink:fedora.im> !info PR to add rocm backend for onnx has been merged, will start showing up in the Fedora onnx packages 2025-03-27 16:59:14 <@dherrera:fedora.im> I was researching to do this trick on onnxruntime https://www.pixelbeat.org/docs/conflicting-rpms.html but in the end I stumbled with some optimizations that made it infeasible to use on libraries 2025-03-27 16:59:14 <@dherrera:fedora.im> This trick is used on coreutils though, so it should work for packages like sourceextratorc++-rocm 2025-03-27 17:00:48 <@trix:fedora.im> this seemed similar to alternatives, but don't know alternatives that well, just i have _heard_ of alternatives and not this approach. 2025-03-27 17:02:14 <@tflink:fedora.im> it sounds like different approaches to the same problem - one using RPM and the other using symlinks+config 2025-03-27 17:02:40 <@dherrera:fedora.im> tbh, whatever works is fine for me ^^ 2025-03-27 17:03:00 <@dherrera:fedora.im> tbh, whatever that works is fine for me ^^ 2025-03-27 17:03:20 <@tflink:fedora.im> I don't have anything insightful to add, either. just wondering if alternatives might be better to look at first since that's a more general solution 2025-03-27 17:03:37 <@tflink:fedora.im> but I also think that'll be up to whoever is doing the work :) 2025-03-27 17:04:07 <@tflink:fedora.im> is there anything else to cover on this topic? 2025-03-27 17:04:27 <@trix:fedora.im> i guess i try some stuff out and report back on list and in the next meeting. 2025-03-27 17:04:34 <@dherrera:fedora.im> https://docs.fedoraproject.org/en-US/packaging-guidelines/Alternatives/ 2025-03-27 17:04:34 <@dherrera:fedora.im> btw, alternatives you mean this, right? 2025-03-27 17:05:39 <@tflink:fedora.im> yeah, that looks right. I thought it was a general method and that doc makes it sound fedora-specific 2025-03-27 17:07:16 <@tflink:fedora.im> !info more investigation is needed before this conversation goes father, will resume the discussion at a later date once that investigation is done 2025-03-27 17:07:37 <@tflink:fedora.im> anything else here or any suggested changes to my summary attempts? 2025-03-27 17:07:48 <@trix:fedora.im> sounds good. 2025-03-27 17:08:02 <@tflink:fedora.im> ok, that's the last topic on today's agenda 2025-03-27 17:08:06 <@tflink:fedora.im> moving on to 2025-03-27 17:08:09 <@tflink:fedora.im> !topic open floor 2025-03-27 17:08:15 <@tflink:fedora.im> any other topic that folks want to bring up? 2025-03-27 17:12:44 <@tflink:fedora.im> ok, thanks for coming everyone 2025-03-27 17:12:52 <@tflink:fedora.im> !endmeeting