• 0 Posts
  • 10 Comments
Joined 1 year ago
cake
Cake day: December 14th, 2023

help-circle

  • That seems kind of like pointing to reverse engineering communities and saying that binaries are the preferred format because of how much they can do. Sure you can modify finished models a lot, but what you can do with just pre trained weights vs being able to replicate the final training or changing training parameters is just an entirely different beast.

    There’s a reason why the OSI stipulates that code and parameters used to train is considered part of the “source” that should be released in order to count as an open source model.

    You’re free to disagree with me and the OSI though, it’s not like there’s 1 true authority on what open source means. If a game that is highly modifiable and moddable despite the source code not being available counts as open source to you because there are entire communities successfully modding it, then all the more power to you.


  • It’s worth noting that OpenR1 have themselves said that DeepSeek didn’t release any code for training the models, nor any of the crucial hyperparameters used. So even if you did have suitable training data, you wouldn’t be able to replicate it without re-discovering what they did.

    OSI specifically makes a carve-out that allows models to be considered “open source” under their open source AI definition without providing the training data, so when it comes to AI, open source is really about providing the code that kicks off training, checkpoints if used, and details about training data curation so that a comparable dataset can be compiled for replicating the results.


  • It really comes down to this part of the “Open Source” definition:

    The source code [released] must be the preferred form in which a programmer would modify the program

    A compiled binary is not the format in which a programmer would prefer to modify the program - it’s much preferred to have the text file which you can edit in a text editor. Just because it’s possible to reverse engineer the binary and make changes by patching bytes doesn’t make it count. Any programmer would much rather have the source file instead.

    Similarly, the released weights of an AI model are not easy to modify, and are not the “preferred format” that the internal programmers use to make changes to the AI mode. They typically are making changes to the code that does the training and making changes to the training dataset. So for the purpose of calling an AI “open source”, the training code and data used to produce the weights are considered the “preferred format”, and is what needs to be released for it to really be open source. Internal engineers also typically use training checkpoints, so that they can roll back the model and redo some of the later training steps without redoing all training from the beginning - this is also considered part of the preferred format if it’s used.

    OpenR1, which is attempting to recreate R1, notes: No training code was released by DeepSeek, so it is unknown which hyperparameters work best and how they differ across different model families and scales.

    I would call “open weights” models actually just “self hostable” models instead of open source.


  • BakedCatboy@lemmy.mltoMemes@lemmy.mlThanks dad
    link
    fedilink
    English
    arrow-up
    3
    ·
    7 months ago

    The creation tool also just lets you save the iso - but for some reason the media creation tool gives you a different iso than if you spoofed a non-windows user agent on the windows download website so that it gives you a direct link to the iso instead of getting you to install the creation tool. And for some reason only one of them worked with DISM to repair my system in order to be able to run windows update successfully.


  • BakedCatboy@lemmy.mltoMemes@lemmy.mlThanks dad
    link
    fedilink
    English
    arrow-up
    17
    ·
    7 months ago

    Also add DISM to that - if it’s corrupted it could cause you to be unable to install windows updates.

    DISM /Online /Cleanup-Image /CheckHealth

    It’s also possible that fixing it may require the original windows installation media matching your windows version build number - which means if you’ve since installed a major windows update there may not be an available installation iso that meets the requirement. Happened to me and I was lucky enough to coax an iso out of the windows installer download page that satisfied the repair tool (oddly enough downloading an iso using the Windows Media creation tool didn’t work, but spoofing Linux on the page to directly download an iso gave me a different iso that worked)

    These are the things I do to maintain my last windows machine.




  • I think that text is from melroy, so according to him. From seeing his interactions in the kbin issue tracker I get somewhat of an egotistical impression of him, because he would often take an issue that has just been opened and not triaged or discussed what the best fix is, and he would open a PR with how he thinks it should be fixed, and it sounds like his frustration is that his hasty PRs weren’t getting merged quickly because people wanted to come to a consensus.

    Maybe I’m just reading into it but it felt like he just wanted his name on something and it wasn’t happening with kbin.

    Edit: I want to add that I don’t mean to shit on him as a dev or as a person - it’s possible that I’ve only seen a one-sided view of his interactions as a busy contributor who just wants to whittle down the issue list as fast as possible and that he’s got good intentions, and regardless he seems like a very capable dev. It’s just that based on my perusing of issues and discussions I’ve come across, it doesn’t seem fun to work with him to contribute, and if I were to treat the contributors list as a scoreboard and had the goal of having my name on as many commits as possible, I think it would be hard to tell us apart. I was just going to keep my thoughts about this to myself but I’ve seen some other people comment similar things in other threads about mbin so maybe it’s worth sharing my skepticism about mbin. Take from it what you will.