• 6 Posts
  • 294 Comments
Joined 2 years ago
cake
Cake day: December 31st, 2023

help-circle

  • I’m surprised that you’re talking about models being CUDA-specific or AMD-specific. I’ve had a bunch of models running on my amd-only pc, using ollama, lemonade, and lm-studio, through either rocm or vulkan. None of these models were billed as AMD-specific. I had to do some config tweaking for ollama to use my graphics card but that’s more because I have a weird in-between-generations card that also predates the LLM hype (6700XT).

    However, I did generally need to look for the GGUF format versions of things - usually accounts like unsloth have them uploaded on huggingface barely a day or two after the original version gets posted.







  • I agree with most of what you’re saying, but:

    • forgejo is working on federation. They’ve been working on it for a certain amount of time by now, but I do think we can expect some concrete version of what you describe in terms of community to materialize in the next decade as long a people want it and are motivated enough

    • when talking about code that is stored in a version control software that supports decentralized state (git, mercurial are the 2 I have working knowledge of) the “easy” fix for low bus factors is to just fork/mirror the software you want to see continue to exist. Source code is not that voluminous, I would be surprised if [the collective we] can’t manage to store multiple copies of the sources for software we deem useful. It’s a question of changing habits, not finding some miracle tech

    Of course, habits aren’t necessarily easy to change.







  • You have all my sympathies. Someone in another post/thread brought up the idea of a support group for burned out devs/tech workers in general. I definitely think there’s something between that and unionization that is both needed and starting to be possible. Heck, even in the hackernews comments for this article there was at least one person telling another “welcome to luddism!” as both resonated with the spirit of the article itself.


  • That’s wild. Your managers’ reaction to “the project made by AI has created 2-4 years of work by experienced engineers, perhaps up to 6 of them, before it’s ready” was “why don’t you use more AI??”?

    I’m starting to think Mao had a point when he sent the business owners to do farm work. Barring a revolution, I can only hope the effective cost of inference rises du much as to make these dipshits back off from wanting it to do all the labor ever.


  • I don’t know about heavier vehicles like vans or trucks, but in my parent’s Renault Zoé the Regen braking is strong enough to slow the car down from like 50km/h to 30km/h when going downhill. It might be enough to bring the car to a standstill, I’ve never actually tried letting it be - usually there’s a car behind me or I need to get somewhere in time so I can’t afford to experiment.

    Brakes are still important for emergency/manual speed adjustments, of course. Just wanted to share my experience with “how well does regen braking work downhill?”