• 0 Posts
  • 3 Comments
Joined 4 months ago
cake
Cake day: October 15th, 2024

help-circle

  • Arkthos@pawb.socialtoPrivacy@lemmy.worldPlease, don't!
    link
    fedilink
    English
    arrow-up
    6
    ·
    2 days ago

    You can offload them into ram. The response time gets way slower once this happens, but you can do it. I’ve run a 70b llama model on my 3060 12gb at 2 bit quantisation (I do have plenty of ram so no offloading from ram to disk at least lmao). It took like 6-7 minutes to generate replies but it did work.


  • An external cartridge processing and providing battery power seems like a much better idea than the current solution of wearing the computer on your face. A small shoulder strapped device weighing a few hundred grams with a headset with more of a BSB sort of profile would be ideal for me.

    I’m glad to see Apple experimenting with some ideas like this, not so much because I want an Apple headset, but because if it turns out to be a popular idea others will jump on board.