LocalAI LLM Testing: Can 6 Nvidia A4500's Take on the WizardLM 2 8x22b?

แชร์
ฝัง
  • เผยแพร่เมื่อ 1 ก.พ. 2025

ความคิดเห็น • 8

  • @jeffersonnunn
    @jeffersonnunn 6 หลายเดือนก่อน +1

    Wild.

  • @jackflash6377
    @jackflash6377 6 หลายเดือนก่อน +1

    Great info for a guy looking to create my own AI server for the larger models.
    What about a 3090Ti 24GB plus a 4060 16GB (or two)?

    • @RoboTFAI
      @RoboTFAI  6 หลายเดือนก่อน +2

      I don't have a 3090 in the lab, but looking to sell/trade a few 4060's for one since everyone defaults to the 3090.

  • @kazadori164
    @kazadori164 6 หลายเดือนก่อน

    server parts list please, also thise vids analyzing the performance of various hardware and models are interesting, but what do you use this ai sever to actually accomplish????

    • @RoboTFAI
      @RoboTFAI  6 หลายเดือนก่อน +1

      I touch on the different node specs in several of the videos, but have been meaning to do short videos on each of them.
      As far as usage for these nodes, depends which one...but endless.
      Mostly api backends to different applications/automations I build and play with that mimic OpenAI standards.
      Research and Development
      Daily coding tasks
      Agent/Crew/Team based projects using several models all at once - my own agent based engineering team (or is the goal...)
      Proving me wrong
      Making up things that don't exist in code
      etc, etc

  • @omarsh82
    @omarsh82 6 หลายเดือนก่อน

    What about the P100? it has a bigger bit rate.

    • @RoboTFAI
      @RoboTFAI  6 หลายเดือนก่อน

      I don't have any pascal generation cards in the lab, well an old p2000 but that wouldn't stand up to a P100 for tests

    • @omarsh82
      @omarsh82 6 หลายเดือนก่อน

      @@RoboTFAI I would love to watch a video comparing tokens per second vs memory type, if HBM2 vs GDDR