I’ve been using llama.cpp, gpt-llama and chatbot-ui for a while now, and I’m very happy with it. However, I’m now looking into a more stable setup using only GPU. Is this llama.cpp still still a good candidate for that?

  • Hudsonius
    link
    fedilink
    English
    arrow-up
    2
    ·
    2 years ago

    GPTQ-for-llama with ooba booga works pretty well. I’m not sure to what extent it uses CPU, but my GPU is at 100% during inference so it seems to be mainly that.

    • biaOP
      link
      fedilink
      English
      arrow-up
      1
      ·
      2 years ago

      I’ve looked at that before. Do you use it with any UI?

      • Hudsonius
        link
        fedilink
        English
        arrow-up
        2
        ·
        2 years ago

        Yea it’s called Text Generation web UI. If you check out the Ooba Booga git, it goes into good details. From what I can tell it’s based on the automatic1111 UI for stable diffusion.

        • dragonfyre13@sh.itjust.works
          link
          fedilink
          English
          arrow-up
          2
          ·
          2 years ago

          It’s using Gradio, which is what auto1111 also uses. Both of these are pretty heavy modifications/extensions that do a lot to push Gradio to it’s limits, but that’s package being used in both. Note, it also has an api (checkout the --api flag I believe), and depending on what you want to do there’s various UIs that can hook into the Text Gen Web UI (oobabooga) API in various ways.

        • biaOP
          link
          fedilink
          English
          arrow-up
          1
          ·
          edit-2
          2 years ago

          I’ve tried and failed before, and the same this time. It’s nothing but problems when i use it. It seems to be popular, so it’s frustrating.

  • gh0stcassette@lemmy.world
    link
    fedilink
    English
    arrow-up
    2
    ·
    edit-2
    2 years ago

    Llama.cpp recently added CUDA acceleration for generation (previously only ingesting the prompt was GPU accelerated), and in my experience it’s faster than GPTQ unless you can fit absolutely 100% of the model in VRAM. If literally a single layer is CPU offloaded, the performance in GPTQ immediately becomes like 30-40% worse than an equivalent CPU offload with llama.cpp

    • biaOP
      link
      fedilink
      English
      arrow-up
      1
      ·
      2 years ago

      Haven’t been able to test that out, but saw the change. Particularly interesting for my use case.

      • gh0stcassette@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        edit-2
        2 years ago

        What use case would that be?

        I can get like 8 tokens/s running 13b models in q_3_k_L quantization on my laptop, about 2.2 for 33b, and 1.5 for 65b (I bought 64gb of RAM to be able to run larger models lol). 7B was STUPID fast because the entire model fits inside my (8gb) GPU, but 7b models mostly suck (wizard-vicuna-uncensored is decent, every other one I’ve tried was Not).