Run AI models locally without an expensive GPU

Поделиться
HTML-код
  • Опубликовано: 25 янв 2025

Комментарии • 25

  • @alwayssmile1778
    @alwayssmile1778 2 дня назад

    Will try it surely!

  • @nufh
    @nufh День назад

    Hey, thanks for the video.

  • @BadalPradhan-h1v
    @BadalPradhan-h1v День назад +1

    Kindly make a video how to setup and make own deepseek r1 api

    • @JoeMama-k6u
      @JoeMama-k6u 2 часа назад

      It's like 0.0014 per prompt just get it lol things unbelievable I was stuck for 3 days with clide and chatgpt and cascade. All of them. Could t handle the size of my algo any more. Thjs thing did what I was trying to figure out in three days. Did it in 3 prompts

    • @zenvanriel
      @zenvanriel  2 часа назад

      Deepseek r1, definitely the full model you can access via web/api, is really difficult to host on your own hardware. You'll need to start with a smaller model like Phi-3.5/4 or a smaller Llama model.

  • @clemngene2313
    @clemngene2313 2 дня назад +1

    What's the difference between using docker and ollama? If you could give a detailed explanation, I'd like that

    • @zenvanriel
      @zenvanriel  2 дня назад +3

      The important distinction here would be that this guide shows you how to use LocalAI with Docker. Similarly, you can use Ollama with Docker. Docker is just meant to give you an easy one-command way to start an environment consistently.
      So I'm going to take the liberty to answer the difference between Ollama and LocalAI, because those are actually the systems you are interested in comparing. For large language models, Ollama is actually just as good if not even better supported by the community. You can use Ollama just fine in Docker as well: hub.docker.com/r/ollama/ollama
      However, LocalAI supports more types of AI models than just language models in one package. I aim to show more of those possibilities later on hence I focus on LocalAI. If you just want to use language models, definitely try out Ollama!

    • @clemngene2313
      @clemngene2313 2 дня назад

      @zenvanriel oh, LocalAI. I understand now. Thanks

  • @CarlosRodriguez-xk9ot
    @CarlosRodriguez-xk9ot День назад

    Awesome content, could you make a video where you customize a reasoning model It further like connecting It to a folder of PDF files as database

    • @zenvanriel
      @zenvanriel  День назад +1

      This is indeed all in the backlog of the next videos! Stay tuned.

  • @zenvanriel
    @zenvanriel  3 дня назад

    How does this model perform on your device? Let me know!

  • @zensajnani
    @zensajnani День назад +1

    nice name zen

    • @zenvanriel
      @zenvanriel  День назад +1

      We share the same first name based on your handle? If so, love it!

    • @zensajnani
      @zensajnani День назад

      @ yes we do 🤝🏼

  • @ultrasound1459
    @ultrasound1459 14 часов назад +2

    The voice is not synchronised with your video

    • @zenvanriel
      @zenvanriel  12 часов назад

      Oops, looks like it is slightly off indeed. Will make it better next time.

  • @toby645pa
    @toby645pa 12 часов назад

    I always start my prompting with the question “Was 2024 a leap year”. On simple local machines it’s telling you fairy tales. That’s disappointing. But thanks for your good explanation 👏

    • @zenvanriel
      @zenvanriel  12 часов назад

      Yeah, for local devices a RAG implementation makes more sense. I'll explain that later on.

  • @Ai-inarabic
    @Ai-inarabic День назад

    How can i contact u zen

    • @zenvanriel
      @zenvanriel  День назад

      In a few weeks I'll make sure we can all get in touch. Stay tuned.

  • @엠케이-p3p
    @엠케이-p3p 19 часов назад

    phi...ok but not that powerful, i thought there might be something i don't know in this video...i would rather use openwebui or anythingllm with ollama if you just need a chat interface

    • @zenvanriel
      @zenvanriel  11 часов назад

      That's right, I will show in later content that LocalAI allows you to use other models not just LLMs. If you just want a language model, you can use ollama

  • @CarlosRodriguez-xk9ot
    @CarlosRodriguez-xk9ot День назад

    How can you make an interface or make it receive input files?