local install neural network model

I need to use neural network completely out of the realm of internet and it doesnt seem there are any helpful tutorials how to achieve this?

So how to do it?

I would give the AI a lot of locally stored pictures to process and then it should output stuff it lelarned from those pictures.

It would be stable diffusion but trained by my own set instead of being readily trained.

Beware Cat Shirt $21.68

Rise, Grind, Banana Find Shirt $21.68

Beware Cat Shirt $21.68

  1. 2 weeks ago
    Anonymous

    LoRAs are what you're looking for, and yes it works for SD. There is literally a whole general dedicated to this that could have told you that.

    • 2 weeks ago
      Anonymous

      https://i.imgur.com/vFYNR6i.jpg

      I need to use neural network completely out of the realm of internet and it doesnt seem there are any helpful tutorials how to achieve this?

      So how to do it?

      I would give the AI a lot of locally stored pictures to process and then it should output stuff it lelarned from those pictures.

      It would be stable diffusion but trained by my own set instead of being readily trained.

      [...]
      >/lmg/ - A general dedicated to the discussion and development of local language models.
      >FAQ: https://wikia.schneedc.com
      >►Glossary: https://archive.today/E013q
      >►Links: https://rentry.org/LocalModelsLinks
      >►Official /lmg/ card: https://files.catbox.moe/cbclyf.png

      >►Getting Started
      >https://rentry.org/llama-mini-guide
      >https://rentry.org/8-step-llm-guide
      >https://rentry.org/llama_v2_sillytavern
      >https://rentry.org/lmg-spoonfeed-guide

      https://justine.lol/oneliners/
      You may find this useful.

      I was sceptical but it really does seem to work

      Gotta put it in a more powerful computer though

  2. 2 weeks ago
    Anonymous

    I will never know what happiness feels like.

  3. 2 weeks ago
    Anonymous

    tons of options. If you know a bit of python or julia, you could also train your own and tinker with the parameters. check out the pytorch tutorials

    • 2 weeks ago
      Anonymous

      >train your own
      LMAO

      • 2 weeks ago
        Anonymous

        I will never know what happiness feels like.

        so I have to download some 40 gigabyte model and run that on command line?

        but what if I want to create my own model from scratch without internet resources?

        and run self trained LLM from command line using this program

        https://justine.lol/oneliners/
        You may find this useful.

        • 2 weeks ago
          Anonymous

          From scratch you can't unless you've got a datacenter full of Nvidia GPUs. You can train a LoRA using a single GPU though. There are plenty of guides on YT and online, mostly about Khoya-SS. You'll need enough good quality images (don't believe people saying 10 images are enough) and some time to write the captions, unless the auto-captioners work well for your images (Qwen VL Max works well for mine, but it's not a local model, sadly. I haven't tried Qwen VL, which can be downloaded). Stable diffusion is less than 12 go.

          • 2 weeks ago
            Anonymous

            is there any 5 gigabyte model for starters? when I see system is working I could get a 40 gigabyte LLM later

          • 2 weeks ago
            Anonymous

            Oh you want a text generator? Then try Mistral 7B. It's a bit more than 4gb https://huggingface.co/TheBloke/Mistral-7B-v0.1-GGUF
            I haven't tried the fine-tuned version of this model.

          • 2 weeks ago
            Anonymous

            no it needs to be a picture generator but it could be pretty much nonsense in the beginning because I want to test how to make these things work and cant download 40 gigabytes right now

          • 2 weeks ago
            Anonymous

            listen to the anons that are legit trying to answer your question. LoRAs and SD are your best bet, and no you're not going to be able to make your own SD without internet resources or a shitload of GPUs. That's not how this stuff works, kid. Even if you did have the resources to make your own, you would still need to download the frameworks, and libraries and everything else necessary to make them. This is not happening without some degree of internet access. You don't need the internet to run the models, but you require the internet to make or acquire them. If you don't have at least this much it's a no-go and you aren't going to get a different answer unless someone is trolling you.

          • 2 weeks ago
            Anonymous

            why does IQfy always post pictures of men?

            From scratch you can't unless you've got a datacenter full of Nvidia GPUs. You can train a LoRA using a single GPU though. There are plenty of guides on YT and online, mostly about Khoya-SS. You'll need enough good quality images (don't believe people saying 10 images are enough) and some time to write the captions, unless the auto-captioners work well for your images (Qwen VL Max works well for mine, but it's not a local model, sadly. I haven't tried Qwen VL, which can be downloaded). Stable diffusion is less than 12 go.

            https://justine.lol/oneliners/
            You may find this useful.

            at least this command line tool which enables local runs seem to work

            havent downloaded any model yet because this tool already was 4GB, yikes

            but its a start and now I know the program itself seem to work

            I just renamed it into .exe file and started it in Windows

          • 2 weeks ago
            Anonymous

            If that's llava, the models are going to be pretty big, bro. llava is a vision encoder built on top of vicuna, and it is neither small nor particularly resource unintensive.

          • 2 weeks ago
            Anonymous

            yeah I am testing whatever because I havent run AI systems before

            looks like it works in the first place so I can then move onto some more suitable model, this was a test

            I wasnt sure I could have anything working at all in Windows 10 from the year 2015 (I cant use more recent version)

  4. 2 weeks ago
    sage

    [...]

    >/lmg/ - A general dedicated to the discussion and development of local language models.
    >FAQ: https://wikia.schneedc.com
    >►Glossary: https://archive.today/E013q
    >►Links: https://rentry.org/LocalModelsLinks
    >►Official /lmg/ card: https://files.catbox.moe/cbclyf.png

    >►Getting Started
    >https://rentry.org/llama-mini-guide
    >https://rentry.org/8-step-llm-guide
    >https://rentry.org/llama_v2_sillytavern
    >https://rentry.org/lmg-spoonfeed-guide

  5. 2 weeks ago
    Anonymous

    https://justine.lol/oneliners/
    You may find this useful.

  6. 2 weeks ago
    Anonymous

    why does IQfy always post pictures of men?

  7. 2 weeks ago
    Anonymous

    Give me the source immediately.

    • 2 weeks ago
      South Delhi

      sk8bae

      • 2 weeks ago
        Anonymous

        how did you figure it out from just one picture with a weird pose

Your email address will not be published. Required fields are marked *