I’d like to play with a personal instance of Stable Diffusion. I could install it directly on this Debian GNU/Linux box, but that requires a weird dance involving installing a downgrade to Python (among other things). Anyone know of a downloadable VM from virtualboxes.org or somewhere? Otherwise I will just roll my own, but one that has been optimized and tested by many others would presumably save me some time.

Thanks.

  • brisvag
    link
    fedilink
    arrow-up
    1
    ·
    1 year ago

    No need to downgrade python, and no weird dance shoukd be needed. Use virtual environments! Conda is great if it’s your first dabbbling with virtualenvs. What other things are blicking you?

    • nitpicking@kbin.socialOP
      link
      fedilink
      arrow-up
      1
      ·
      1 year ago

      Miniconda drives me crazy and the one time I tried to use it, it changed my default python environment system-wide. I have been increasingly annoyed by Python’s having so absurdly many fractional versions, and its scripts refusing to work unless you supply them with version N.nnnn.nnn exactly. (Yes, I am exaggerating.) Here is the acceptable pace to make incompatible versions of a programming language: no more than every 2 years.

      • brisvag
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        Yup, packaging is the worst side of python. Conda having a default base environment is a huge mistake on their part, many people are burned by this constantly. I suggest you try mamba/micromamba: it’s the same, much faster, and has no base env by default.

        Btw, python hasn’t had backward incompatibe versions for more than a decade now. The problem comes from libraries being too strict or too loose with their restrictions, with at the same time no way for the interpreter to simply use different versions of the same library at once (which other languages like julia can do)