Conversation

Notices

1. Embed this notice
   Johnny Peligro (mischievoustomato@varishangout.net)'s status on Monday, 19-Sep-2022 21:57:39 JST Johnny Peligro

   in reply to

   • iced depresso
   • Coyote
   @icedquinn @Coyote i think that only nvidia is supported
   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 19-Sep-2022 21:57:40 JST iced depresso

     in reply to

     • Coyote
     @Coyote i have amd but its like 14gb vram
   • Embed this notice
     Coyote (coyote@social.singing.dog)'s status on Monday, 19-Sep-2022 21:57:49 JST Coyote

     in reply to

     • iced depresso

     @icedquinn It’s pretty easy on Linux. Git clone the code repo, download the model weights (you’ll need a hugging face account), set up the conda environment by following the README, and call the txt2img.py script. There’s a little more involved if you want to disable the watermark or NSFW filter, and you’ll need an Nvidia GPU with a bunch of VRAM to actually run the thing (I can barely run it with 8 GBs and that’s at half precision). I’m also pretty sure some people have made easier to setup forks, and I think someone has even made a Windows GUI program for it, too.

   • Embed this notice
     iced depresso (icedquinn@blob.cat)'s status on Monday, 19-Sep-2022 21:57:51 JST iced depresso
     pondering how difficult it is to install stable diffusion to generate some stuff for the test project