Conversation

Notices

1. LS (lain@lain.com)'s status on Sunday, 29-Oct-2023 19:45:20 JST LS
   The interop you can do between elixir and python is actually quite amazing, you can install an erlang message handler in python and then send messages back and forth as if it was an erlang process. So you can have a GenServer start up a python SD model loader, register a message handler, then send messages to python to use that model to actually generate images, then have python message it back once it's ready.
   • entropealabs (entropealabs@mastodon.social)'s status on Sunday, 29-Oct-2023 21:09:40 JST entropealabs

     in reply to

     @lain or check out bumblebee :) https://github.com/elixir-nx/bumblebee

   • LS (lain@lain.com)'s status on Sunday, 29-Oct-2023 21:09:40 JST LS

     in reply to

     • entropealabs
     @entropealabs i tried to use that, but it doesn't seem to work very well for me.
   • entropealabs (entropealabs@mastodon.social)'s status on Sunday, 29-Oct-2023 21:12:03 JST entropealabs

     in reply to

     @lain shoot, I don't have much experience with it, but I've heard some pretty good success stories moving from Python ML to an all Elixir stack.

     LS likes this.
   • LS (lain@lain.com)'s status on Sunday, 29-Oct-2023 21:14:00 JST LS

     in reply to

     • entropealabs
     @entropealabs one issue for me is that i'm using the gpu in my macbook, and that is barely supported in python, and not at all in elixir as fast as i can see.