Conversation

Notices

1. ✡️ ((fuggy)) :swastika: (fuggy@skippers-bin.com)'s status on Friday, 25-Aug-2023 10:00:22 JST ✡️ ((fuggy)) :swastika:

   KoboldAI with LLaMA seems to work on ROCm fairly well too, I wasn't able to get Oobabooga to work at all though it just segfaulted

   • ✡️ ((fuggy)) :swastika: (fuggy@skippers-bin.com)'s status on Friday, 25-Aug-2023 10:13:32 JST ✡️ ((fuggy)) :swastika:

     in reply to

     Had to load some of the layers on the CPU though, I think bitsandbytes not working for some reason

   • ✡️ ((fuggy)) :swastika: (fuggy@skippers-bin.com)'s status on Friday, 25-Aug-2023 10:34:43 JST ✡️ ((fuggy)) :swastika:

     in reply to

     Yeah couldn't get that to work, I'm just going to back back to using CPP LLaMA or whatever