Conversation

Notices

1. Embed this notice
   kaia (kaia@brotka.st)'s status on Tuesday, 24-Oct-2023 15:54:38 JST kaia
   looking for GPUs feels like checking on drug prices :puniko_wtf3:
   • Embed this notice
     kaia (kaia@brotka.st)'s status on Tuesday, 24-Oct-2023 16:19:41 JST kaia

     in reply to

     • condret :verified: :ancom:

     @condret generating tomboys important AI research

   • Embed this notice
     condret :verified: :ancom: (condret@fedi.absturztau.be)'s status on Tuesday, 24-Oct-2023 16:19:50 JST condret :verified: :ancom:

     in reply to
     @kaia gpus for what purpose?
   • Embed this notice
     condret :verified: :ancom: (condret@fedi.absturztau.be)'s status on Tuesday, 24-Oct-2023 16:24:25 JST condret :verified: :ancom:

     in reply to
     @kaia as long as you don't want to heavily train models, it could work with some older gpu just fine, i believe.
     
     a friend runs some llama models on cpu.
     kaia likes this.
   • Embed this notice
     kaia (kaia@brotka.st)'s status on Tuesday, 24-Oct-2023 16:25:05 JST kaia

     in reply to

     • condret :verified: :ancom:
     @condret
     llama is CPU though. I need cold hard CUDA cores for StableDiffusion
   • Embed this notice
     kaia (kaia@brotka.st)'s status on Tuesday, 24-Oct-2023 16:39:51 JST kaia

     in reply to

     • condret :verified: :ancom:
     @condret yeah and it sucks so much :sadcat:
   • Embed this notice
     condret :verified: :ancom: (condret@fedi.absturztau.be)'s status on Tuesday, 24-Oct-2023 16:39:52 JST condret :verified: :ancom:

     in reply to

     • condret :verified: :ancom:
     @kaia i mean https://github.com/ROCm-Developer-Tools/HIPIFY exists, so you don't have to necessarily rely on cuda apis
     kaia likes this.
   • Embed this notice
     condret :verified: :ancom: (condret@fedi.absturztau.be)'s status on Tuesday, 24-Oct-2023 16:39:54 JST condret :verified: :ancom:

     in reply to
     @kaia why cuda?
   • Embed this notice
     ロミンちゃん (romin@shitposter.club)'s status on Tuesday, 24-Oct-2023 16:46:02 JST ロミンちゃん

     in reply to

     • condret :verified: :ancom:
     @kaia @condret
     >llama is CPU though
     It isn't? If you meant llama.cpp, they've supported gpus for a while.
     kaia likes this.