you are viewing a single comment's thread
view the rest of the comments
[–] 4 points 2 months ago*

llama.cpp (and by extension Ollama) has a Vulkan backend now, which works fine for my usecases on AMD

  • source
  • parent