3 ms·

Neither of which run LLaVA as far as I can tell. Use this until that happens. Also, this is way easier to set up. No dependencies. Easy.

• v3ss0n 3y ago
  why not? both uses llamacpp as backend
  • trzy 3y ago
    llama.cpp does not have interactive mode support for LLaVA nor the web server yet. So if they're just invoking the app, there is a lot of overhead for each call until they add that in. This server doesn't have that issue.