you are viewing a single comment's thread
view the rest of the comments
[–] 2 points 5 months ago (2 children)

Until now, ollama model router implementation is still better. There is still no model queue system in llama-server.

llama-server is also released around last year only.

Maybe after https://github.com/ggml-org/llama.cpp/issues/21678#issuecomment-4592982609 then there is not much reason to use ollama.

Anyway i stopped using ollama and just deal with the issue now. Because most huggingface models work better with llama.cpp.

But i dont understand why everyone just keep bashing ollama. Does no one run multiple models?

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 5 months ago

    But i dont understand why everyone just keep bashing ollama. Does no one run multiple models?

    Because for the longest time they were just a wrapper around llama.cpp and not until pushed they acknowledged the devs behind llama.cpp. They also took a lot of the spot and funding for what actually was llama.cpp features.

  • source
  • parent