Yet another reminder why you should not use Ollama (github.com)

and use the original llama.cpp directly. Its infinitely more easy to setup and use now

Georgi's relevant comment: https://github.com/ggml-org/llama.cpp/pull/19324#issuecommen...