Quick Start section in README.md is a bit misleading
It shows cloning ik_llama.cpp, but then llama-server command example is for mainline llama.cpp (includes -fit that is not supported in ik_llama.cpp).
Yes you are correct. If you have an updated set of commands consolidated and labeled for both ik and mainline feel free to submit a PR!
fwiw he CLI arguments on ik_llama.cpp have changed for the Qwen3.5 MoE numerous times in the past week including -sm graph available now
i also assume people using these quants will have some idea what they are doing or be active in other areas to get the updated information, the quick start is more of a general guidance.
But yes you are correct.
I just did some updates, hopefully it is slightly less manic of a quickstart lol...
it probably requires this PR for now though on ik: https://github.com/ikawrakow/ik_llama.cpp/pull/1368 which I'm currently testing