When using llama-server, does sampling parameters from incoming request overrides the ones we set while starting the server?

[enesaltinkaya]

If i start the server with llama-server -m model.gguf -temp 0.5
and the incoming http request has {"temperature":1.0}
which one will be used?

And can i force it to use only command line parameters and ignore http request parameters?

Thanks.

[ggerganov]

The request's one.

And can i force it to use only command line parameters and ignore http request parameters?

No, this is not supported atm.

[enesaltinkaya]

Alright.

Thanks a lot for your hard work!