Guide or Script to “tune” args/parameters to main?

[Free-Radical]

As a noob to ML/AI I am not finding good examples (or simple explanations) showing what values to set the Args/parameters (top_p, ctx, etc) to main.
I have read the doc but it's really not explanatory enough for me to tune and run the different llms efficiently on my hardware.
And from the questions I have seen, many people are having the same problem.

Is there a decent comprehensive guide with examples somewhere that explains how to pass parameters to main per model and hardware? or perhaps a script that does an "auto-tune" ?

If not, someone who can do this ( I would be very happy to help) really needs to setup such a resource. It will be a tremendous benefit to promoting "edge" inference.
Look forward to responses. Thank you everyone and especially @ggerganov for your excellent work.