How to disable prompt cache?

[mscheong01]

Hi, I'm running some offline inference benchmarks using llama-cpp-python, and the prompt cache that was implemented here (#158) is getting in the way of measuring prompt evaluation time. Is there an option to disable it?

[mscheong01]

I was able to do this by calling model.reset()