How to use the stopSequence parameter with the Llama.cpp backend?

[synw]

I use the Llama.cpp backend and tried to use the stopSequence parameter, which is a string, but it does not seem to work as expected. Maybe I am misusing it. Question: is this parameter supposed to work like the stop one in the Python Llama.cpp lib? It is and array of strings in this lib, quote from their doc:

A list of strings to stop generation when encountered.

Example: I have a prompt that ends like this:

### Response: ```json

and I would like the LM to stop inference when it produces the ` token. I tried to set the stopSequence parameter to the ` or ``` string but it does not stop the inference. How to use the stopSequence parameter?