no that's incorrect - llama.cpp has support for providing a context free grammar while sampling and only samples tokens that would conform to the grammar, rathe

no that's incorrect - llama.cpp has support for providing a context free grammar while sampling and only samples tokens that would conform to the grammar, rather than sampling tokens that would violate the grammar