use llama.cpp quantization with QAT finetune

[anyinlover]

I'm finetuning a 1B model with QAT using torchtune library, but I wonder is it match to llama.cpp quantization algorithm, does anyone have some experience?