While dabbling with different models, I have started to understand that there are different parts of the model & KV cache, and possibly also different kinds of layers within the model, which all can be quantised differently. Does it make sense to split this question into two parts: Dense models and
