Wrote an essay on what Q, K, and V matrices are and the roles they play in LLMs. Since I’m on a spree to understand LLMs, I dug deeper into the nuances and put together this distilled yet detailed version. This essay also supplements my How LLM Inference Works write-up.
I’m no expert, but if you’re looking for a starting point with LLMs, this would help. Give it a read.