On KL-Divergence and Context Size Optimization in GGUF Quantization

note: This paper was written with AI, and the exploration it describes was done collaboratively with AI. The "we" described here is us; me and a few models. With llama.cpp model quantization, properly adjusting models to keep their performance after ...