Missing files
#5
by CoruNethron - opened
Hello. Thank you for your great quants.
Readme mention separate CPU vs GPU quants, but there is actually only one set of files and both CPU and GPU links to the same files in the repo.
Thank you for the careful read.
The CPU and GPU recommendations share a subset of models. In general, llama.cpp IQ datatypes perform best on GPUs, while KQ datatypes are better suited for CPUs. That said, if a KQ model also performs well on GPUs, we include it in the GPU recommendations as well.
In total, the repository contains 5 IQ-dominated models and 5 KQ-dominated models.
Thank you for clarification, I'll take into account [^I]Q vs IQ in model names in addition to their distribution in readme tables. Please close the issue if it's required.
Ali93H changed discussion status to closed