Broken M quants
#2
by Artefact2 - opened
None of the medium variants in the repo work, probably because of https://github.com/ggerganov/llama.cpp/pull/4927
Could you delete/reupload these files so that users don't get confused?
I have uploaded fixed models for Q3_K_M and Q4_K_M, along with a IQ3_XXS quantization.
Artefact2 changed discussion status to closed