make it clear you need to make a chat template jinja file

upload smol-IQ4_KSS and post all perplexity data

Upload folder using huggingface_hub

Add AesSedai IQ3_XXS which has good perplexity!

add tensor parallel example for 2x48 GB GPUs

add ggml-org Q4_K_M to perplexity graph

add link to working chat template

Upload folder using huggingface_hub

uploading smol-IQ2_KS

update perplexity graph with official version

update perplexity graph with experimental data

add some perplexity data

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload imatrix-Step-3.5-Flash-BF16.dat with huggingface_hub

working now on ik_llama.cpp

Upload folder using huggingface_hub

add note and upload mainline compatible IQ4_XS

add IQ4_XS only for testing for now

redoing convert

initial commit