add link to 13B QPTQ

try turning on hosted inference

add much improved options for inference

replace GPTQ with act order false (for faster generation)

create model card

add gptq files

Upload tokenizer

initial commit