This is GGML format quantised 4bit model of Eric Hartford's 'uncensored' training of Wizard-Vicuna 13B.
This version is quantized using older version of llama.cpp (cf348a6) so it can be used with oobabooga/text-generation-webui.