Base model: https://huggingface.co/lmsys/vicuna-13b-v1.5-16k

Note that running this model with llama.cpp requires passing in non-default RoPE scaling parameters. For example:

-c 16384 --rope-freq-base 10000 --rope-freq-scale 0.25  

Note that fully offloading this model may require more VRAM than expected from the weight file sizes alone.