Include the llama2-7b model and use gptq quantization. Ensure you run it on gpu above 8GB of VRAM.