can you give more info? llama.cpp vs vllm? config? i wanna try specifically this model