3 ms·Puma: Secure Inference of LLaMA-7B in Five Minutes4 points by tosh 3y agovillgax 3y agoWay worse performance than llama.cpp based LLM deployments