I infini-ai-lab.github.io SEQUOIA: Exact Llama2-70B on an RTX4090 with half-second per-token latency 131 points zinccat 2 years ago 61 comments
