Top New Best Ask Show Jobs

Full F16 precision 34B Code Llama at >20T/s on M2 Ultra | Hacker News Reader

Full F16 precision 34B Code Llama at >20T/s on M2 Ultra

twitter.com

17 points·Jimmc414··2 comments

Open articleView on HN

https://nitter.net/ggerganov/status/1697262700165013689 (full thread without requiring login)

https://github.com/ggerganov/llama.cpp/pull/2926

Reply on news.ycombinator.com