News for hackers. Stuff that matters.
Re:Show HN: Exafunction NLP – Run models like GPT-J for 6x c... by ml_hardware on Tuesday September 28, 2021 @03:10PM (#28686260)
Wow! How are you able to achieve the cost reductions? Is it different hardware, software optimizations, or both? Also does this suggest that OpenAI is charging 6x markups on their models... >:(
Re:Show HN: Exafunction NLP – Run models like GPT-J for 6x c... by varunkmohan on Tuesday September 28, 2021 @03:25PM (#28686441)
Mainly efficient usage of cloud hardware right now, with some software optimizations. Looking at more cost-effective hardware is on our roadmap too.