LLM inference, visually explained: why a 13B model reads 26 GB per token

 (youtube.com)

4 points | by ryansakhala 8 hours ago ago

1 comments