??????? #tensors

Re-quantizing a local LLM 14x faster by skipping the tensors that didn't change

4 points, 0 comments on Hacker News