A complete Llama2 inference engine that fits in 1356 bytes of x86 assemblygithub.com27 points·monax··0 commentsOpen articleSaveView on HN