Researchers baked 3x inference speedups directly into LLM weights — without speculative decoding

Researchers from the University of Maryland, Lawrence Livermore, Columbia and TogetherAI have developed a training technique that triples LLM inference speed without auxiliary models or infrastructure ...
GitHub

ssibb/PDF-Diff-Viewer

PDF Diff Viewer, a side-by-side, visual highlight, sync-scroll, PDF comparer, written in Python. Open source, mostly powered by PyMuPDF and Tkinter. Optional support for git diff, for a better ...