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

floret: fastText + Bloom embeddings for compact, full-coverage vectors with spaCy

floret is an extended version of fastText that can produce word representations for any word from a compact vector table. It combines: fastText's subwords to provide embeddings for any word Bloom ...
IEEE

Loop detection and automated route aggregation in distance vector routing

Abstract: In this paper, a distance vector routing algorithm for IP computer networks is presented which is able to aggregate routes automatically and detect and prevent routing loops even on the ...
GitHub

Flash LiDAR Vehicle Detection with MATLAB Dataset and Deep Learning Benchmarks

This repository provides code and workflows to test several state-of-the-art vehicle detection deep learning algorithms —including YOLOX, SalsaNext, RandLA-Net, and VoxelRCNN— on a Flash Lidar dataset ...
IEEE

An example of modeling local loop transfer function in DSL environment

Abstract: This paper shows an example of estimation and calculation of transfer function of twisted pair channel in DSL environment, from provider point of view, instead of commonly used manufacturers ...