hugging-apps/nvidia-cosmos3-edge
6
Demo of nvidia/Cosmos3-Edge, the 4B Reasoner tower of NVIDIA's omnimodal Cosmos 3 world foundation model for Physical AI. Give it a text instruction plus an image or video and it plans, describes, localizes, and reasons about motion, causality, and spatial relationships — with an optional chain-of-thought — via the Transformers Cosmos3EdgeForConditionalGeneration (AutoModelForImageTextToText) integration.
