This repository provides training code, pre-trained checkpoints, and deployment scripts for general-purpose goal-conditioned visual navigation policies. The models include GNM, ViNT (a visual navigation transformer foundation model), and NoMaD (a diffusion-based navigation policy). These policies can control multiple robot platforms in zero-shot and support efficient fine-tuning on new robots. The codebase includes training scripts, trajectory processing utilities, and ROS-based deployment examples for TurtleBot and LoCoBot.