jetson-containers

Why we included this project

Running modern AI tooling on a Jetson edge device often means chasing CUDA versions, JetPack releases, and broken wheels. This project sidesteps that with a modular container build system that produces pre-built Docker images for Jetson, covering the usual ML libraries like PyTorch and TensorFlow plus heavier serving stacks such as vLLM, SGLang, llama.cpp, and Whisper. Instead of maintaining your own Dockerfiles and fighting ARM and CUDA toolchains, you pull a container matched to your hardware and JetPack release and get straight to work. Teams doing on-device inference, computer vision, edge LLM serving, or ROS robotics get tested images and reproducible build recipes they can rely on. The repo also works as a handy reference for which package versions work on which Jetson when you are planning a deployment.