What is jetson-containers?

Nvidia Jetson devices are small, energy-efficient computers built around an Nvidia GPU. They are popular for running AI workloads at the edge, meaning on a device in the real world rather than in a data center. Getting AI and machine learning software installed on Jetson devices is often difficult because these libraries are compiled specifically for the hardware architecture and the combination of the Linux operating system and Jetson firmware they ship with. This repository provides a system for building Docker containers that have these libraries pre-configured and ready to run on Jetson hardware. Docker containers are self-contained software packages that include an application and everything it needs to run. By using a container, you avoid the complicated process of compiling each library from source and resolving dependency conflicts yourself. The collection covers a wide range of AI software categories. For machine learning frameworks there are containers for PyTorch, TensorFlow, and JAX. For running large language models locally there are containers for tools like llama.cpp, Ollama, vLLM, and several others. For vision-language models that can understand images and text together there are LLaVA and VILA among others. For robotics applications there is ROS and several robot learning frameworks. There are also containers for simulation environments, computer vision tools, and vector databases used in retrieval-augmented generation setups. The project is maintained by a Nvidia engineer and is tied to an accompanying website called Jetson AI Lab, which provides tutorials and benchmarks. Installation starts with a setup script that configures the system and installs a command-line tool for building and running any of the available containers. This is primarily useful for researchers, engineers, or hobbyists who have Jetson hardware and want to run AI experiments or build applications without manually dealing with library compilation. The containers save significant setup time and provide tested combinations of software versions known to work together on Jetson.