A modular container build system providing the latest AI/ML packages for NVIDIA Jetson and JetPack-L4T.
Jetson Containers is a modular Docker container build system and repository that provides pre-built, optimized machine learning and AI software stacks for NVIDIA Jetson edge computing platforms. It solves the problem of complex dependency management and cross-compilation for edge AI by offering hundreds of ready-to-use container images for frameworks like PyTorch, TensorFlow, ROS, and various LLM inference engines.
Developers, researchers, and engineers building and deploying AI, machine learning, robotics, or computer vision applications on NVIDIA Jetson devices (e.g., Orin, Nano).
It dramatically reduces setup time and complexity for Jetson development by providing a vast, curated library of compatible containers and a flexible build system, eliminating the need to manually compile dependencies or manage environment conflicts.
Machine Learning Containers for NVIDIA Jetson and JetPack-L4T
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Allows easy combination of packages like PyTorch, ROS, and Transformers into custom containers, as demonstrated in the build command example for creating tailored environments.
Includes hundreds of pre-configured packages for ML, LLMs, robotics, simulation, and more, detailed in extensive tables covering everything from vLLM to Isaac Sim.
Supports different CUDA, cuDNN, TensorRT, and Python versions via environment variables, enabling compatibility across JetPack releases without manual compilation.
Provides helper scripts like `jetson-containers` and `autotag` to simplify pulling, building, and running containers, reducing setup time and errors.
Only tested and supported for JetPack 6.2 and 7 on NVIDIA Jetson devices, making it unsuitable for other edge hardware or non-CUDA environments.
Docker images add memory and storage costs, which can be significant on resource-constrained edge devices, potentially impacting performance for lightweight deployments.
While modular, creating highly specialized containers requires deep understanding of package dependencies and build system intricacies, which can be daunting for newcomers.