dusty-nv/jetson-containers

Provides a modular, layer-based container build system that bundles CUDA-optimized ML frameworks (PyTorch, TensorFlow, JAX), LLM inference engines (vLLM, Ollama, llama.cpp), vision models (LLaVA, SAM, CLIP), and robotics/simulation tools—all pre-configured for Jetson's resource constraints. Integrates with NVIDIA's L4T base images and leverages TensorRT for inference optimization across diverse edge AI workloads from perception to embodied AI applications.

4,465 stars. Actively maintained with 81 commits in the last 30 days.