Skip to content

🐳 Docker Environments

Documentation for using inference-models in Docker containers.

Recommended for Jetson

Docker is the recommended installation method for NVIDIA Jetson devices. See Hardware Compatibility for details.

Work in Progress

Docker builds for inference-models are currently in progress. Some builds are automated, while others are manual. We encourage you to test these images and raise issues if you encounter any problems.

📦 Available Docker Images

Pre-built experimental Docker images are available on Docker Hub under the roboflow/inference-exp repository.

x86_64 / AMD64 Images

Image Tag Base CUDA Version Status
roboflow/inference-exp:cpu-latest
roboflow/inference-exp:cpu-<version>		 Ubuntu 22.04 N/A (CPU only) 🤖 Automated
roboflow/inference-exp:cu118-latest
roboflow/inference-exp:cu118-<version>		 Ubuntu 22.04 CUDA 11.8 🤖 Automated
roboflow/inference-exp:cu124-latest
roboflow/inference-exp:cu124-<version>		 Ubuntu 22.04 CUDA 12.4 🤖 Automated
roboflow/inference-exp:cu126-latest
roboflow/inference-exp:cu126-<version>		 Ubuntu 22.04 CUDA 12.6 🤖 Automated
roboflow/inference-exp:cu128-latest
roboflow/inference-exp:cu128-<version>		 Ubuntu 22.04 CUDA 12.8 🤖 Automated

Image Tags

  • -latest tags point to the most recent release
  • -<version> tags (e.g., cpu-0.17.3) pin to a specific version for reproducibility

Jetson Images

Image Tag JetPack Version Status
roboflow/roboflow-inference-server-jetson-5.1.1:0.62.5-experimental JetPack 5.1 (L4T 35.2.1) ✋ Manual (experimental)
roboflow/inference-exp:jp61-* JetPack 6.1 (L4T 36.x) 🚧 In development

Image Status

  • 🤖 Automated - Built automatically on releases and main branch pushes
  • ✋ Manual - Built manually; not part of automated pipeline
  • 🚧 In development - Coming soon

🚀 Quick Start

CPU-only

docker run -it roboflow/inference-exp:cpu-latest python3

GPU (CUDA 12.8)

docker run --gpus all -it roboflow/inference-exp:cu128-latest python3

Jetson (JetPack 5.1 - Experimental)

docker run -it \
  --runtime nvidia \
  roboflow/roboflow-inference-server-jetson-5.1.1:0.62.5-experimental \
  python3

Jetson 5.1 Experimental Build

This is an experimental build that works but is not part of the automated pipeline. Future updates will be delivered later.

🔨 Building Custom Images

You can build your own Docker images using the Dockerfiles in the inference_models/dockerfiles/ directory:

💡 Usage Tips

Mounting Volumes

Mount your model cache for cache persistency:

docker run -it \
  -v ~/.cache/inference:/root/.cache/inference \
  roboflow/inference-exp:cpu-latest \
  bash

🐛 Reporting Issues

These Docker images are experimental. If you encounter issues:

  1. Check the Hardware Compatibility guide
  2. Verify your Docker and NVIDIA runtime setup
  3. Open an issue with:
  4. Image tag used
  5. Error message
  6. Steps to reproduce

🚀 Next Steps