Bonjour, I am currently facing a technical challenge in creating a Docker image on ARM architecture (an NVIDIA Jetson AGX Xavier Developer Kit device) using the ultralytics/ultralytics:latest-jetson-jetpack5 image from Ultralytics. This base image currently weighs 13 GB, with an 11 GB base. Our goal is to create a complete Docker image of approximately 22 GB, which exceeds the available capacities of an NVIDIA card. We are seeking specific recommendations, best practices, or any relevant documentation to optimize the use of the ultralytics/ultralytics:latest-jetson-jetpack5 image on ARM architecture while reducing the size of the final image. Any suggestions, advice, or documentation you could provide would be greatly appreciated to help us resolve this issue.
What device are you using? Considering SSD?
Have you had a look at this?
Alternatively you might be able to take some stuff out of the dockerfile and end up with a smaller image after build. Just some thoughts.
Hi @gmeeusen
I am using an NVIDIA Jetson AGX Xavier Developer Kit device.
@boutheina.anavid - it has a M.2 Key M connector for high speed NVMe storage.
Your 32GB eMMC memory won’t be enough in most cases.
Or an ssd doesn’t fit in your power budget?