Right now the Docker image nvcr.io/nvidia/l4t-base:r32.5.0 has many packages missing and we have to do hacks to make it confirm to the installed Jetpack image.
Can you provide the log file, and tell me more about what the missing package are?
The request is that all packages that are available in the Jetpack are available in the Docker image. Your engineers should be able to figure this out. One such example is nvidia-l4t-multimedia
nvidia-l4t-multimedia package has historically been a platform-dependent package (i.e. it is installed from a chip-specific apt repo like https://repo.download.nvidia.com/jetson/t194 or https://repo.download.nvidia.com/jetson/t210 for Nano), which means it wasn’t able to be installed into platform-independent containers. However if you want, you could simply mount it into the container like the rest of JetPack (including CUDA/cuDNN/TensorRT/ect) is mounted into JetPack 4.x containers.
For JetPack 5.x containers, we’re migrating nvidia-l4t-multimedia to be platform-independent and be in the https://repo.download.nvidia.com/jetson/common apt repo so that it can be installed into the containers for future versions of JetPack 5.x.
The point is that I don’t want to mount anything or install any additional packages. I would like to download a docker image that is completely equivalent to a Jetpack SD card image, for example https://developer.nvidia.com/embedded/l4t/r32_release_v6.1/jetson_xavier_nx/jetson-nx-jp46-sd-card-image.zip
On JetPack 4.x, all of the CUDA Toolkit, cuDNN, TensorRT, Visionworks, ect are mounted in (via the CSV files under /etc/nvidia-container-runtime/host-files-for-container.d/)
On JetPack 5.x, these packages get installed inside the container - here is an example JetPack dockerfile doing that:
https://github.com/dusty-nv/jetson-containers/blob/master/Dockerfile.jetpack
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.