Because do-release-upgrade cannot work in this container.
This docker container is based on l4t-ml.
Jetson Docker Container for ROS2 Foxy with CUDA, CUDNN and TensorRT
This docker container is based on l4t-ml.
Docker Hub
https://hub.docker.com/repository/docker/tatsuyai713/jetson_ros2_foxy
please refer this. dusty-nv/jetson-containers#108
And add runtime setting for nvidia gpu
/etc/docker/daemon.json
{
"default-runtime": "nvidia", # add this line
...
...
}
Nvidia container automatically mount host libraries (CUDA, CUDNN and so on). But sometimes, this function block the updating the software inside container. So, you should remove or replace the setting file. please refer this site. https://github.com/NVIDIA/nvidia-docker/wiki/NVIDIA-Container-Runtime-on-Jetson
please keep only one file l4t.csv. /etc/nvidia-container-runtime/host-files-for-container.d/l4t.csv
- first time, you have to setup container.
./launch_container.sh setup
- after setup, you can access the inside container by same username and hostname is Docker- + Host hostname.
- next time, you can enter the inside container with only using
./launch_container.sh
- if you want to commit the change inside container, please type this command.
./launch_container.sh commit
- first of all, you have to execute "upgrade_to_20.04.sh" for upgrading the Ubuntu OS. (execute do-release-upgrade command)
- execute "setup_python3.6.sh" for installing Python3.6. after that you can use pytorch with Python3.6.
- execute "install_nvidia_packages.sh"
- execute "install_ros2_foxy.sh" if you want to use ROS2 Foxy.