Nvidia-docker on Jetson can't access the nvidia toolkit

Paddlepaddle runs without any issues on the host, but I can’t get it to run in a container (I believe because I can’t access the nvidia toolkit from inside the container). Am I doing something wrong here? Any help is much appreciated.

I can see nvidia in the list of runtimes when I run nvidia-docker info.
nvidia-docker run -it -v $PWD:/paddle --shm-size=16G --network=host --name nv_ubuntu --runtime=nvidia ubuntu /bin/bash

Host environment info:
nvidia-docker --version: Docker version 19.03.6, build 369ce74a3c
uname -r: 4.9.201-tegra
apt-cache show nvidia-jetpack | grep Version: Version: 4.5.1-b17
dpkg -l | grep nvinfer: 7.1.3-1+cuda10.2
dpkg -l | grep cudnn:

Hi @AmateurPirate, it appears that you are using the ubuntu container image. To use GPU acceleration in JetPack container, can you try running l4t-base container (or a container derived from l4t-base):

$ sudo docker run -it --rm --net=host --runtime nvidia nvcr.io/nvidia/l4t-base:r32.5.0
# python3 -c 'import tensorrt'

If you can run that python3 -c 'import tensorrt' statement inside l4t-base, that means GPU should be working for you.

1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.