Docker with Nvidia 390.67 + Cuda 9.1 + cudnn7 + datascience system libraries

Tested with the following Host :

• Ubuntu 18.04
• Installed Nvidia driver = 390.67

No need for cuda on the Host. It will be installed inside the docker image.

Steps

1. Install nvidia docker 2

Follow those instructions : https://github.com/NVIDIA/nvidia-docker#quickstart

2. Build the docker image

sudo docker build --build-arg http_proxy=$http_proxy -t cuda9.1 .

3. Execute jupyter within the docker

Let's say :

• your user is user
• your home is /home/user
• your jupyter path is /home/user/anaconda3/bin/jupyter

And that you already created a configuration file for jupyter with the user user :

jupyter notebook --generate-config
jupyter notebook password

Then you can launch it with:

sudo docker run -p 4242:8888 --runtime=nvidia  -v /usr/local/cuda:/usr/local/cuda -v /usr/local/cuda-9.1/:/usr/local/cuda-9.1/ -v /usr/bin/nvidia-smi:/usr/bin/nvidia-smi -v /etc/passwd:/etc/passwd -v /etc/group:/etc/group -v /etc/shadow:/etc/shadow -v /home/user:/home/user -it cuda9.1 /sbin/runuser -l user -s /bin/bash -c "/home/user/anaconda3/bin/jupyter lab --notebook-dir='/home/user' --config='/home/user/.jupyter/jupyter_notebook_config.py'"

Your Jupyter will be accessible from all the local network from http://your_ip:4242.