docker_instruction.md

Docker instruction

You need a GPU for installing the docker container. If you don't have docker installed:

• Install docker: https://docs.docker.com/engine/install/ubuntu/
• Create docker user:
  • sudo groupadd docker
  • sudo usermod -aG docker $USER
  • newgrp docker
• Download NVIDIA Container Toolkit:
  • sudo apt-get install -y nvidia-container-toolkit
  • sudo nvidia-ctk runtime configure --runtime=docker
  • sudo systemctl restart docker

If docker is already installed, you can proceed by downloading the docker image: docker pull vargas95/opensim:opensim-tf

To run the container (make sure that you are in the folder that contains dlc-docker file):

GPU=0 bash ./dlc-docker run -d -p 2359:8888 --cpuset-cpus="0-15" -v /media1:/media1 --name containername vargas95/opensim:opensim-tf

You can modify the previous command by:

• changing port: (i.e. 2359 can be 777, etc)
• change which GPU to use (check which GPU you want to use in the terminal by running nvidia-smi)
• change the name: --name containername can be anything you want
• change the volume: -v /media1:/media1 can be changed to any volume

Enter the container via the terminal (to get terminal access in container):

docker exec --user $USER -it containername /bin/bash

Jupyter and docker

You can access Jupyter by going to the port you specified (e.g. http://localhost:2359) in Google Chrome.

To get the token for entry, back in the terminal, look at the docker log:

docker logs containername 

Copy and paste the value after "token=".