Switch branches/tags
Find file History
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
..
Failed to load latest commit information.
group_vars
roles
README.md
inventory
setup.yml

README.md

Setting up a host for nvidia-docker containers

You May Already Be Done

  1. You may already have nvidia-docker setup. You can test via:

Run nvidia-docker run --rm graphistry/cljs:1.1 npm test

Expected output:

> cljs@0.0.1 test /app
> ./tests/add.js

Ran
Success
  1. Check a more involved demo with visual verification:
  • Run sudo nvidia-docker run -p 80:3001 -d graphistry/cljs:1.0
  • Go to 257.257.257.257/?mode=opencl
  • You should see two pandas, one in black and white. The compute time should be < 20ms: refresh the page if not.
  • Stop the test: docker ps to get the Container ID and then docker stop MY_CONTAINER_ID

Ubuntu Xenial

  1. Follow http://docs.ansible.com/ansible/intro_installation.html#latest-releases-via-apt-ubuntu to install Ansible. (In short, sudo apt-add-repository ppa:ansible/ansible && sudo apt-get update && sudo apt-get install ansible.)

  2. Make your remove server's private key readable at ~/.ssh/nvidia-docker-host.pem.

  3. Change the contents of the file inventory to the host ip address / cname. If your IPv4 address is 257.257.257.257, echo 257.257.257.257 > inventory.

  4. Run ansible-playbook setup.yml -i inventory -e 'ansible_python_interpreter=/usr/bin/python3'. This includes some tests; if this finishes cleanly, you will have run edge detection using Javascript on the CPU and OpenCL compiled to the GPU in a container.

  5. For additional verification, try to run the cljs convolution demo above.