Orquestra Resource To Run Madness with Tequila

• uses the docker container: kottmanj/tequila-madness
• example workflows are provided in examples
• implementations follow doi.org/10.1021/acs.jpclett.0c03410
• tequila is installed on the custom image and does not need to be added as a requirement in setup.py (can be done, but then the path.append command in src/python/qemadtequila/_madness_tequila.py should be commented out)

Extending/Improving this

• make a fork of this repo
• adapt the github address in the example_workflow accordingly (so that it is now your fork)
• later make a pull request

Using this

Execute like this

qe submit workflow my_workflow.yaml

Get results

qe get workflowresults WORKFLOW-ID

Debug like this

# get information about STEP-ID and which step crashed
qe get workflow WORKFLOW-ID
# get error logs
qe get logs WORKFLOW-ID -s STEP-ID

See Orquestra docs for more.

If you are running from your own fork, adapt the corresponding parts in resource definition of the workflows (see example workflows).

TODO

• need to have tequila molecules (at least the madness ones) JSON serializable (so that we can dump and load)
• make some example workflows (run madness and then do several tq tasks in parallel with the obtained molecule)
• figure out how to make own language in orquestra, so that we can use the tequila installed in the docker container instead of re-installing (easier version control)

Dockerfile

Here you can find the Dockerfile that produces the image hosten on kottmanj/tequila-madness.
You don't need this file, but you can modify it and host your own image. Works like this:

1. Modify the Dockerfile
2. Make a docker image (takes a while)

sudo docker image -t username/imagename:version Dockerfile

3. Push the image to dockerhub

sudo docker push username/imagename:version

If you host your own image on Dockerhub, you need to change it accordingly in your workflows (see the example workflow, and replace kottmanj/madness-tequila with username/imagename:version.

The Dockerfile currently uses ubuntu. Starting from another base might be beneficial for lighter containers.