EZ-EXPERIMENTr

Running reproducible research experiments at scale

EZ-EXPERIMENTr is a toolkit under-development with the aim to help researchers run automated experiments on a cluster.

Check out the code!

• The tool is designed to dispatch many experiments as reproducible singularity containers to run on a specified cluster.
• The first iteration of this toolkit will be using the University of Melbourne IaaS (that’s where I am doing my PhD)
• The DockerFile in this repo is to setup an instance of mlflow. It currently doesn’t support using a database service.

To run the mlflow container use the following commands:

docker build -t ez-experimentr/mlflow .
docker run -p 5000:5000 ez-experimentr/mlflow

To run experiments:

# Run entry_point.sh
bash bin/entry_point.sh

Architecture

Who should use EZ-EXPERIMENTr?

• Researchers trying to scale up their experiments
• Data Scientists/ML Engineers who are looking to run many experiments at scale

Developers

• Vivek Katial