ferrarodav / docker-scientific

A docker container to run experiment with Sacred tool and scientific packages

Experiment with Data Science

The Docker container is based on the python:3.6 container and has many packages to do numerical computation and visualize data: Numpy, Cython, OpenCV (and opencv-python), Matplotlib, Pandas, Seaborn, Scipy, Scikit-learn, Scikit-image, Jupyter (and the extensions configurator), Xgboost, H5py, Requests, Pillow, Networkx.

It also has Sacred, a tool to help you configure, organize, log and reproduce experiments developed at IDSIA. The packages needed to keep track of the code (git), use a mongo database, yaml configuration and a telegram bot notifier are included.

Jupyter is run on port 8080.

Run with database and omniboard

Omniboard is a web dashboard for the Sacred machine learning experiment management tool. You can download the docker-compose.yml file in this repository to run the Jupyter server, the database and omniboard all together.

It needs two environment variables:

• CODE: the folder to mount in the Jupyter server
• DB_NAME: the name of the database that omniboard should look to.

Here an example:

export CODE="/ABSOLUTE/PATH/TO/CODE/FOLDER"
export DB_NAME="EXPERIMENT"
docker swarm init
docker stack deploy -c docker-compose.yml STACKNAME

Run Jupyter server alone

You can run it directly:

docker pull ferrarodav/scientific
docker run -p 8080:8080 -v /PATH/TO/FOLDER:/home/workspace --name CONTAINERNAME ferrarodav/scientific