Initializing a UnionML App

UnionML ships with app templates that you can use to quickly set up a UnionML app. In this guide, you’ll learn a little bit more about the anatomy of a complete UnionML app project.

Basic App Template

Let’s create a simple app that performs hand-written digits classification. UnionML ships with a command-line interface that you can use to quickly create an app:

unionml init my_app

You should see a new directory my_app with the following structure:

my_app
├── Dockerfile        # docker image for packaging up your app for deployment
├── README.md         # app project readme
├── app.py            # app script
├── data              # directory containing sample feature data
└── requirements.txt  # python dependencies for your app

Create a Virtual Environment

Create a virtual environment for your app so that you can isolate its dependencies:

cd my_app
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

To make sure everything’s working as expected, we can run app.py as a Python script:

python app.py

Expected output

   LogisticRegression(max_iter=10000.0)
   {'train': 1.0, 'test': 0.9666666666666667}
   [6.0, 9.0, 3.0, 7.0, 2.0]

Next

Now that we’ve created our UnionML app, we can now go deeper into how it works by looking at how a Dataset object is defined.