Creating the project#

Wine classification example#

To demonstrate the essential elements of a Union project, we will start with a simple model training workflow called wine-classification. It consists of three steps:

  1. Get the classic wine dataset using scikit-learn.

  2. Process the data that simplifies the 3-class prediction problem into a binary classification problem by consolidating class labels 1 and 2 into a single class.

  3. Train a LogisticRegression model to learn a binary classifier.

Create the project using unionai init#

We will use the unionai (the CLI tool that ships with unionai) to quickly initialize the project, from a template. The wine-classification example is among the installable examples published in the GitHub repository flyteorg/flytekit-python-template.

Install the example, and cd into it:

[~]:wine-classification
$ unionai init --template wine-classification wine-classification

[~]:wine-classification
$ cd wine-classification

[~/wine-classification]:wine-classification
$

Note

If you need to use a Dockerfile in your project instead of ImageSpec, you can use the Dockerfile template:

unionai init --template basic-template-dockerfile my_project

Project structure#

If you examine the wine-classification directory you’ll see the following file structure:

[~/wine-classification]:wine-classification
$ tree
.
├── LICENSE
├── README.md
├── local-requirements.txt
├── image-requirements.txt
└── workflows
    ├── __init__.py
    └── example.py

Note

You can create your own conventions and file structure for your Union projects. The unionai init command just provides a good starting point.

Install the local dependencies#

We will explain the significance of the two requirements files later. For now, you just need to install the local dependencies. Make sure that you have activated your wine-classification conda environment, then install the dependencies from local-requirements.txt:

[~/wine-classification]:wine-classification
$ pip install -r local-requirements.txt