Make a new dataset like those found in for example ucimlr/ By subclassing either ucimlr.classification_datasets.ClassificationDataset or ucimlr.regression_datasets.RegressionDataset, it will automatically be added to the lists of dataset and be tested. Read the superclass definition to understand what the subclass needs to implement.


Add a short version of the dataset (~100 rows) to dataset_test_stubs to allow automatic testing in the cloud without repeated downloads from UCI's servers. For classification datasets, make sure all classes are present with multiple examples each as this will be tested. Stage and commit this file.

Documentation generation

First install the document generation tool:

pip install git+

Then run


Inspect documentation by running:

mkdocs serve