Continuous Validation Smoke Test

To ensure no future merge or Python library version update causes any regression, we should have a basic smoke test to perform the following:
* Create two small subsets S1 and S2 of dataset
* Train a model on S1
* Test the model on S2
* Cache the test score and compare to those of previous runs. Ensure scores are within rough margin of error
** To begin, fine to compare against hard-coded metrics