here’s a gist for setting up a github workflow. It runs a 5-fold cross-validation on your training data, and then comments on your PR with a table of results.
This is useful for a continuous integration setup, so you can make sure your model performance doesn’t drop when you add new training data or change your NLU config.
