`rasa train` vs`rasa train nlu`?

Sorry if this an obvious question, but what is the difference between running

rasa train and rasa train nlu?

I see if I run rasa train nlu, I have a new model in data that starts with nlu

Is one better than the other? Should I do both?

rasa train nlu will train the nlu component only.

rasa train will train both the nlu and core (the dialogue engine) components.

If you are only interested in using rasa for its nlu capabilities, then rasa train nlu is sufficient.

1 Like

Thank you, that’s kind of what I was thinking, but then I was following this tutorial and it has both commands in there.

In that tutorial, the author first wants to train the NLU component, so she calls rasa train nlu (Step 1). In Step 7, the author wants to train the dialogue management component, so she calls rasa train which will train both the NLU and dialogue management component.

It’s worth pointing out that if you have called rasa train nlu before and have made no changes to your nlu training data since that call, then calling rasa train will skip the NLU training and only train the dialogue management component.

1 Like