Fine tuning Tesorflow-embeddings

Rasa Open Source
1

I’m trying to tune parameters in intent_classifier_tensorflow_embedding . i read RASA-NLU documentation but no where they mentioned how to fine tune those parameters like they provided for sklearn intent classifier using gridseachcv.

Is there a way to automate parameter tuning for tensorflow embedding using Python API or Command line ?

2

@amn41 Maybe you have some suggestions? :slight_smile:

3

Sure! I’ve had some good results with hyperopt. We plan to create a proper tutorial for this, but will probably be early next year before we get to it.

1 Like
4

I’ve some doubts on defining objective function do you have any sample code for tuning tensorflow embedding parameters using hyperopt?