What are the rasa_nlu.train tensorflow parameters?

Hi all! I’ve been playing around with python -m rasa_nlu.train recently with my personal datasets and the models its creating work very, very well.

I know that it uses tensorflow to create the model, however I want to know exactly what the parameters are. For example, what neurons are being used (sigmoid, tanh, relu), what cost function is being used, what’s the gradient descent, how is overfitting handled, etc. How can I find out what these parameters are so I can tweak them?

Thanks for the help - I love using Rasa, I just want to understand how it’s doing what it’s doing so I can tweak it.

see policies in the documentation at https://rasa.com/docs/core/policies/ or keras_policy at https://github.com/RasaHQ/rasa_core/tree/master/rasa_core/policies

I suppose @sirusss is looking for the policies of rasa NLU not Core

This might help

This might help you http://rasa.com/docs/nlu/pipeline/#intent-classifier-tensorflow-embedding

Following are the parameters you can tweak for the classifier using the config file without changing the code

1 Like

Ooops :slight_smile:

This helps - thanks!