Fine-tuning policy configuration


I’m trying to find the best configuration for training my bot. I saw on the rasa website that there are some scripts for fine-tuning the policy configuration. I couldn’t find those scripts.

Could anyone direct me to more info about fine-tuning scripts?

Thanks :slight_smile:

@Aviad You can refer to this Rasa NLU in Depth: Hyperparameter Tuning . But this is for NLU model. Can you add some details about your data, maybe there are some recommendations for the patterns that you spot.

Thanks @dakshvar22! I am referring more to the Core fine-tuning. NLU model works fine, but I really want to know how to find the best combination of config parameters such as epochs, max history, attention etc.

I was wondering if rasa is already offering some ways for quick fine tuning of the Core part.

No, we don’t support anything out of the box for that.