I’ve looked everywhere for this answer and cannot find it anywhere.
When do I need to retrain? Say I add a slot to the domain, or add new response types through interactive learning… I’ve noticed those new response types don’t appear in the Rasa X dropdown after I add a new one, is that because it needs a retrain? Does that mean I need to retrain after each story I add that adds a new intent or response type?
Sorry if this is a simple answer but I’ve spent 20 minutes trying to find an answer… a lot of the documentation seems like it’s written for people who have a really firm grasp of all these concepts already. I may well have seen the answer and just not understood it.
@UnbridledGames No worries, your question is very generic, and let me try to help you to understand.
For Rasa open source:
You always need to train the bot, if you had updated any single code in the rasa server which is running on port 5005, i.e if you have changed in domain.yml, nlu.yml, stories.yml, or rules.yml. So, you need to train the bot every time, if you are using Rasa 3.X the bot will train fast as rasa 3.x using cache now which helps in training. On the other hand, If you are using Rasa 2.x it will take a regular time as depends on the changes you have made in the code.
Command:rasa train --debug
If you have updated any code in the custom action server which is running on port 5055, again you need to run the action server every time. It will not take much time for running.
Command:rasa run actions --debug
For Rasa X:
If you have updated anything whilst using Rasa x you need to commit to GitHub and then need to train, every time you have updated anything, make sure all the changes reflect in Github and the same in Rasa X.
I hope this will help you to understand the concept of re-train. Good Luck!