Why I need to wait after the training of some files and then an error ( many errors ) appear(s) and the training will repeat again?
Or does rasa repeat the training for all components if an error appears on one file for example?
What’s the specific error?
For example this error:
ActionNotFoundException: Cannot access action 'utter_utter_categories', as that name is not a registered action for this domain. Available actions are:
|| - action_listen|
|---|---|
|| - action_restart|
|| - action_session_start|
|| - action_default_fallback|
|| - action_deactivate_loop|
|| - action_revert_fallback_events|
|| - action_default_ask_affirmation|
|| - action_default_ask_rephrase|
|| - action_two_stage_fallback|
|| - action_unlikely_intent|
|| - action_back|
|| - ...|
It’s true that it would be better to check for errors before training. I myself wasted so many hours because of that lol.
Then I learned to do rasa data validate before training to make sure there’s at least this kind of error (which is the most common error while training) to stop worrying about.
Maybe you could suggest this change by opening an issue on GitHub.
rasa data validate detects many problems. But I have a case that doesn’t detect as an example:
If you add an intent under the intents keyword which has a typo, then you add a correct intent name inside the payload in button. After that running rasa data validate will not detect the typo!
If you running the shell, you receive a message UserWarning: Interpreter parsed an intent 'category_A1' which is not defined in the domain. Please make sure all intents are listed in the domain. More info at https://rasa.com/docs/rasa/domain
Maybe I can open an issue and see what will they say for that.
1 Like
Hi @Pain. Yes we need to train again and again if we are updating the code or having some errors. I guess that’s how our model refine? On the other hands, maximum error are self explanatory with the hint, what is missing, what you need to do with that error to get running the code.
As, your last post; category_A1 is not mentioned in the domain.yml file check please or sometime it contradicts with the older trained models also.
If you have updated small changes in stories or domain etc, the model train fast also 
it only focuses on the particular code update
Hi @nik202,
thanks for replying. I know that.
But in my case, it cannot detect the error as I discussed.
I opened an issue on Github for that as a feature request and I will see what happens.
I agree with you.
But here at the forum, we can try to help to solve your issues, but if the problem is with Rasa itself, better report it on GitHub 
@Pain what is your current error, can you share please?