Failed trainning with entity roles

Hi!

I’m trying to build a form (motor_form) with three slots (amount_of_potencia, amount_of_velocidad, alimentacion) and two of the three slots are numbers (amount_of_potencia, amount_of_velocidad), so I’m using entities roles as I saw in this post. But the trainning failed. I’m using rasa 2.6.2 and rasa-sdk 2.6.0.

Here’s the form

image

Here’s the trainning error

It failed when i declared the entities in this part.

image

Here’s how I declared the slots

image

I tried to update rasa to 2.8.6 but that wasn’t the problem.

I hope someone could help me :slight_smile: , thanks !!

Hi @tomimartin01 remove - before intent in first screenshoot under required_slots and also see this link: Forms

@tomimartin01 After intent you mention type but where in entity ? Further type should have -type again check the ref: Forms

I guess you only need to do a proper syntax and synchronised the flow of type, entity and intent you are good to go. Good Luck!

Hi @nik202 ! Thanks for reply me. Sorry, but I could understand your first correction → " remove - before intent in first screenshoot under required_slots and also see this link: Forms ".

Also, I belive that your second correction means this

image

I remove the ignored intents flags, because I wrote a rule for stop the form.

The problem isn’t solve, and if I comment the code of third photo of mi last post the trainning starts ok.

Thanks!

@tomimartin01 Yes, I can see you had updated the code as per the reference link I shared with you, but you did not followed my second comment, please again check the links and how they have mention the type, entity, roles and intent and then entity.

@tomimartin01 What is your third photograph reflect is it slots? if yes try mention type to text

This is the third photo, in the entities section of the domain.

image

If I comment line 326 to 329 the trainning begins,

@tomimartin01 Right, Can you please share what is the training example of these roles velocidad & potency | Please see this also for your ref: NLU Training Data

@tomimartin01 Even add more training data, in same format as shown in the ref link.