ERROR Can not train 'TEDPolicy'. No data was provided. Skipping training of the policy

What could be the cause of this error? I do have a valid YAML file with stories.

config.yml

default pipeline

policies:
  - name: RulePolicy
  - name: TEDPolicy
    max_history: 5
    epochs: 100
    constrain_similarities: true

If anyone can help, thank you :slight_smile:

@yem why you not mentioned - name: MemoizationPolicy?

No particular reason. I’m still testing stuff. Is the MemoizationPolicy needed for the TEDPolicy to work?

@yem try mention and provide more training data for β€œnlu_fallback” please and re-train again while deleting previous models.

@nik202 Thank you for your responses. I added the MemoizationPolicy but the Can not train 'TEDPolicy'. No data was provided. Skipping training of the policy. still persists.

@yem You need to provide training data for it, if you not required that policy just comment out on the TEDPolicy and train the model.

Check more details here: Policies OR Unpacking the TED Policy in Rasa Open Source

I figured it out. Somewhere along the way I must have accidentally deleted the stories: keyword in my stories.yml file.

File format causing the error:

version: "2.0"
- story: happy path
  steps:
  - intent: greet
  - action: utter_greet
  - intent: mood_great
  - action: utter_happy

Correct format:

version: "2.0"
stories:
- story: happy path
  steps:
  - intent: greet
  - action: utter_greet
  - intent: mood_great
  - action: utter_happy

Thank you, this saved me a lot of headaches while searching! :slight_smile: I did the same by mistake

1 Like