Core component gets trained before NLU component

Hi, Recently I started using conveRT. I observed that Core model is getting trained before the NLU model. It’s generally the NLU model which gets trained first right? Is it ok or is there something I am doing wrong here?

My pipeline config:

language: en
  - name: ConveRTTokenizer
  - name: ConveRTFeaturizer
  - name: RegexFeaturizer
  - name: LexicalSyntacticFeaturizer
  - name: CountVectorsFeaturizer
  - name: CountVectorsFeaturizer
    analyzer: "char_wb"
    min_ngram: 1
    max_ngram: 4
  - name: DIETClassifier
    epochs: 200
  - name: EntitySynonymMapper
  - name: ResponseSelector
    retrieval_intent: smalltalk
    epochs: 200
    scale_loss: false
# Configuration for Rasa Core.
  - name: MemoizationPolicy
  - name: TEDPolicy
    max_history: 5
    epochs: 100
  - name: MappingPolicy

@Akhil This sounds normal to me. The Core model is usually trained first and the NLU second. If only NLU data is changed, then only the NLU model will train. If only Stories data is changed, then only the Core model will train

Thank you @tyd. Got it. But, I see that even if I don’t change NLU/stories, the Core model is getting trained if I run rasa train --debug.