Rasa train takes too long


I am building an application with chatbot as a core component. As the domain is vast, the core data and nlu has to be constantly changed , however to test the new changes(sometimes only domain.yml) I need to run the rasa train to create new training model. Every time I run the command (rasa train) it takes long time even thought there are existing models.

Could you please suggest is there is any other commands or alternative for faster training which will help in accelerating the development?

rasa 2.1.1 Config.yml language: en pipeline:

  • name: WhitespaceTokenizer
  • name: CRFEntityExtractor
  • name: EntitySynonymMapper
  • name: CountVectorsFeaturizer

token_pattern: (?u)\b\w+\b

  • name: SpacyNLP model: en_core_web_md case_sensitive: false
  • name: DIETClassifier epochs: 100
  • name: FallbackClassifier threshold: 0.4 ambiguity_threshold: 0.1
  • name: EntitySynonymMapper dimensions:
  • number


  • name: TEDPolicy epochs: 100
  • name: AugmentedMemoizationPolicy
  • name: RulePolicy

Hi @raviaradhi. How many training examples do you have in your nlu.yml file as well as stories.yml?

Hi @Juste, Below are the detials
NLU : distributed among 6 yml files with 40 Intents with average of 30 examples each, Stories: stories: distributed among 6 yml files, 51 stories

Please let me know if any other details required

Any inputs on my earlier query or is it expected to take so long for train command to run.