Help needed with Bot core training

Rasa Open Source
1

RASA Version: 1.1.5

Tried trainnig the model using the train function as below.

model_path = train(domain_filepath, config_filepath, data_filepath, models_filepath, {‘augmentation_factor’:40})

Intent recognistion works fine. But the core training doesnt seem to be complete. I get the responses for only the first message and then fallback is triggered for all the messages.

policies:

  • name: “MemoizationPolicy” max_history: 7
  • name: “KerasPolicy”
  • name: “FallbackPolicy” nlu_threshold: 0.5 core_threshold: 0.3 fallback_action_name: “action_fallback”
  • name: “FormPolicy”

I’m using the same training data with the same policies for another bot with RASA Nlu: 0.13.7 RASA Core: 0.12.1 and everything works fine there. Any suggestions would be really helpful.

2

Hi @Aaru229. With the Rasa 1.0 you should use the Rasa CLI to train and test your models. So, instead of writing a script to train the model, you should run the rasa train in your project directory. You can find more documentation on that here. Have a look and let me know if you still have any issues.