Rasa nlu: RuntimeError: Calling `rasa.nlu.train` using docker


I am using rasa-nlu for my project. I am running rasa-nlu using docker. I am using the below command to train the rasa-nlu.

docker run
-v $(pwd):/app/project
-v $(pwd)/models:/app/models
python -m rasa_nlu.train -c config.yml --data project/data/MyModel -o models --fixed_model_name MyProject --project current --verbose

Whenever I am trying to train the rasa-nlu model I getting the following issue:

  1. /app/rasa_nlu/init.py:12: UserWarning: The ‘rasa_nlu’ package hase been renamed. You should change your imports to use ‘rasa.nlu’ instead.
  2. RuntimeError: Calling rasa.nlu.train directly is no longer supported. Please use rasa train nlu instead.

I was not facing the issue earlier while using the same command. I think issue is because of the latest update on 23 April 2019 [Ref. Link: https://rasa.com/docs/nlu/changelog/].

Please help me to resolve this.

Hi @Anurag53 I was having the same issue while running my chatbot in docker. I wasted a lot of my time is solving this. Please use rasa/rasa_nlu:0.14.4-full image from docker hub instead of rasa_nlu : latest. They have made some changes in that. Your issue will get resolved surely. Please get back if any further issue is there.

Hi @sank1992, thank you so much for your suggestion it worked for me.