如何在 Docker 中的 RASA 中训练和测试 nlu 模型？

Question

Having issues with trainig RASA inside Docker. I enter it: docker run -it -v C:\\Users\\User\\Documents:/app --entrypoint=/bin/bash rasa/rasa

Then I provide a command line to train with exact path to traning_data (it exists!) rasa train --data %cd%\\data\\split_train\\train_data.md nlu But every time I got this

2019-12-16 17:35:39 WARNING  rasa.cli.utils  - '%cd%datasplit_traintrain_data.md' does not exist. Using default value 'data' instead.
2019-12-16 17:35:39 WARNING  rasa.cli.utils  - 'nlu' does not exist. Using default value 'data' instead.

What should I do to finally be able totrain and test my utterances in C:\\Users\\User\\Documents\\data\\split_train\\training_data.md?

Answer 1

You need to execute rasa train inside the running docker container.

check your running containers by executing:

docker ps

now execute rasa train inside the container with your model :

docker exec <container-name> rasa train <arguments>

test if the model was trained:

docker exec -it <container-name> rasa shell <arguments>

Reference to docker exec in the Docker Docs .

See also the Rasa DocsRunning Rasa with Docker .