How to run a project that using MultiProjectImporter

Rasa Open Source
1

I want to running my multiple project that can be combined using MultiProjectImporter. But I confused how to run this project after following the guidance from the rasa documentation.

Does anyone know how to solve this? Thankyou.

Sorry for my bad english.

2

Hi @paraparata, can you explain what is ambiguous and what you didn’t understand from the documentation. I can then clarify that part.

3

The root folder on the guidance looks like this:

image

There’s two projects, first is GreetBot and second is ChitchatBot. In which folder I should rasa run the project? I tried to run it in root folder but it gave me an error result.

Thank you

4

Can you specify the exact command you used and the error log that you get?

5

Hi @dakshvar22 Can you please specify what command do we need to use in the root directory for running the multi project importer? I have few other questions? 1.) when you say individual projects? Do they need to be separately trained in the projects folder and then imported in the config file in the root directory? 2.)When I run rasa train by just following the project structure specified on your account I get an error message that the model is not trained. 3.) I have the data folder specified in the individual projects as specified in the directory structure in your documents and the data folder is deleted from the root directory is this the correct method?

Thank you Regards, Awaiting your reply.

6

Came here to leave a positive comment for the MultiProjectImporter (we finally got it working and dont want it to get removed D: ). Devs, please we need this. Don’t remove it in the future.

Let me try to help kedarnad as well.

  1. No, no independent training. You choose a primary or dominant project, and invoke the rasa train command on that project. But the config of the dominant project should import the others. During training time all data will be merged and one single model will be trained.
  2. post your error please. that would make it easier for people to spot issues (along with code in github if possible for a dummy test case)
  3. I’m not sure if there is a correct way, but we mash up things a lot, but manage to get the training working though command line args to the rasa train command. There is no wrong if it works I hope :slight_smile: