Rasa shell: Could not load model due to Error initializing graph component

Hi everybody,

I succeed to train my model but when I run rasa shell from command line I get the following error message:

rasa.core.agent  - Could not load model due to Error initializing graph component for node 'domain_provider'..

I am running :

Rasa Version      :         3.0.0
Minimum Compatible Version: 3.0.0
Rasa SDK Version  :         3.0.0
Rasa X Version    :         None
Python Version    :         3.7.11
Operating System  :         Windows-10-10.0.19041-SP0

Thankful for any help…

Stefano

@stefanod hello and welcome to the forum!

Did you create the rasa init or using your own project? can you share the model file (*.tar.gz) how it looking ?

The error was caused by “forms” entry in domain.yml file. It seems that with Rasa 3.0 something like: forms: my_order_form: {} cause this issue even tough the model is trained without error messages. As a matter of fact I am experiencing other issues probably caused by the use of no more supported statements I wonder whether you could suggest any tutorial on Rasa 3.0. At the moment, I am finding only material about Rasa 2.x. I would need some working examples with forms and custom actions. Many thanks Stefano D.

@stefanod sure please find this repo for form examples: https://github.com/RasaHQ/rasa-2.x-form-examples (in 2.x) try experiment with this? thou personally I have not started exploring rasa much as my bot is in production, and I am working on 2.8.1 only. I hope this will help you. Hope this will solve your issue?

Further, there is much change in forms, but yes ref the Rasa 3.x Documentation and try implement your use case.

Hi @stefanod,

I got the same error message and in my case deleting the trained models and training the bot again resolved the issue. Maybe you give it a try :slight_smile:

3 Likes

I have the same issue, can cannot solve it now. Could someone help me, please?

I had this error in case the download cache directory of the bert language model was not writable, e.g. if created by root. This is the directory which is referred to by the environment variable TRANSFORMERS_CACHE or the directory ./.cache, if the variable is not set.

1 Like

One possible source of this error in general is a mismatch in python library versions between build and run environments.

To debug, generate a list of installed libraries on all your environments (via e.g. pip list or pip freeze) and compare, looking for version differences in important packages like rasa, tensorflow, and scipy.