Error in using my customized tokenizer

Rasa Open Source
1

Hi. I am moving from rasa 2.x to rasa 3. And i meet this problem:

InvalidConfigException: Can't load class for name 'MyTokenTokenizer'. Please make sure to provide a valid name or module path and to register it using the '@DefaultV1Recipe.register' decorator.

Here is the infomation you may need:

  1. currently rasa 2.7 works while rasa 3.0 fails, but since i got this projects and envs from an other guy, i dont know what he changed in rasa 2.7
  2. MyTokenTokenizer: just treat it as a common nlp tokenizer, writen by my colleague.
  3. i only use rasa nlu
  4. here is my config.yml config.yml (1.7 KB)

i think it should just a normal problem about how to apply personal customize file into rasa config, but i looked around and around and cant find it : (

2

Hi @QuantumCoder,

Rasa 3.0 contained major changes to the way components are implemented, so the error you see could come from using the Rasa 2 component with Rasa 3. When migrating to Rasa 3.0, did you adapt your MyTokenTokenizer class to the new graph architecture?

If not, please check the migration guide and requirements for custom graph components to adapt your component.

1 Like
3

Great. I will check it. Thanks!