Hello,
I want to use a custom bert model in the pipeline as below. When I use the whitespacetokenizer in the pipeline, I can train the bot and get a response. When I don’t use a tokenizer, I get the error “rasa.engine.exceptions.GraphSchemaValidationException: 1 components are missing required components which have to run before themselves”.
Languagemodeltokenizer has been deprecated in Rasa 3.x. Bert Tokenizer is wordpiece tokenizer, so I dont want to use whitespacetokenizer in the pipeline, I want to use the tokenizer which belongs to Bert I use. How can I do that?
Can you share your thoughts on this? Am I missing something here?
