Train custom spacy model with rasa train


I am trying to train a spacy model instead of DIET to recognise entities using rasa train. Is that possible by simply using the Spacy classes (EntityExtractor, Tokenizer, Featurizer) as described here Using Custom spaCy components | The Rasa Blog | Rasa?

I guess the alternative is to train either using spacy and load the trained model or using a custom component similar to How to Enhance Rasa NLU Models with Custom Components | Rasa Blog | The Rasa Blog | Rasa. In the former case, how would you convert the data to spacy format?