Most likely. Technically you could use the DIET classifier. You don’t really need a pre trained model for it.
You can potentially prune the fast text model vectors by importing it into spacy and that should reduce the size, I did that but like 3 years ago I am not sure how it works with spacy 3.0