Show HN: WinkNLP delivers 600k tokens/second speed on browsers (MBP M1)

I remember using it. Pretty fast and accurate. My only concern was its failure to deal with verbs used as adjectives, like "bleeding edge" as VRB NN

The documentation is amazingly clear and thorough.

Two questions:

Is there a way to train custom models?

Is there a way to convert models trained with spacy or PyTorch to WinkNLP?