how to keep subwords only in final model?

[ZizhenWang]

Hi there

I find the output model contains not only ngram subwords, but also original whole word, which makes the vocabulary size much bigger. I want to just keep the ngram subwords, how can I do that?

Thanks

[Celebio]

Hi @ZizhenWang , What is your use case? Do you want to have a smaller model file? Or you need this feature for something else?

Best regards, Onur

[ZizhenWang]

@Celebio Thanks for your reply, I want to have a smaller model file. I think the subword is necessary in fasttext, but the whole word is not, because it can be inferred by its subwords. Am I right? So I want to keep only the subwords in model file, and remove all whole words.