Generate sequence from a trained LM

Hi,

I have trained a language model using the LanguageModel class and saved it (“best.th”).
I would like to generate some sequences with it: the most likely sequence using beam search.

I loaded the model together with its vocabulary. I also created a dummy instance which contains the token. How can I generate further tokens? I don’t see a straightforward way using the Predictor. Is there a tutorial/code snippets for this?

Thank you.