Feature/shared encoder for input

emanjavacas / pie

A fully-fledge PyTorch package for Morphological Analysis, tailored to morphologically rich and historical languages.

MIT License

22 stars 10 forks source link

Feature/shared encoder for input #71

Closed PonteIneptique closed 3 years ago

PonteIneptique commented 4 years ago

Proposal for fixing #70

It's unittested which seemed to be the best way to check things out. Only the tagging/training is not unittested.

I did not give the option to force reusing (and not caring about new characters...), maybe this is a feature you think could be useful ? (training on two different dataset one smaller than the other ?)

PonteIneptique commented 3 years ago

Closed as per #70 discussion: not worth the complication.