raphaelmerx/xlm-roberta-large-tetun

Tetun BERT model

A fine-tune of xlm-roberta-large trained on Tetun data with a masked language modelling objective.

Tetun data used: MADLAD tet clean split (~40k documents).

Trained for 10 epochs with hyper params from the MasakhaNER paper (lr 5e-5 etc).