If you would like to re-use your MaChAmp finetuned model in another toolkit, you can extract the
transformers model. We provide an example script in scripts/misc/extract_automodel.py, which
needs to be ran from the root of this repo. Usage is as follows:
cp scripts/misc/extract_automodel.py .
python3 extract_automodel.py logs/ewt/*/model.pt mBERT_finetuned_on_EWT
Now the models including its configuration and tokenizer will be written in a folder titled
mBERT_finetuned_on_EWT.