Try out BERT cause why not.
Only if transformers had longer sequence length, the world would have been a better place.
dataset.py and train.py is pretty general for any kind of BERT fine-tuning task. Change the model and
and you are good to go.
Hardcoded main.py cause I was lazy. Feel free get inspired and change as required.
- Change
main.pyas required. - Make changed to the model in
train.py pip install -U transformers tezpython3 main.pyand watch the loss go brr