The main simple longformer training notebook.
Notebook that generates more data by back-translation
More clean one: https://www.kaggle.com/dinowun/eda-simplified-feedback-prize More comprehensive one: https://www.kaggle.com/erikbruin/nlp-on-student-writing-eda
Create virtualenv
python -m venv venv # or your way of creating it
source venv/bin/activate
pip install --upgrade pip # cause why not
Install requirements
pip install -r requirements.txt
Get model
bash get_model.sh
Run train
python train.py