A BERT based classifier for sentiment classification. The code in this repository contains the script for training the classifier. For training the classifier, the BERT backbone can either be kept frozen or fine-tuned by specifying the appropriate argument.
The model is optimized using the AdamW optimizer implemented in optimizer.py.
For training the classfier, run the following command:
python3 classifier.py --option [freeze/finetune] --epochs NUM_EPOCHS --lr LR --train {train_path} --dev {dev_path} --test {test_path}