VA BERT and VA-R BERT

This repository contains the code and pipelines to train named entity recognition and relation classification models for visual acuity in clinical letters.

Training VA BERT

## Training
Train with default config (defined in `config/train_config.json`)
```bash
python train.py

Override config's parameters by passing optional arguments

CUDA_VISIBLE_DEVICES=1 python train.py \
	--checkpoint 'models/name' \
	--max_epoch 150 

Training VA-R BERT

## Training
USE -- jupyter notebook provided in notebook folder.

Other model/checkpoint can be from huggingface: dmis-lab/biobert-v1.1, bert-base-cased, etc.

Note: Check config/train_config.json before training