This repository introduces and implements MemoryBank, a Question Answering system with improved consistency and accuracy. This code runs with Python 3.9 on a google cloud A2 compute instance.
To pull BeliefBank data:
./get_beliefbank.sh
Set up virtual environment with dependencies:
python3 -m venv env
source env/bin/activate
pip install -r requirements.txt
Run Memory bank on validation dataset
python run.py --mode 'val'
Run Memory bank on test dataset
python run.py --mode 'test'
To adjust batch size:
python run.py --batch_size 10
To visualize results, run:
tensorboard --logdir=runs
And open http://localhost:6006/