• implementation of continual learning for repeated reference is in /models/
• interactive, web-based reference game w/ model is in /behavioral_experiments/
• experiments on model behavior are in /listener_experiments/ and /speaker_experiments/
• csv output from both kinds of experiments goes in /data/
• scripts for analyzing/visualizing data from experiments are in /analysis/
• paper is in /writing/

1. install dependencies and the COCO API:

conda install pytorch torchvision -c pytorch
conda install Cython nltk
pip install -U symspellpy
git clone https://github.com/pdollar/coco.git
cd coco/PythonAPI/
make
python setup.py build
python setup.py install
cd ../../

1. clone repo and download coco annotations and images for experiments

git clone https://github.com/hawkrobe/continual-adaptation.git
sh download-coco.sh

1. Extract pretrained model weights and vocabulary files to ./data/preprocess/

For more information on pretrained model, see tutorial here

Model code depends on PyTorch >=1.2.0