DRGN

COLING 2022: Dynamic Relevance Graph Network for Knowledge-Aware Question Answering

How to train the code?

Install Dependency files

• Python == 3.7
• Pytorch == 1.12.1
• Transformers == 4.21.3
• Torch-geometric

Run the following commands to create a conda environment:

CUDA Version: 11.6
conda create -n drgn python=3.7
source activate drgn
pip install numpy==1.18.3 tqdm
pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu116
pip install transformers nltk spacy==2.1.6
python -m spacy download en

#for torch-geometric
pip install --upgrade torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric -f https://data.pyg.org/whl/torch-1.12.1+cu116.html

config

The config files:

• data_preprocessing: preprocess.py
• utils: utils/parser_utils.py
• hyper-parameters: qa_drgn.py
• script (CommonsenseQA): drgn_run_script/run_drgn_csqa.sh
• script (OpenbookQA): drgn_run_script/run_drgn_obqa.sh

Data

Download all the raw data -- ConceptNet, CommonsenseQA, OpenBookQA -- by

./download_raw_data.sh
python preprocess.py -p <num_processes>

Or

./download_preprocessed_data.sh

Training the code:

For CommonsenseQA, run

cd drgn_run_script/
sh run_drgn_csqa.sh

For OpenBookQA, run

cd drgn_run_script/
sh run_drgn_obqa.sh