SUPPLEMENTARY MATERIAL
======================

Track Name: SUSTAINLP 2021

Paper ID: 4

REQUIREMENTS
============

NOTE: This code has been tested using Python 3.7.  When using Python >=3.8, an undiagnosed interaction between the multiprocessing, pickle and scipy.sparse modules causes fatal errors.

USAGE 
=====

export N=10
export C=1.0
export DATA=data/PRM45k

# If requirements are met (see Pipfile), do:
./linkpred.py ${DATA} ${N} ${C} --stage --build --fit --score test.txt

# If you plan to use pipenv, do:
pipenv install

# If torch fails to install with the error,
#     "ModuleNotFoundError: No module named 'tools.nnwrap'"
# we find issuing the following command registers the torch library, and that is sufficient to 
# make use of the DataLoader class, which is all we require.
pipenv install torch

# Run with
pipenv run ./linkpred.py ${DATA} ${N} ${C} --stage --build --fit --score test.txt


ARCHIVE
=======

tar --exclude 'output/*'  -czvf low-resource-kge.tar.gz low-resource-kge 
