anonymous

Train the model

python main.py --cuda --do_train --do_valid --do_test --data_path data/DATASET-ind --geo GEO --tasks "1p.2p.3p.2i.3i.ip.pi.2u.up" -max_n 32 -ee -es -se

GEO: the reasoning method, vec for GQE, box for Query2box

max_n: max neighbors for getting embedding in inductive setting

ee, es, se: the type of queries. The front represents the anchor, and the back is the answer. e represents emerge, s for seen

Test the model

python main.py --cuda --do_test --data_path data/DATASET-ind --geo GEO --tasks "1p.2p.3p.2i.3i.ip.pi.2u.up" --checkpoint_path PATH -max_n 32 -ee -es -se