Skip to content

Latest commit

 

History

History
60 lines (50 loc) · 1.36 KB

README.org

File metadata and controls

60 lines (50 loc) · 1.36 KB

Training TransE on NAACL data

Setting up Python on cannon

source new-environment/bin/activate
bash

Open session and move to the right directory

cd ~/workspace/SME

Download corpora

FB15k

wget -P FB15k https://everest.hds.utc.fr/lib/exe/fetch.php?media=en:fb15k.tgz
tar xvfz FB15k/fetch.php\?media\=en\:fb15k.tgz 

WN

wget -P WN https://everest.hds.utc.fr/lib/exe/fetch.php?media=en:wordnet-mlj12.tar.gz
tar xvfz WN/fetch.php?media=en:wordnet-mlj12.tar.gz

Parse data

NAACL

python NAACL/NAACL_parse.py

FB15k

python FB15k/FB15k_parse.py

Train models

TransE

FB15k
python FB15k/FB15k_TransE.py
NAACL
cd NAACL
python NAACL_TransE.py

Evaluate model

FB15k

python FB15k_evaluation.py FB15k_TransE/best_valid_model.pkl 

NAACL

python NAACL_evaluation.py NAACL/best_valid_model.pkl