muhaochen / on2vec Goto Github PK

Hi, I am reading your code and trying to regenerate the result in your paper these days and here are some questions I met. Thanks for answering!

1. I couldn't find the full dataset as you mentioned in your code, like the directory 'onto2vec-dataset'. Since the dataset seems to be preprocessed, could you send me the link of the full dataset?
2. In 'training_yago.py', you use
   35 more_filt = ['../../../onto2vec-dataset/Yago/yagoFactsFiltCDTTest.tsv']
46 for f in more_filt:
this_data.record_more_data(f)
   while the file is also used as the test file in 'test_yago_ver1_rel.py' where:
   25 test_data = '../../../onto2vec-dataset/Yago/yagoFactsFiltCDTTest.tsv'

I suppose it will cause a quite high accuracy if a test file is used as training data. In fact, I used 'yagoFactsFilt15kTest.tsv' in your zip file and run a result as follws:

1.0_test_yago_rel.txt:
Rel
0.9984560455466563
trsym
1.0
hier
0.9969272133666218

There must be something wrong :-(
Could you please tell me the reason?
Thanks a lot !

mldl@ub1604:/ub16_prj/on2vec/yago_wohm$ PYTHONPATH=../src python training_yago.py
/usr/local/lib/python2.7/dist-packages/h5py/init.py:36: FutureWarning: Conversion of the second argument of issubdtype from float to np.floating is deprecated. In future, it will be treated as np.float64 == np.dtype(float).type.
from ._conv import register_converters as _register_converters
Loaded ../on2vec-yg60k/Yago/yagoFactsFilt15kTest.tsv to triples_record.
Traceback (most recent call last):
File "training_yago.py", line 53, in
m_train.build(this_data, dim=100, batch_size=500, save_path = model_path, data_save_path = data_path, L1=False)
File "/home/mldl/ub16_prj/on2vec/src/trainer.py", line 36, in build
L1=self.L1)
File "/home/mldl/ub16_prj/on2vec/src/model.py", line 44, in init
self.build()
File "/home/mldl/ub16_prj/on2vec/src/model.py", line 146, in build
A_loss_matrix = tf.sub(
AttributeError: 'module' object has no attribute 'sub'
mldl@ub1604:/ub16_prj/on2vec/yago_wohm$