1.
Please install the tensorflow library
2.
Clone this repository.
3.
Create folder named glove.6B.
4.
Download data and put them in the folder above: http://nlp.stanford.edu/data/glove.6B.zip.
5.
Run python2 train1.py.
1.
Change l1-distance to elementwise-l1 distance in utils.py as discribed in paper.
2.
Allow various value of num_filter_A.
3.
Add small lambda before feeding into tf.sqrt to avoid nans in loss.