Glove Co-occurrence matrix issue

[ali alk]

Hello friends,

I am having a large corpus (16 G) and would like to train Glove word embeddings.I am using this implementation to train it ( https://github.com/hans/glove.py )

The problem is that building the co-occurrence matrix is so slow, it will take around two weeks to build the matrix. please any help??

is it possible to use multithreading concept in python to overcome this problem.. Thank you

[Ivan Menshikh]

Hi,

This algorithm isn't a part of gensim, for your case, I'll recommend using another glove implementation (more faster and mature), for example, original GLoVe from Stanford, this implementation should work significantly faster!