kkrasnas

Results 4 comments of kkrasnas

Document representaion.

Yes! After some epochs, the dirichlet loss freezes and every document belongs to all topics equally.

Giving same representaion for every document.

I have the same issue!

why is loss_nce NaN through out the training phase ?

I have this problem for my tests. I use Bloomberg news from financial news dataset https://github.com/philipperemy/financial-news-dataset, where I am taking only 10 documents for testing. And I want 10 documents...

Please add an example

Please do. Based on the issues created here the code doesn't work properly, so we need more insights.