Is the Learning Speaker Representations with Mutual Information paper ever going to be implemented?

As in the title. I am interested in seeing how it could be implemented, as I have tried to myself and the results weren't as good as the ones described in the paper, even with a huge dataset, and am not sure how to make the model yield better results.
Have a good day.