Does this pytorch implementation of MuseGAN give similar results to the original?

[Kasidit0052]

Hi, I tried to train this implementation of MuseGAN, but midi results seem to be repetitive compare to results from original repo. I wonder does this pytorch implementation of MuseGAN give similar results to the original tensorflow version where it can produce more longer notes, and variety music pieces. If no, could you elaborate which part of original MuseGAN is responsible of that task.

[lcc-404]

This version of the code seems to be a bit less than the original?