You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I tried your code, hparams not changed, the wavs generated of 44k steps is understandable, but the line noise appear as many people reported in the nvidia code branch.
But your samples sounds great and suffer from this problem little, did you add post-process to the wavs? You have any tips ? You change the value of n_group or the sigma?
By the way, I tried the b04901014's branch, using lr=0.001, after about 30k, the model converge, and the total loss is about -5.8
The text was updated successfully, but these errors were encountered:
I tried your code, hparams not changed, the wavs generated of 44k steps is understandable, but the line noise appear as many people reported in the nvidia code branch.
But your samples sounds great and suffer from this problem little, did you add post-process to the wavs? You have any tips ? You change the value of n_group or the sigma?
By the way, I tried the b04901014's branch, using lr=0.001, after about 30k, the model converge, and the total loss is about -5.8
The text was updated successfully, but these errors were encountered: