Problems with other datasets.

[a897456]

Hi
I processed the noisy samples through an encoder - decoder (thinking it was a good idea), and the length of the samples became a multiple of 300 (closest to the original sample length). However, the clean samples still have their original lengths. It seems that I can't train SGMSE properly. Do you know the reason?