Questions about the using the pre-trained model

[leepokai]

@kaituoxu I've been using a voice separation model provided by Baidu to separate two overlapping voices into individual audio tracks. However, I've noticed that the separated audio contains a lot of noise. What could be the possible reasons for this issue?

I've uploaded my modified test code and the resulting output on the following GitHub page: https://github.com/leepokai/Conv_TasNet_trying.

Thank you for your help!

[InnerPeace-code]

me too, I didn't meet this problem when I use TasNet.