Questions About Phase/Magnitude Computation and Hop Size Sensitivity

[EuiYeonKim]

Hello,

I am very interested in your research and would love to explore it further. While reviewing your code, I had a couple of questions that I hope you can clarify:

1. I noticed that instead of using torch.angle() or torch.abs() to compute phase and magnitude, you implemented these computations directly. Is there a specific reason for this choice?

2. In another issue, I saw that the model is highly sensitive to hop_size when applied to different tasks. Could you please explain why this sensitivity occurs and how it manifests?

Thank you for your time and for sharing your work!