Right now you need to input a factor that evenly divides the shape for subsample to work, and any other value doesn't work
Either
- Remake the interface so that the user inputs the factor they want to downsample by so that you don't need to directly input the number of channels, or,
- Change the downsampling algorithm to handle different sizes. FFT downsampling? interpolating?
Right now you need to input a factor that evenly divides the shape for subsample to work, and any other value doesn't work
Either