Can't give torch.Tensor Input
#13
by
ehmargondal
- opened
Calculated padded input size per channel: (0). Kernel size: (1). Kernel size can't be greater than actual input size
This error occurs whenever I input a wav file that I have loaded from directory
It looks like your audio has two channels (first dimension of the waveform tensor is 2) - could you try converting it to mono and passing it through the model? The simplest way of doing this is by summing up the first and second channels along the channel dimension (dim 0):
mono_waveform = torch.sum(waveform, dim=0)
(note that this is pretty naive and might distort the amplitude, but will give you an idea of whether the channels are the issue)