WaveRNN
WaveRNN copied to clipboard
changing downsample factor?
What do I need to change in order to increase the downsample factor from 64 to 128? It seems it is not as simple as adding an additional layer to the encoder. What other variables need to be adjusted?
the downsampling happens in multiples of four, i.e. you can have a factor of 4, 16, 64 or 256. for implementing 256, you need to add another encoder layer, decoder layer, and adapt the inference function.
In this code it does not appear as simple as adding additional layers. If you have successfully changed the downsample factor please share the full solution.