how can i get the grad of the input using the hifigan.decode_batch() method ?
#2
by
chenjojo
- opened
hi, i am using the code you provided "waveforms = hifigan.decode_batch(mel_output)" in the Model Card, i set the attributes of the mel_output "requires_grad=True", but it seems cannot return the grad throungh backward(), even i remove the "with torch.no_grad()" in the decode_batch() method. how can i return the grad of mel_output?