Open aidosRepoint opened 1 year ago
Hi!
in model.module line 128, we have
model.module
if duration_target is not None: x, mel_len = self.length_regulator(x, duration_target, max_len) duration_rounded = duration_target else: duration_rounded = torch.clamp( (torch.round(torch.exp(log_duration_prediction) - 1) * d_control), min=0, ) x, mel_len = self.length_regulator(x, duration_rounded, max_len) mel_mask = get_mask_from_lengths(mel_len, self.device)
This means, that in train mode, there will always be duration_target. Does it mean that the output from VarianceAdaptor's forward method will always return the true value for durations? Does it mean that the loss calculation is wrong?
Hi!
in
model.module
line 128, we haveThis means, that in train mode, there will always be duration_target. Does it mean that the output from VarianceAdaptor's forward method will always return the true value for durations? Does it mean that the loss calculation is wrong?