No alignment without linear spectrograms

[JRMeyer]

>>> geneing
[May 21, 2019, 4:49pm]

I need help understanding something.

I removed the linear spectrogram part of the loss function, along with
postnet that generates it. I didn't need the linear spectrogram for the
vocoder and removing the linear spectrogram part save a LOT of GPU
memory during training. However, the reduced model doesn't produce
reasonable attention even after 50K steps. For the full model, attention
was reasonable after only a few thousand steps.

Why isn't mel spectrogram part of the loss not enough to train the
attention?

[This is an archived TTS discussion thread from discourse.mozilla.org/t/no-alignment-without-linear-spectrograms]

[JRMeyer]

>>> erogol
[May 23, 2019, 12:13am]

could you check gradient norms between these two runs? Maybe , the scale
of loss value has been changed and it requires a new learning rate. If
you have tensorboard files, you can also share them.

[Archived Post]