-
The following questions arose when implementing the paper:
**1. Regarding Post-Net**
In the Tacotron paper the authors cite, Post-Net is used after their so-called 'Synthesizer', the module that p…
-
The following branch implements a part of the parrel that can be referred to as a reference
https://github.com/vincentherrmann/pytorch-wavenet/blob/parallel/wavenet_training.py
-
また、本質的?な指摘ですが、r_channelsやs_channelsなどはWaveNetの推論用CUDA kernelであるnv-wavenetにchainerが対応することを見越して、同じ名前を付けています。READMEにも書いてあるのでチュートリアルでは触れなくても良いかもですが、参考に載せても良いかと思います。
-
https://github.com/vincentherrmann/pytorch-wavenet/blob/091396443bf656b348d9f9c17696d5aedc252eb9/wavenet_training.py#L259
-
Thanks for this useful summary page. Have you considered providing also links to the relevant GitXiv entries, when available?
If you are not familiar with GitXiv, have a look: [http://www.gitxiv.com…
-
I use GTA synthesis by Tacotron-2 to predict mel spectrograms and then train the wavenet vocoder with **default hparams**. I have two problems:
1.The loss < 0
2.When iterating after 196k the traini…
-
当我进行waveNet训练时，进行到67456step时出现以下问题：
Step 67456 [1.078 sec/step, loss=6.73029, avg_loss=6.65925]
Generated 64 train batches of size 4 in 0.298 sec
Loss exploded to 378655.87500 at step 674647500, …
-
I set input_type='raw' .
when I train WaveNet with LJSpeech. the loss is negative quick.
Is something wrong ?
-
i got negative loss when training wavenet with THSCH-30 dataset.
```
Wavenet Train
###########################################################
Checkpoint_path: i:/data/voice/logs-Tacotron-2\…
-
when i try train my mandarin model using one speaker voice by python train.py python tra…