-
Currently blimpy errors out (in a confusing manner) if the filename doesn't end in .fil or .h5, even if it's a filterbank or hdf5 file. Maybe have a filetype command switch?
-
Below is the exec info:
````
batched-wav-nnet3-cuda --acoustic-scale=1.0 --add-pitch=true --beam=15.0 --feature-type=mfcc --frame-subsampling-factor=3 --frames-per-
chunk=150 --gpu-feature-extract=…
-
**## How to solve it? I thought it was about import problem. But 'import h5py' code couldn't help me.**
To assist reproducing bugs, please include the following:
* Operating System Ubuntu 16.…
-
Hi there, I have found little difference between what the paper said and what I found in code.
In Section 2.2 Paragraph 2 of Tacotron2 paper,
"Prior to log compression, the filterbank output magni…
-
Thanks for your excellent work!
- question 1
I trained with new dataSet, but loss is 'inf'
train.cfg details, on linux ubuntu 16.04
```
--datadir=/data/CNSpeech/
--tokensdir=/data/CNSpeech…
-
Hi, I have found difference between default value of min_level_db and the comment about this in the paper.
In Section 2.2 Paragraph 2 of Tacotron2 paper,
"Prior to log compression, the filterbank …
-
When a frame following the first frame is decoded it uses the previous frame. How is the first frame decoded, and what is used instead of the previous frame?
ghost updated
4 years ago
-
Various system packages must be installed, including:
* ffmpeg
* sox
* Kaldi for pitch features
-
Hello, if i want to train the filterbanks of CNN, how should I do? By executing the script tensorflow_net/dnn-filterbank/train_dnn_filterbank_gpu0 ? How this script is to train the filters of DNN.
-
I tried to complete the training with the example training configuration, but when I used the trained model to decode, there are no candidates output.
>[root@dd88309d-124c-4caf-b665-f87d8b1b66d1 /d…