Hi,
I’m new to the TTS/Tensorflow world. My objective is first to be able to use existing dataset to test my environment and learn how the TTS training process works.
After a lot of attempt I was able to run the TTS command
python3 TTS/bin/distribute.py --script train_tacotron.py --config_path TTS/tts/config/config.json . In parallel I was also able to run tensorboard. I executed this command on the LJ-Speech-Dataset dataset.
Overall the output of the command and Tensorbord shows it’s progressing, from time to time in the output I see following warning but it’s doesn’t stop the command from running: Warning: NaN or Inf found in input tensor
I know it’s only at 2K steps … but when I run the synthesizer, as expected it generated some robotic audio.
However, my issue is that on the tensorboard “Audio” tab, I was expecting to hear the a similar audio as the synthesizer, however it’s just dead air I doesn’t produce any audio , it’s muted. Even if I try to download and listen to it via VLC there is nothing.
I was wondering if this is expected or there is something I need to configure?
I’ve attached some screenshot
