search

NVIDIA / flowtron

Flowtron is an auto-regressive flow-based generative network for text to speech synthesis with control over speech variation and style transfer
https://nv-adlr.github.io/Flowtron
Apache License 2.0
887 stars 177 forks source link

Model not producing any sound #111

Closed reubzdubz closed 3 years ago

reubzdubz commented 3 years ago

Hi, I have been fine tuning the Flowtron LJS model for about 39 epochs with a data set of about 1.2 hrs (single speaker) I set attention prior as true and with n_flows equals 2 attention weight 0 attention weight 1 The attention weights seem to be quite visible already, however when I try to infer a text, the model does not seem to be producing any sound. Is there some sort of configuration I forgot to do?