search

as-ideas / ForwardTacotron

⏩ Generating speech in a single forward pass without any attention!
https://as-ideas.github.io/ForwardTacotron/
MIT License
578 stars 113 forks source link

Bad Attention! #96

Closed MohamedAlmahmood closed 1 year ago

MohamedAlmahmood commented 1 year ago

I'm training on an Arabic dataset that I've created. When running !python train_forward.py I get the following: Using 287 train files. Filtered 1681 files due to bad attention!

I don't know what's causing this. Could someone please help!