Fixes ability to train a pitch model

openvpi / DiffSinger

An advanced singing voice synthesis system with high fidelity, expressiveness, controllability and flexibility based on DiffSinger: Singing Voice Synthesis via Shallow Diffusion Mechanism

Apache License 2.0

2.73k stars 288 forks source link

Fixes ability to train a pitch model #219

Closed AnAndroNerd closed 1 week ago

AnAndroNerd commented 1 week ago

LYNXNet update broke the abillity to train a pitch model, this PR fixes that. Thank you cippbi for the fix.

cippbi commented 1 week ago

confirmed that this fixes the issue for lynxnet