Remove several warning and better floating point representation on the progress bar

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

2.62k stars 275 forks source link

Remove several warning and better floating point representation on the progress bar #115

Closed hrukalive closed 11 months ago

hrukalive commented 11 months ago

Got rid of several warning messages at the beginning of the training.
Automatic floating point representation (positional and scientific) on the TQDM progress bar based on precision.