search

NVIDIA / mellotron

Mellotron: a multispeaker voice synthesis model based on Tacotron 2 GST that can make a voice emote and sing without emotive or singing training data
BSD 3-Clause "New" or "Revised" License
853 stars 187 forks source link

number of epoch issue #81

Closed yuzuda283 closed 3 years ago

yuzuda283 commented 3 years ago

The default epoch number in hparams.py is 50000, does that really need so much epoches? because i can't use the distributed run and fp16 run, and i try to use nn.DataParallel() to use multi-gpu, but bugs happened. so if i use one P100 to train the model on 10000 speeh , it may take 200 days with 50000 epoches!

rasenganai commented 2 years ago

How many epochs did you give?