r9y9/deepvoice3_pytorch

Increasing Training steps

Arafat4341 opened this issue · 2 comments

Hello!
How can I increase training steps. I can't see any parameter for this. My training is completed upto 960,000 steps. But I want more improved results.
Please help!

r9y9 commented

Change

nepochs=2000,
. But I don't think you can get big improvements if you just train more (I guess your model already reached the limitation of the model)

@r9y9 Thanks a lot.
But the thing is, I get unnatural sound.
I used deepvoice3_ljspeech.json as preset for pre-processing, training.
Later I changed the "frontend"="jp" in the deepvoice3_nikls.json file and passed it to synthesize.py.
Can you tell me what to do?