Advice for using on Consumer Level Hardware

NVIDIA / flowtron

Flowtron is an auto-regressive flow-based generative network for text to speech synthesis with control over speech variation and style transfer

https://nv-adlr.github.io/Flowtron

Apache License 2.0

887 stars 177 forks source link

Advice for using on Consumer Level Hardware #114

Open TercioOfParma opened 3 years ago

TercioOfParma commented 3 years ago

Hi,

I am looking experiment and to use flowtron for a project I am working on. I have a GTX 1660 TI for my laptop. I understand it's a much lower power system than the GPU recommended in the paper. I've been mucking around with different batch_sizes and this seems to have a positive impact.

How should I tweak the settings, if at all?

Thanks

deepglugs commented 3 years ago

see #119. For inference, you don't need nearly as much VRAM, but the longer the utterance, the more VRAM you'll need in general.