search

winddori2002 / TriAAN-VC

TriAAN-VC: Triple Adaptive Attention Normalization for Any-to-Any Voice Conversion
MIT License
129 stars 12 forks source link

Training time/compute for provided model #3

Closed nateraw closed 1 year ago

nateraw commented 1 year ago

Hi there, thanks for sharing this codebase and the associated models :)

Was curious about a couple things:

  1. how long did it take to train on the VCTK dataset, and
  2. What compute was used? (Type+number of gpus)

This info might help folks who want to look into training on a larger dataset. I tried finding this info in the paper/here but didn't find anything (sorry if I missed something!) Thanks!

winddori2002 commented 1 year ago

Hi,

We used a single RTX3090 GPU, and it took about 45 hours for training TriAAN-VC on the VCTK dataset (with the paper training setting, batch size: 32, epoch: 500).

Thank you.

nateraw commented 1 year ago

Thank you!! :)

Blakey-Gavin commented 1 year ago

Hi,

We used a single RTX3090 GPU, and it took about 45 hours for training TriAAN-VC on the VCTK dataset (with the paper training setting, batch size: 32, epoch: 500).

Thank you.

Hi,

The training settings in your paper are not batch_ Size=64, epochs=400?

image