Transfer Learning using pre-trained checkpoints

chrisdonahue / wavegan

WaveGAN: Learn to synthesize raw audio with generative adversarial networks

MIT License

1.32k stars 282 forks source link

Open moiseshorta opened 3 years ago

moiseshorta commented 3 years ago

Hi,

I'm giving WaveGAN another go this year.

Mainly, I'm wondering if it's possible to save compute time and perform transfer learning from the provided pre-trained checkpoints?

I've tried directing the folder to where the example pre-trained checkpoints are (i.e. drums) but the experiment keeps starting from scratch.

Thanks

luisarandas commented 3 years ago

Any available script for this?

luisarandas commented 3 years ago

@moiseshorta if you have any updates on this please share! I'm training this from scratch and (of course) is taking a long time.. Thanks!