search

chrisdonahue / wavegan

WaveGAN: Learn to synthesize raw audio with generative adversarial networks
MIT License
1.33k stars 282 forks source link

Overtraining #114

Open JordanDavis-1 opened 2 years ago

JordanDavis-1 commented 2 years ago

Has anyone run into the problem of overtraining with their models? I’m doing quite a bit of different datasets and starting to worry about it with some of the results coming back extremely similar.

I’m using ~500-1000 unique .wav per model

kcrosley-leisurelabs commented 2 years ago

How do you even get this to run anymore? I just became familiar with it and can’t get a Jupyter notebook up and running in Colab to train. Have one that works here in 2022?

I’ll send you coffee, cigs, and donuts. (Seriously.)

mattjwarren commented 2 years ago

I have created an updated fork of this repository migrated to run on Tensorflow 2. You may have more success getting it to run in modern environments, especially if you are aiming to train on newer GPU hardware. https://github.com/mattjwarren/wavegan