Lots of them.
-
Firstly, to download the WaveGAN generator model you should be on Windows because I call PowerShell commands, but you can easily work around this if you're on Linux by either writing a shell script or downloading the models manually and putting them in the waveGAN_models folder.
-
Secondly, install Tensorflow. I think any version above 2.0 should work but I installed 2.5. To install TF along with cuda and cuDNN follow this tutorial (fantastic channel). TF is only used to initially generate data.
-
Thirdly, install PyTorch. I installed the version that was relevant as of august 2021.
-
Finally, install everything else. I really should make "requirements.txt" file. All the other libraries are standard.
Download the pretrained WaveGAN Generator as a .ckpt file by running download_wavegan.py but before that, make sure you are on Windows and that you have running PowerShell commands enabled. See link on how to do that. If you're on Linux you should be able to just run the wget commands in the terminal (I can write a shell script but I can't test it 😥). The folder download_scritps has PowerShell scripts ready for use but download_wavegan.py just runs the same commands from a subrpocess call.
Choose you're DATASET and run generate_data.py
Train your model with the train.py script