This is a PyTorch implementation of Categorical Info WaveGAN (ciwGAN) and Featural Info WaveGAN (fiwGAN) from Beguš, 2021. The original code (in Tensorflow 1.12) can be found here. In this fork I invetigate the use of a 2nd Q network
python -m pdb -c c train_Q2.py --ciw --Q2 --num_categ 11 --datadir ~/notebooks/talker_variability/TIMIT_padded/ --logdir logs_q2
Add --cont last to the end of the training statement to continue from the last found state
Add --cont epoch_number to continue from the state corresponding to epoch_number