igul222 / improved_wgan_training

Code for reproducing experiments in "Improved Training of Wasserstein GANs"
MIT License
2.35k stars 668 forks source link

reproduce unsupervised inception scores #58

Open fjxmlzn opened 6 years ago

fjxmlzn commented 6 years ago

Hi,

I enjoyed reading this WGAN-GP paper so much! I am trying to reproduce the unsupervised inception scores 7.86+-0.07 on CIFAR-10. I just want to make sure what I did was correct. Should I just change CONDITIONAL = True and ACGAN = True to CONDITIONAL = False and ACGAN = False in gan_cifar_resnet.py, and keep ALL other hyperparameters unchanged?

Thank you very much! @igul222