Closed PharrellWANG closed 5 years ago
We compare our method with the VPN baseline and we've mentioned it in the experiment section.
Thanks for the reply. I see.
@xzr12 In experiment section: “Due to the difficulty of reproducing the experimental results of the VPN”
May I ask what exactly is the “difficulty” mentioned in the paper? (Insufficient GPU memory for 256 channels in convLSTM or prohibitive converging time for 19.2M training sequences? Or others?)
The difficulty mentioned in the paper mainly lies in the reproduction of PixelCNN decoder so we adopt the baseline version of VPN.
@xzr12 Hi Ziru,
Thanks for your paper and code. After skimming your paper, I have a question: Are you comparing to "VPN baseline" or "VPN"?
Best, Pharrell