Skip to content

Latest commit

 

History

History
9 lines (6 loc) · 543 Bytes

README.md

File metadata and controls

9 lines (6 loc) · 543 Bytes
Raw

PixelCNN implementation based on the original van der Oord paper

  • the network output is actually a mixture of logistics following the PixelCNN++ paper by Salimans et al.

  • I took the code for loss computation from the OpenAI pixelcnn_pp repository for stable loss computation but I implemented it myself as well

    • for reasons of stability the OpenAI loss provides better samples

  • here's a sample from a network trained with the OpenAI loss on the CIFAR-10 dataset:

pixelcnn_sample_openai_loss