Gated PixelCNN
A TensorFlow implementation of the gated variant of PixelCNN (Gated PixelCNN) from "Conditional Image Generation with PixelCNN Decoders" (https://arxiv.org/abs/1606.05328). The Gated PixelCNN matches the log-likelihood of PixelRNN on both CIFAR and ImageNet while requiring less than half the training time.
Training the Network
MNIST (default)
python main.pyColor-MNIST
python main.py --data color-mnist --gated_conv_num_layers 7 --gated_conv_num_feature_maps 48 --output_conv_num_feature_maps 96 --q_levels 4CIFAR-10
python main.py --data cifar --gated_conv_num_layers 15 --gated_conv_num_feature_maps 126 --output_conv_num_feature_maps 1020 --q_levels 256| Configuration Parameter | Quick MNIST | MNIST | COLOR-MNIST | CIFAR-10 | Description |
|---|---|---|---|---|---|
| batch_size | 100 | 100 | 100 | 100 | Size of a batch. |
| gated_conv_num_layers | 1 | 7 | 7 | 15 | The number of gated conv layers. |
| gated_conv_num_feature_maps | 4 | 16 | 48 | 126 (128 in paper) | The number of input / output feature maps in gated conv layers. Must be multiple of two, should be multiple of two times num_channels. |
| output_conv_num_feature_maps | 16 | 32 | 96 | 1020 (1024 in paper) | The number of output feature maps in output conv layers. Must be multiple of two, should be multiple of two times num_channels. |
| q_levels | 4 | 4 | 4 | 256 | The number of quantization levels in the output. |
| data | "mnist" | "mnist" | "color-mnist" | "cifar" | Name of dataset. |
Gated Convolutional Layers
