update MNIST experiments

README.md

@@ -13,7 +13,7 @@ This repository implements the following:
 * Improved Training of Wasserstein GANs: [paper](https://arxiv.org/abs/1704.00028), [code](https://github.com/igul222/improved_wgan_training)
 * Deep Regret Analytic Generative Adversarial Networks: [paper](https://arxiv.org/abs/1705.07215), [code](https://github.com/kodalinaveen3/DRAGAN)
 * Conditional Image Synthesis With Auxiliary Classifier GANs: [paper](https://arxiv.org/abs/1610.09585)
-* Learning to Discover Cross-Domain Relations with Generative Adversarial Networks: [paper](https://arxiv.org/pdf/1703.05192.pdf), [code](https://github.com/SKTBrain/DiscoGAN)
+* Learning to Discover Cross-Domain Relations with Generative Adversarial Networks: [paper](https://arxiv.org/abs/1703.05192), [code](https://github.com/SKTBrain/DiscoGAN)
 * Minibatch Stat Concatenation: [paper](https://arxiv.org/abs/1710.10196), [code](https://github.com/tkarras/progressive_growing_of_gans)
 
 ## Requirements

deep_networks/models/blocks.py

@@ -355,9 +355,10 @@ def __init__(self,
                 with tf.variable_scope('conv{}'.format(i + 1)):
                     if i == len(downsamples) - 1:
                         stds = std_eps(outputs)
+                        stds = tf.reduce_mean(stds, axis=-1, keep_dims=True)
                         stds = tf.tile(stds,
                                        tf.concat(
-                                           [tf.shape(outputs)[:-1], [1]],
+                                           [tf.shape(outputs)[:1], [1, 1, 1]],
                                            axis=0))
                         outputs = tf.concat([outputs, stds], axis=-1)