small update

examples/DeepQNetwork/DQN.py

@@ -90,7 +90,7 @@ class Model(ModelDesc):
                  .Conv2D('conv3', out_channel=64, kernel_shape=3)
 
                  # the original arch is 2x faster
-                 # .Conv2D('conv0', image, out_channel=32, kernel_shape=8, stride=4)
+                 # .Conv2D('conv0', out_channel=32, kernel_shape=8, stride=4)
                  # .Conv2D('conv1', out_channel=64, kernel_shape=4, stride=2)
                  # .Conv2D('conv2', out_channel=64, kernel_shape=3)
 

examples/GAN/Image2Image.py

@@ -42,6 +42,11 @@ LAMBDA = 100
 NF = 64  # number of filter
 
 
+def BNLReLU(x, name=None):
+    x = BatchNorm('bn', x)
+    return LeakyReLU(x, name=name)
+
+
 class Model(GANModelDesc):
     def _get_inputs(self):
         SHAPE = 256
@@ -54,8 +59,7 @@ class Model(GANModelDesc):
         with argscope(BatchNorm, use_local_stat=True), \
                 argscope(Dropout, is_training=True):
             # always use local stat for BN, and apply dropout even in testing
-            with argscope(Conv2D, kernel_shape=4, stride=2,
-                          nl=lambda x, name: LeakyReLU(BatchNorm('bn', x), name=name)):
+            with argscope(Conv2D, kernel_shape=4, stride=2, nl=BNLReLU):
                 e1 = Conv2D('conv1', imgs, NF, nl=LeakyReLU)
                 e2 = Conv2D('conv2', e1, NF * 2)
                 e3 = Conv2D('conv3', e2, NF * 4)
@@ -89,16 +93,13 @@ class Model(GANModelDesc):
     def discriminator(self, inputs, outputs):
         """ return a (b, 1) logits"""
         l = tf.concat([inputs, outputs], 3)
-        with argscope(Conv2D, nl=tf.identity, kernel_shape=4, stride=2):
+        with argscope(Conv2D, kernel_shape=4, stride=2, nl=BNLReLU):
             l = (LinearWrap(l)
                  .Conv2D('conv0', NF, nl=LeakyReLU)
                  .Conv2D('conv1', NF * 2)
-                 .BatchNorm('bn1').LeakyReLU()
                  .Conv2D('conv2', NF * 4)
-                 .BatchNorm('bn2').LeakyReLU()
                  .Conv2D('conv3', NF * 8, stride=1, padding='VALID')
-                 .BatchNorm('bn3').LeakyReLU()
-                 .Conv2D('convlast', 1, stride=1, padding='VALID')())
+                 .Conv2D('convlast', 1, stride=1, padding='VALID', nl=tf.identity)())
         return l
 
     def _build_graph(self, inputs):