[rllib] format with yapf (#2427)

* initial yapf

* manual fix yapf bugs

python/ray/rllib/models/visionnet.py

@@ -22,14 +22,27 @@ class VisionNetwork(Model):
         with tf.name_scope("vision_net"):
             for i, (out_size, kernel, stride) in enumerate(filters[:-1], 1):
                 inputs = slim.conv2d(
-                    inputs, out_size, kernel, stride,
-                    activation_fn=activation, scope="conv{}".format(i))
+                    inputs,
+                    out_size,
+                    kernel,
+                    stride,
+                    activation_fn=activation,
+                    scope="conv{}".format(i))
             out_size, kernel, stride = filters[-1]
             fc1 = slim.conv2d(
-                inputs, out_size, kernel, stride,
-                activation_fn=activation, padding="VALID", scope="fc1")
-            fc2 = slim.conv2d(fc1, num_outputs, [1, 1], activation_fn=None,
-                              normalizer_fn=None, scope="fc2")
+                inputs,
+                out_size,
+                kernel,
+                stride,
+                activation_fn=activation,
+                padding="VALID",
+                scope="fc1")
+            fc2 = slim.conv2d(
+                fc1,
+                num_outputs, [1, 1],
+                activation_fn=None,
+                normalizer_fn=None,
+                scope="fc2")
             return flatten(fc2), flatten(fc1)