name scope clean-ups in regularization and resnet (#340)

examples/ResNet/imagenet_resnet_utils.py

@@ -157,19 +157,20 @@ def eval_on_ILSVRC12(model, model_file, dataflow):
 
 
 def image_preprocess(image, bgr=True):
-    if image.dtype.base_dtype != tf.float32:
-        image = tf.cast(image, tf.float32)
-    image = image * (1.0 / 255)
-
-    mean = [0.485, 0.456, 0.406]    # rgb
-    std = [0.229, 0.224, 0.225]
-    if bgr:
-        mean = mean[::-1]
-        std = std[::-1]
-    image_mean = tf.constant(mean, dtype=tf.float32)
-    image_std = tf.constant(std, dtype=tf.float32)
-    image = (image - image_mean) / image_std
-    return image
+    with tf.name_scope('image_preprocess'):
+        if image.dtype.base_dtype != tf.float32:
+            image = tf.cast(image, tf.float32)
+        image = image * (1.0 / 255)
+
+        mean = [0.485, 0.456, 0.406]    # rgb
+        std = [0.229, 0.224, 0.225]
+        if bgr:
+            mean = mean[::-1]
+            std = std[::-1]
+        image_mean = tf.constant(mean, dtype=tf.float32)
+        image_std = tf.constant(std, dtype=tf.float32)
+        image = (image - image_mean) / image_std
+        return image
 
 
 def compute_loss_and_error(logits, label):
@@ -177,8 +178,9 @@ def compute_loss_and_error(logits, label):
     loss = tf.reduce_mean(loss, name='xentropy-loss')
 
     def prediction_incorrect(logits, label, topk=1, name='incorrect_vector'):
-        return tf.cast(tf.logical_not(tf.nn.in_top_k(logits, label, topk)),
-                       tf.float32, name=name)
+        with tf.name_scope('prediction_incorrect'):
+            x = tf.logical_not(tf.nn.in_top_k(logits, label, topk))
+        return tf.cast(x, tf.float32, name=name)
 
     wrong = prediction_incorrect(logits, label, 1, name='wrong-top1')
     add_moving_summary(tf.reduce_mean(wrong, name='train-error-top1'))

tensorpack/models/regularize.py

@@ -46,14 +46,15 @@ def regularize_cost(regex, func, name='regularize_cost'):
     # If vars are replicated, only regularize those in the current tower
     params = ctx.filter_vars_by_vs_name(params)
 
-    costs = []
-    for p in params:
-        para_name = p.name
-        if re.search(regex, para_name):
-            costs.append(func(p))
-            _log_regularizer(para_name)
-    if not costs:
-        return tf.constant(0, dtype=tf.float32, name='empty_' + name)
+    with tf.name_scope('regularize_cost'):
+        costs = []
+        for p in params:
+            para_name = p.name
+            if re.search(regex, para_name):
+                costs.append(func(p))
+                _log_regularizer(para_name)
+        if not costs:
+            return tf.constant(0, dtype=tf.float32, name='empty_' + name)
     return tf.add_n(costs, name=name)