diff --git a/dygraph/configs/fcn_hrnet/fcn_hrnetw18_cityscapes_1024x512_100k.yml b/dygraph/configs/fcn_hrnet/fcn_hrnetw18_cityscapes_1024x512_100k.yml
index 7f9cfed0f5fcf0cf904f8e3cbab25e2275d6468d..fb4c531bb4e6cbb498988be9685c95415873f610 100644
--- a/dygraph/configs/fcn_hrnet/fcn_hrnetw18_cityscapes_1024x512_100k.yml
+++ b/dygraph/configs/fcn_hrnet/fcn_hrnetw18_cityscapes_1024x512_100k.yml
@@ -7,3 +7,6 @@ model:
   num_classes: 19
   backbone_channels: [270]
   backbone_pretrained: pretrained_model/hrnet_w18_imagenet
+
+optimizer:
+  weight_decay: 0.0005
diff --git a/dygraph/utils/config.py b/dygraph/utils/config.py
index e0577a6e0eaaa5353c680bfded30094cacd969ab..4b1cf410e8c4baebcb3131c098d4ca1267defba5 100644
--- a/dygraph/utils/config.py
+++ b/dygraph/utils/config.py
@@ -137,6 +137,12 @@ class Config(object):
             lr = self.learning_rate
             args = self.optimizer_args
             args.setdefault('momentum', 0.9)
+            weight_decay = args.get('weight_decay', None)
+            if weight_decay is not None:
+                args.pop('weight_decay')
+                regularization = fluid.regularizer.L2DecayRegularizer(
+                    regularization_coeff=weight_decay)
+            args.setdefault('regularization', None)
             return fluid.optimizer.Momentum(
                 lr, parameter_list=self.model.parameters(), **args)
         else: