diff --git a/exps/LFNA/basic-same.py b/exps/LFNA/basic-same.py
index 99a7861..b565f4d 100644
--- a/exps/LFNA/basic-same.py
+++ b/exps/LFNA/basic-same.py
@@ -58,6 +58,7 @@ def main(args):
         # build model
         model = get_model(**model_kwargs)
         print(model)
+        model.analyze_weights()
         # build optimizer
         optimizer = torch.optim.Adam(model.parameters(), lr=args.init_lr, amsgrad=True)
         criterion = torch.nn.MSELoss()
@@ -168,7 +169,7 @@ if __name__ == "__main__":
     parser.add_argument(
         "--epochs",
         type=int,
-        default=1000,
+        default=300,
         help="The total number of epochs.",
     )
     parser.add_argument(
diff --git a/lib/models/xcore.py b/lib/models/xcore.py
index 08c03a6..143278c 100644
--- a/lib/models/xcore.py
+++ b/lib/models/xcore.py
@@ -40,13 +40,10 @@ def get_model(config: Dict[Text, Any], **kwargs):
         norm_cls = super_name2norm[kwargs["norm_cls"]]
         sub_layers, last_dim = [], kwargs["input_dim"]
         for i, hidden_dim in enumerate(kwargs["hidden_dims"]):
-            sub_layers.extend(
-                [
-                    norm_cls(last_dim, elementwise_affine=False),
-                    SuperLinear(last_dim, hidden_dim),
-                    act_cls(),
-                ]
-            )
+            if last_dim > 1:
+                sub_layers.append(norm_cls(last_dim, elementwise_affine=False))
+            sub_layers.append(SuperLinear(last_dim, hidden_dim))
+            sub_layers.append(act_cls())
             last_dim = hidden_dim
         sub_layers.append(SuperLinear(last_dim, kwargs["output_dim"]))
         model = SuperSequential(*sub_layers)
diff --git a/lib/xlayers/super_module.py b/lib/xlayers/super_module.py
index 58a6993..c99be99 100644
--- a/lib/xlayers/super_module.py
+++ b/lib/xlayers/super_module.py
@@ -66,6 +66,15 @@ class SuperModule(abc.ABC, nn.Module):
             container.append(name, buf, False)
         return container
 
+    def analyze_weights(self):
+        with torch.no_grad():
+            for name, param in self.named_parameters():
+                shapestr = "[{:10s}] shape={:}".format(name, list(param.shape))
+                finalstr = shapestr + "{:.2f} +- {:.2f}".format(
+                    param.mean(), param.std()
+                )
+                print(finalstr)
+
     @property
     def abstract_search_space(self):
         raise NotImplementedError