[Fix]Fix arch.freeze/unfreeze and change eval of control_arm example (P…

…addlePaddle#673)
huohuohuohuohuo123 · Nov 30, 2023 · 7772ef9 · 7772ef9
1 parent b1cc770
commit 7772ef9
Show file tree

Hide file tree

Showing 3 changed files with 5 additions and 5 deletions.
diff --git a/docs/zh/examples/control_arm.md b/docs/zh/examples/control_arm.md
@@ -48,7 +48,7 @@
 
 | 预训练模型  | 指标 |
 |:--| :--|
-| [inverse_x_axis_pretrained.pdparams](https://paddle-org.bj.bcebos.com/paddlescience/models/control_arm/inverse_x_axis_pretrained.pdparams) | loss(geo_eval): 0.02505<br>MAE.lambda_(geo_eval): 0.01580<br>MAE.mu(geo_eval): 0.01984 |
+| [inverse_x_axis_pretrained.pdparams](https://paddle-org.bj.bcebos.com/paddlescience/models/control_arm/inverse_x_axis_pretrained.pdparams) | loss(geo_eval): 0.02505<br>L2Rel.lambda_(geo_eval): 0.06025<br>L2Rel.mu(geo_eval): 0.07949 |
 
 ## 1. 背景简介
 

diff --git a/examples/control_arm/inverse_parameter.py b/examples/control_arm/inverse_parameter.py
@@ -107,7 +107,7 @@ def train(cfg: DictConfig):
             "batch_size": cfg.EVAL.batch_size.validator,
         },
         ppsci.loss.MSELoss("sum"),
-        metric={"MAE": ppsci.metric.MAE()},
+        metric={"L2Rel": ppsci.metric.L2Rel()},
         name="geo_eval",
     )
     validator = {geom_validator.name: geom_validator}
@@ -216,7 +216,7 @@ def evaluate(cfg: DictConfig):
             "batch_size": cfg.EVAL.batch_size.validator,
         },
         ppsci.loss.MSELoss("sum"),
-        metric={"MAE": ppsci.metric.MAE()},
+        metric={"L2Rel": ppsci.metric.L2Rel()},
         name="geo_eval",
     )
     validator = {geom_validator.name: geom_validator}

diff --git a/ppsci/arch/base.py b/ppsci/arch/base.py
@@ -126,14 +126,14 @@ def register_output_transform(
 
     def freeze(self):
         """Freeze all parameters."""
-        for param in self.named_parameters():
+        for param in self.parameters():
             param.stop_gradient = True
 
         self.eval()
 
     def unfreeze(self):
         """Unfreeze all parameters."""
-        for param in self.named_parameters():
+        for param in self.parameters():
             param.stop_gradient = False
 
         self.train()