fix bug; test=develop (#17733)

python/paddle/fluid/dygraph/nn.py

@@ -1461,8 +1461,8 @@ class GRUUnit(layers.Layer):
             sigmoid=1,
             tanh=2,
             relu=3, )
-        activation = activation_dict[activation]
-        gate_activation = activation_dict[gate_activation]
+        self.activation = activation_dict[activation]
+        self.gate_activation = activation_dict[gate_activation]
 
         self._dtype = dtype
         size = size // 3
@@ -1494,8 +1494,8 @@ class GRUUnit(layers.Layer):
                 'Hidden': updated_hidden,
             },
             attrs={
-                'activation': 2,  # tanh
-                'gate_activation': 1,  # sigmoid
+                'activation': self.activation,
+                'gate_activation': self.gate_activation,
             })
 
         return updated_hidden, reset_hidden_pre, gate
@@ -2053,7 +2053,7 @@ class Conv2DTranspose(layers.Layer):
             self._filter_size = [filter_size_h, filter_size_w]
         else:
             self._filter_size = utils.convert_to_list(
-                self._output_size, 2, 'conv2d_transpose.filter_size')
+                self._filter_size, 2, 'conv2d_transpose.filter_size')
 
         if self._output_size is None:
             self._output_size = []

python/paddle/fluid/layers/math_op_patch.py

@@ -42,7 +42,9 @@ def monkey_patch_variable():
                 'shape': shape,
                 'value': value,
                 'force_cpu': force_init_on_cpu()
-            })
+            },
+            stop_gradient=True)
+        var.stop_gradient = True
         return var
 
     def create_scalar(block, value, dtype):
@@ -68,7 +70,10 @@ def monkey_patch_variable():
                 'value': value,
                 'input_dim_idx': batch_dim,
                 'output_dim_idx': batch_dim
-            })
+            },
+            stop_gradient=True)
+
+        var.stop_gradient = True
         return var
 
     def astype(self, dtype):