Fix

jointContribution/XPINNs/XPINN_2D_PoissonsEqn.py

@@ -10,6 +10,8 @@
 from matplotlib import patches
 from matplotlib import tri
 
+import ppsci
+
 # For the use of the second derivative: paddle.cos, paddle.exp
 paddle.framework.core.set_prim_eager_enabled(True)
 
@@ -32,7 +34,7 @@ def __init__(self, layer_list):
             layer_list[2], "layers3"
         )
 
-    def convert_tensor(self, dataset):
+    def preprocess_data(self, dataset):
         X_ub, ub, X_f1, X_f2, X_f3, X_fi1, X_fi2 = dataset
         self.x_ub = paddle.to_tensor(X_ub[:, 0:1], dtype=paddle.float64)
         self.y_ub = paddle.to_tensor(X_ub[:, 1:2], dtype=paddle.float64)
@@ -49,7 +51,7 @@ def convert_tensor(self, dataset):
         self.y_fi2 = paddle.to_tensor(X_fi2[:, 1:2], dtype=paddle.float64)
 
     def forward(self, dataset):
-        self.convert_tensor(dataset)
+        self.preprocess_data(dataset)
         self.ub1_pred = self.net_u1(self.x_ub, self.y_ub)
         self.ub2_pred = self.net_u2(self.x_f2, self.y_f2)
         self.ub3_pred = self.net_u3(self.x_f3, self.y_f3)
@@ -119,9 +121,7 @@ def initialize_NN(self, layers, name_prefix):
             W = self.create_parameter(
                 shape=[layers[l], layers[l + 1]],
                 dtype="float64",
-                default_initializer=paddle.nn.initializer.XavierNormal(
-                    layers[l], layers[l + 1]
-                ),
+                default_initializer=self.w_init((layers[l], layers[l + 1])),
             )
             b = self.create_parameter(
                 shape=[1, layers[l + 1]],
@@ -135,6 +135,7 @@ def initialize_NN(self, layers, name_prefix):
                 is_bias=True,
                 default_initializer=paddle.nn.initializer.Constant(0.05),
             )
+
             self.add_parameter(name_prefix + "_W_" + str(l), W)
             self.add_parameter(name_prefix + "_b_" + str(l), b)
             self.add_parameter(name_prefix + "_a_" + str(l), a)
@@ -143,6 +144,14 @@ def initialize_NN(self, layers, name_prefix):
             A.append(a)
         return weights, biases, A
 
+    def w_init(self, size):
+        in_dim = size[0]
+        out_dim = size[1]
+        xavier_stddev = np.sqrt(2 / (in_dim + out_dim))
+        param = paddle.empty(size, "float64")
+        param = ppsci.utils.initializer.trunc_normal_(param, 0.0, xavier_stddev)
+        return lambda p_ten, _: p_ten.set_value(param)
+
     def neural_net_tanh(self, X, weights, biases, A):
         num_layers = len(weights) + 1
 
@@ -303,7 +312,7 @@ def train(self, nIter, X_star1, X_star2, X_star3, u_exact2, u_exact3):
             loss = loss1_value + loss2_value + loss3_value
             loss.backward()
             self.optimizer.step()
-            loss.clear_grad()
+            self.optimizer.clear_grad()
 
             if it % 20 == 0:
                 # Predicted solution

ppsci/utils/initializer.py

@@ -91,8 +91,10 @@ def norm_cdf(x):
         _tensor.erfinv_()
 
         # Transform to proper mean, std
-        _tensor = paddle.multiply(_tensor, paddle.to_tensor(std * math.sqrt(2.0)))
-        _tensor = paddle.add(_tensor, paddle.to_tensor(mean))
+        _tensor = paddle.multiply(
+            _tensor, paddle.to_tensor(std * math.sqrt(2.0), _tensor.dtype)
+        )
+        _tensor = paddle.add(_tensor, paddle.to_tensor(mean, _tensor.dtype))
 
         # Clamp to ensure it"s in the proper range
         _tensor = paddle.clip(_tensor, min=a, max=b)