Backend paddle: Support deeponet and other examples

deepxde/icbc/boundary_conditions.py

@@ -175,13 +175,11 @@ class PointSetBC:
         component: Integer or a list of integers. The output components satisfying this BC.
             List of integers only supported for the backend PyTorch.
         batch_size: The number of points per minibatch, or `None` to return all points.
-            This is only supported for the backend PyTorch.
+            This is only supported for the backend PyTorch and PaddlePaddle.
         shuffle: Randomize the order on each pass through the data when batching.
     """
 
-    def __init__(
-        self, points, values, component=0, batch_size=None, shuffle=True
-    ):
+    def __init__(self, points, values, component=0, batch_size=None, shuffle=True):
         self.points = np.array(points, dtype=config.real(np))
         self.values = bkd.as_tensor(values, dtype=config.real(bkd.lib))
         self.component = component
@@ -193,13 +191,11 @@ def __init__(
         self.batch_size = batch_size
 
         if batch_size is not None:  # batch iterator and state
-            if backend_name != "pytorch":
+            if backend_name not in ["pytorch", "paddle"]:
                 raise RuntimeError(
-                    "batch_size only implemented for pytorch backend"
+                    "batch_size only implemented for pytorch and paddle backend"
                 )
-            self.batch_sampler = data.sampler.BatchSampler(
-                len(self), shuffle=shuffle
-            )
+            self.batch_sampler = data.sampler.BatchSampler(len(self), shuffle=shuffle)
             self.batch_indices = None
 
     def __len__(self):
@@ -218,15 +214,9 @@ def error(self, X, inputs, outputs, beg, end, aux_var=None):
                     outputs[beg:end, self.component : self.component + 1]
                     - self.values[self.batch_indices]
                 )
-            return (
-                outputs[beg:end, self.component]
-                - self.values[self.batch_indices]
-            )
+            return outputs[beg:end, self.component] - self.values[self.batch_indices]
         if isinstance(self.component, numbers.Number):
-            return (
-                outputs[beg:end, self.component : self.component + 1]
-                - self.values
-            )
+            return outputs[beg:end, self.component : self.component + 1] - self.values
         # When a concat is provided, the following code works 'fast' in paddle cpu,
         # and slow in both tensorflow backends, jax untested.
         # tf.gather can be used instead of for loop but is also slow

deepxde/nn/paddle/deeponet.py

@@ -57,8 +57,7 @@ def __init__(
         if use_bias:
             # register bias to parameter for updating in optimizer and storage
             self.b = self.create_parameter(
-                shape=(1, ),
-                default_initializer=initializers.get("zeros")
+                shape=(1,), default_initializer=initializers.get("zeros")
             )
 
     def forward(self, inputs):
@@ -75,8 +74,12 @@ def forward(self, inputs):
             raise AssertionError(
                 "Output sizes of branch net and trunk net do not match."
             )
-        x = paddle.einsum("bi,bi->b", x_func, x_loc)  # [batch_size, ]
-        x = paddle.reshape(x, [-1, 1])  # reshape [batch_size, ] to [batch_size, 1]
+        # Use the following formula to temporarily replace paddle.einsum()
+        # Because no higher(>=2) orderderivatives for the op now
+        x = paddle.sum(x_func * x_loc, axis=1, keepdim=True)
+        # TODO:
+        # x = paddle.einsum("bi,bi->b", x_func, x_loc)  # [batch_size, ]
+        # x = paddle.reshape(x, [-1, 1])  # reshape [batch_size, ] to [batch_size, 1]
         # Add bias
         if self.use_bias:
             x += self.b
@@ -124,8 +127,7 @@ def __init__(
         self.trunk = FNN(layer_sizes_trunk, self.activation_trunk, kernel_initializer)
         # register bias to parameter for updating in optimizer and storage
         self.b = self.create_parameter(
-            shape=(1, ),
-            default_initializer=initializers.get("zeros")
+            shape=(1,), default_initializer=initializers.get("zeros")
         )
         self.regularizer = regularization
 
@@ -143,7 +145,11 @@ def forward(self, inputs):
             raise AssertionError(
                 "Output sizes of branch net and trunk net do not match."
             )
-        x = paddle.einsum("bi,ni->bn", x_func, x_loc)
+        # Use the following formula to temporarily replace paddle.einsum()
+        # Because no higher(>=2) orderderivatives for the op now
+        x = x_func @ x_loc.T
+        # TODO:
+        # x = paddle.einsum("bi,ni->bn", x_func, x_loc)
         # Add bias
         x += self.b
 

examples/operator/advection_aligned_pideeponet.py

@@ -1,8 +1,23 @@
-"""Backend supported: tensorflow.compat.v1, tensorflow"""
+"""Backend supported: tensorflow.compat.v1, tensorflow, paddle"""
 import deepxde as dde
 import matplotlib.pyplot as plt
 import numpy as np
-from deepxde.backend import tf
+
+if dde.backend.backend_name == "paddle":
+    import paddle
+
+    dim_x = 5
+    sin = paddle.sin
+    cos = paddle.cos
+    concat = paddle.concat
+else:
+    import tf
+
+    dim_x = 2
+    sin = tf.sin
+    cos = tf.cos
+    concat = tf.concat
+
 
 # PDE
 def pde(x, y, v):
@@ -36,7 +51,7 @@ def func_ic(x, v):
 # Net
 net = dde.nn.DeepONetCartesianProd(
     [50, 128, 128, 128],
-    [2, 128, 128, 128],
+    [dim_x, 128, 128, 128],
     "tanh",
     "Glorot normal",
 )
@@ -45,9 +60,7 @@ def func_ic(x, v):
 def periodic(x):
     x, t = x[:, :1], x[:, 1:]
     x *= 2 * np.pi
-    return tf.concat(
-        [tf.math.cos(x), tf.math.sin(x), tf.math.cos(2 * x), tf.math.sin(2 * x), t], 1
-    )
+    return concat([cos(x), sin(x), cos(2 * x), sin(2 * x), t], 1)
 
 
 net.apply_feature_transform(periodic)

examples/operator/advection_aligned_pideeponet_2d.py

@@ -1,8 +1,23 @@
-"""Backend supported: tensorflow.compat.v1, tensorflow"""
+"""Backend supported: tensorflow.compat.v1, tensorflow, paddle"""
 import deepxde as dde
 import matplotlib.pyplot as plt
 import numpy as np
-from deepxde.backend import tf
+
+if dde.backend.backend_name == "paddle":
+    import paddle
+
+    dim_x = 5
+    sin = paddle.sin
+    cos = paddle.cos
+    concat = paddle.concat
+else:
+    import tf
+
+    dim_x = 2
+    sin = tf.sin
+    cos = tf.cos
+    concat = tf.concat
+
 
 # PDE
 def pde(x, y, v):
@@ -41,7 +56,7 @@ def boundary(x, on_boundary):
 # Net
 net = dde.nn.DeepONetCartesianProd(
     [50, 128, 128, 128],
-    [2, 128, 128, 128],
+    [dim_x, 128, 128, 128],
     "tanh",
     "Glorot normal",
 )
@@ -50,9 +65,7 @@ def boundary(x, on_boundary):
 def periodic(x):
     x, t = x[:, :1], x[:, 1:]
     x *= 2 * np.pi
-    return tf.concat(
-        [tf.math.cos(x), tf.math.sin(x), tf.math.cos(2 * x), tf.math.sin(2 * x), t], 1
-    )
+    return concat([cos(x), sin(x), cos(2 * x), sin(2 * x), t], 1)
 
 
 net.apply_feature_transform(periodic)

examples/operator/advection_unaligned_pideeponet.py

@@ -1,8 +1,23 @@
-"""Backend supported: tensorflow.compat.v1"""
+"""Backend supported: tensorflow.compat.v1, paddle"""
 import deepxde as dde
 import matplotlib.pyplot as plt
 import numpy as np
-from deepxde.backend import tf
+
+if dde.backend.backend_name == "paddle":
+    import paddle
+
+    dim_x = 5
+    sin = paddle.sin
+    cos = paddle.cos
+    concat = paddle.concat
+else:
+    import tf
+
+    dim_x = 2
+    sin = tf.sin
+    cos = tf.cos
+    concat = tf.concat
+
 
 # PDE
 def pde(x, y, v):
@@ -36,7 +51,7 @@ def func_ic(x, v):
 # Net
 net = dde.nn.DeepONet(
     [50, 128, 128, 128],
-    [2, 128, 128, 128],
+    [dim_x, 128, 128, 128],
     "tanh",
     "Glorot normal",
 )
@@ -45,9 +60,7 @@ def func_ic(x, v):
 def periodic(x):
     x, t = x[:, :1], x[:, 1:]
     x *= 2 * np.pi
-    return tf.concat(
-        [tf.math.cos(x), tf.math.sin(x), tf.math.cos(2 * x), tf.math.sin(2 * x), t], 1
-    )
+    return concat([cos(x), sin(x), cos(2 * x), sin(2 * x), t], 1)
 
 
 net.apply_feature_transform(periodic)

examples/operator/advection_unaligned_pideeponet_2d.py

@@ -1,8 +1,23 @@
-"""Backend supported: tensorflow.compat.v1"""
+"""Backend supported: tensorflow.compat.v1, paddle"""
 import deepxde as dde
 import matplotlib.pyplot as plt
 import numpy as np
-from deepxde.backend import tf
+
+if dde.backend.backend_name == "paddle":
+    import paddle
+
+    dim_x = 5
+    sin = paddle.sin
+    cos = paddle.cos
+    concat = paddle.concat
+else:
+    import tf
+
+    dim_x = 2
+    sin = tf.sin
+    cos = tf.cos
+    concat = tf.concat
+
 
 # PDE
 def pde(x, y, v):
@@ -39,7 +54,7 @@ def boundary(x, on_boundary):
 # Net
 net = dde.nn.DeepONet(
     [50, 128, 128, 128],
-    [2, 128, 128, 128],
+    [dim_x, 128, 128, 128],
     "tanh",
     "Glorot normal",
 )
@@ -48,9 +63,7 @@ def boundary(x, on_boundary):
 def periodic(x):
     x, t = x[:, :1], x[:, 1:]
     x *= 2 * np.pi
-    return tf.concat(
-        [tf.math.cos(x), tf.math.sin(x), tf.math.cos(2 * x), tf.math.sin(2 * x), t], 1
-    )
+    return concat([cos(x), sin(x), cos(2 * x), sin(2 * x), t], 1)
 
 
 net.apply_feature_transform(periodic)

examples/operator/antiderivative_aligned_pideeponet.py

@@ -1,8 +1,16 @@
-"""Backend supported: tensorflow.compat.v1, tensorflow"""
+"""Backend supported: tensorflow.compat.v1, tensorflow, paddle"""
 import deepxde as dde
 import matplotlib.pyplot as plt
 import numpy as np
-from deepxde.backend import tf
+
+if dde.backend.backend_name == "paddle":
+    import paddle
+
+    transpose = paddle.transpose
+else:
+    import tf
+
+    transpose = tf.transpose
 
 
 dde.config.disable_xla_jit()
@@ -35,9 +43,10 @@ def pde(x, u, v):
     "Glorot normal",
 )
 
+
 # Hard constraint zero IC
 def zero_ic(inputs, outputs):
-    return outputs * tf.transpose(inputs[1])
+    return outputs * transpose(inputs[1], [1, 0])
 
 
 net.apply_output_transform(zero_ic)

examples/operator/antiderivative_unaligned_pideeponet.py

@@ -1,4 +1,4 @@
-"""Backend supported: tensorflow.compat.v1"""
+"""Backend supported: tensorflow.compat.v1, paddle"""
 import deepxde as dde
 import matplotlib.pyplot as plt
 import numpy as np
@@ -30,6 +30,7 @@ def pde(x, u, v):
     "Glorot normal",
 )
 
+
 # Hard constraint zero IC
 def zero_ic(inputs, outputs):
     return outputs * inputs[1]

examples/operator/diff_rec_aligned_pideeponet.py

@@ -1,4 +1,4 @@
-"""Backend supported: tensorflow.compat.v1, tensorflow"""
+"""Backend supported: tensorflow.compat.v1, tensorflow, paddle"""
 import deepxde as dde
 import matplotlib.pyplot as plt
 import numpy as np

examples/operator/diff_rec_unaligned_pideeponet.py

@@ -1,4 +1,4 @@
-"""Backend supported: tensorflow.compat.v1"""
+"""Backend supported: tensorflow.compat.v1, paddle"""
 import deepxde as dde
 import matplotlib.pyplot as plt
 import numpy as np

examples/pinn_forward/Burgers.py

@@ -1,4 +1,4 @@
-"""Backend supported: tensorflow.compat.v1, tensorflow, pytorch"""
+"""Backend supported: tensorflow.compat.v1, tensorflow, pytorch, paddle"""
 import deepxde as dde
 import numpy as np
 

examples/pinn_inverse/elliptic_inverse_field_batch.py

@@ -1,4 +1,4 @@
-"""Backend supported: tensorflow.compat.v1, pytorch"""
+"""Backend supported: tensorflow.compat.v1, pytorch, paddle"""
 import deepxde as dde
 import matplotlib.pyplot as plt
 import numpy as np
@@ -57,7 +57,7 @@ def sol(x):
 plt.plot(x, uhat, "--", label="u_NN")
 plt.legend()
 
-qtrue = -np.pi ** 2 * np.sin(np.pi * x)
+qtrue = -np.pi**2 * np.sin(np.pi * x)
 print("l2 relative error for q: " + str(dde.metrics.l2_relative_error(qtrue, qhat)))
 plt.figure()
 plt.plot(x, qtrue, "-", label="q_true")