[Cleanup][B-8] Replace to_variable (#61581)

---------

Co-authored-by: Nyakku Shigure <sigure.qaq@gmail.com>

test/legacy_test/test_functional_conv1d_transpose.py

@@ -35,12 +35,12 @@ def setUp(self):
 
     def dygraph_case(self):
         with dg.guard():
-            x = dg.to_variable(self.input, dtype=paddle.float32)
-            w = dg.to_variable(self.filter, dtype=paddle.float32)
+            x = paddle.to_tensor(self.input, dtype=paddle.float32)
+            w = paddle.to_tensor(self.filter, dtype=paddle.float32)
             b = (
                 None
                 if self.bias is None
-                else dg.to_variable(self.bias, dtype=paddle.float32)
+                else paddle.to_tensor(self.bias, dtype=paddle.float32)
             )
             y = F.conv1d_transpose(
                 x,

test/legacy_test/test_functional_conv2d.py

@@ -156,9 +156,9 @@ def static_graph_case_2(self):
 
     def dygraph_case(self):
         with dg.guard(self.place):
-            x = dg.to_variable(self.input)
-            weight = dg.to_variable(self.weight)
-            bias = None if self.no_bias else dg.to_variable(self.bias)
+            x = paddle.to_tensor(self.input)
+            weight = paddle.to_tensor(self.weight)
+            bias = None if self.no_bias else paddle.to_tensor(self.bias)
             y = F.conv2d(
                 x,
                 weight,
@@ -534,12 +534,12 @@ def static_graph_case(self):
 
     def dygraph_case(self):
         with dg.guard():
-            x = dg.to_variable(self.input, dtype=paddle.float32)
-            w = dg.to_variable(self.filter, dtype=paddle.float32)
+            x = paddle.to_tensor(self.input, dtype=paddle.float32)
+            w = paddle.to_tensor(self.filter, dtype=paddle.float32)
             b = (
                 None
                 if self.bias is None
-                else dg.to_variable(self.bias, dtype=paddle.float32)
+                else paddle.to_tensor(self.bias, dtype=paddle.float32)
             )
             y = F.conv2d(
                 x,

test/legacy_test/test_functional_conv2d_transpose.py

@@ -154,9 +154,9 @@ def static_graph_case_2(self):
 
     def dygraph_case(self):
         with dg.guard(self.place):
-            x = dg.to_variable(self.input)
-            weight = dg.to_variable(self.weight)
-            bias = None if self.no_bias else dg.to_variable(self.bias)
+            x = paddle.to_tensor(self.input)
+            weight = paddle.to_tensor(self.weight)
+            bias = None if self.no_bias else paddle.to_tensor(self.bias)
             y = F.conv2d_transpose(
                 x,
                 weight,
@@ -542,12 +542,12 @@ def static_graph_case(self):
 
     def dygraph_case(self):
         with dg.guard():
-            x = dg.to_variable(self.input, dtype=paddle.float32)
-            w = dg.to_variable(self.filter, dtype=paddle.float32)
+            x = paddle.to_tensor(self.input, dtype=paddle.float32)
+            w = paddle.to_tensor(self.filter, dtype=paddle.float32)
             b = (
                 None
                 if self.bias is None
-                else dg.to_variable(self.bias, dtype=paddle.float32)
+                else paddle.to_tensor(self.bias, dtype=paddle.float32)
             )
             y = F.conv2d_transpose(
                 x,

test/legacy_test/test_functional_conv3d.py

@@ -156,9 +156,9 @@ def static_graph_case_2(self):
 
     def dygraph_case(self):
         with dg.guard(self.place):
-            x = dg.to_variable(self.input)
-            weight = dg.to_variable(self.weight)
-            bias = None if self.no_bias else dg.to_variable(self.bias)
+            x = paddle.to_tensor(self.input)
+            weight = paddle.to_tensor(self.weight)
+            bias = None if self.no_bias else paddle.to_tensor(self.bias)
             y = F.conv3d(
                 x,
                 weight,
@@ -509,12 +509,12 @@ def static_graph_case(self):
 
     def dygraph_case(self):
         with dg.guard():
-            x = dg.to_variable(self.input, dtype=paddle.float32)
-            w = dg.to_variable(self.filter, dtype=paddle.float32)
+            x = paddle.to_tensor(self.input, dtype=paddle.float32)
+            w = paddle.to_tensor(self.filter, dtype=paddle.float32)
             b = (
                 None
                 if self.bias is None
-                else dg.to_variable(self.bias, dtype=paddle.float32)
+                else paddle.to_tensor(self.bias, dtype=paddle.float32)
             )
             y = F.conv3d(
                 x,

test/legacy_test/test_functional_conv3d_transpose.py

@@ -157,9 +157,9 @@ def static_graph_case_2(self):
 
     def dygraph_case(self):
         with dg.guard(self.place):
-            x = dg.to_variable(self.input)
-            weight = dg.to_variable(self.weight)
-            bias = None if self.no_bias else dg.to_variable(self.bias)
+            x = paddle.to_tensor(self.input)
+            weight = paddle.to_tensor(self.weight)
+            bias = None if self.no_bias else paddle.to_tensor(self.bias)
             y = F.conv3d_transpose(
                 x,
                 weight,
@@ -567,12 +567,12 @@ def static_graph_case(self):
 
     def dygraph_case(self):
         with dg.guard():
-            x = dg.to_variable(self.input, dtype=paddle.float32)
-            w = dg.to_variable(self.filter, dtype=paddle.float32)
+            x = paddle.to_tensor(self.input, dtype=paddle.float32)
+            w = paddle.to_tensor(self.filter, dtype=paddle.float32)
             b = (
                 None
                 if self.bias is None
-                else dg.to_variable(self.bias, dtype=paddle.float32)
+                else paddle.to_tensor(self.bias, dtype=paddle.float32)
             )
             y = F.conv3d_transpose(
                 x,

test/legacy_test/test_gather_nd_op.py

@@ -695,8 +695,8 @@ def test_imperative(self):
         paddle.disable_static()
         input_1 = np.array([[1, 2], [3, 4], [5, 6]])
         index_1 = np.array([[1]])
-        input = base.dygraph.to_variable(input_1)
-        index = base.dygraph.to_variable(index_1)
+        input = paddle.to_tensor(input_1)
+        index = paddle.to_tensor(index_1)
         output = paddle.gather(input, index)
         output_np = output.numpy()
         expected_output = np.array([[3, 4]])

test/legacy_test/test_gelu_op.py

@@ -45,7 +45,7 @@ def _test_case1_cpu(self, approximate):
 
         place = base.CPUPlace()
         with dg.guard(place) as g:
-            x_var = dg.to_variable(x)
+            x_var = paddle.to_tensor(x)
             y_var = F.gelu(x_var, approximate)
             y_test = y_var.numpy()
         np.testing.assert_allclose(y_ref, y_test, rtol=1e-05, atol=1e-08)
@@ -56,7 +56,7 @@ def _test_case1_gpu(self, approximate):
 
         place = base.CUDAPlace(0)
         with dg.guard(place) as g:
-            x_var = dg.to_variable(x)
+            x_var = paddle.to_tensor(x)
             y_var = F.gelu(x_var, approximate)
             y_test = y_var.numpy()
         np.testing.assert_allclose(y_ref, y_test, rtol=1e-05, atol=1e-08)

test/legacy_test/test_grad_clip_minimize.py

@@ -16,8 +16,8 @@
 
 import numpy as np
 
+import paddle
 from paddle import base
-from paddle.base.dygraph.base import to_variable
 from paddle.nn import ClipGradByGlobalNorm, ClipGradByNorm, ClipGradByValue
 
 
@@ -64,8 +64,8 @@ def get_dygrap_global_norm_result(self):
             gloabl_norm_clip = ClipGradByGlobalNorm(self.max_global_norm)
             p_g_var = []
             for p, g in self.para_and_grad:
-                new_p = to_variable(p)
-                new_g = to_variable(g)
+                new_p = paddle.to_tensor(p)
+                new_g = paddle.to_tensor(g)
                 p_g_var.append((new_p, new_g))
 
             new_p_g_var = gloabl_norm_clip(p_g_var)
@@ -136,8 +136,8 @@ def get_dygrap_norm_result(self):
             norm_clip = ClipGradByNorm(self.max_norm)
             p_g_var = []
             for p, g in self.para_and_grad:
-                new_p = to_variable(p)
-                new_g = to_variable(g)
+                new_p = paddle.to_tensor(p)
+                new_g = paddle.to_tensor(g)
                 p_g_var.append((new_p, new_g))
 
             new_p_g_var = norm_clip(p_g_var)
@@ -204,8 +204,8 @@ def get_dygrap_clip_result(self):
             value_clip = ClipGradByValue(max=self.max_value, min=self.min_value)
             p_g_var = []
             for p, g in self.para_and_grad:
-                new_p = to_variable(p)
-                new_g = to_variable(g)
+                new_p = paddle.to_tensor(p)
+                new_g = paddle.to_tensor(g)
                 p_g_var.append((new_p, new_g))
 
             new_p_g_var = value_clip(p_g_var)

test/legacy_test/test_gradient_clip.py

@@ -700,7 +700,7 @@ def test_gradient_clip(self):
         with base.dygraph.guard():
             linear = paddle.nn.Linear(5, 5)
             inputs = paddle.uniform([16, 5], min=-10, max=10).astype('float32')
-            out = linear(base.dygraph.to_variable(inputs))
+            out = linear(paddle.to_tensor(inputs))
             loss = paddle.mean(out)
             loss.backward()
             sgd_optimizer = paddle.optimizer.SGD(
@@ -722,12 +722,8 @@ def setUp(self):
 
     def check_clip_result(self, loss, optimizer):
         # if grad is None
-        x = base.dygraph.to_variable(
-            np.array([2, 3]).astype("float32"), name="x"
-        )
-        y = base.dygraph.to_variable(
-            np.array([3, 4]).astype("float32"), name="y"
-        )
+        x = paddle.to_tensor(np.array([2, 3]).astype("float32"))
+        y = paddle.to_tensor(np.array([3, 4]).astype("float32"))
         assert len(self.clip1([(x, x), (x, y), (x, None)])) == 2
         # get params and grads from network
         opt, params_grads = optimizer.minimize(loss)
@@ -762,10 +758,10 @@ def setUp(self):
 
     def check_clip_result(self, loss, optimizer):
         # if grad is None
-        x = base.dygraph.to_variable(np.array([2, 3]).astype("float32"))
+        x = paddle.to_tensor(np.array([2, 3]).astype("float32"))
         assert len(self.clip([(x, None)])) == 0
         # get params and grads from network
-        self.clip([(base.dygraph.to_variable(np.array([2, 3])), None)])
+        self.clip([(paddle.to_tensor(np.array([2, 3])), None)])
         opt, params_grads = optimizer.minimize(loss)
         _, grads = zip(*params_grads)
         params_grads = self.clip(params_grads)
@@ -791,7 +787,7 @@ def setUp(self):
 
     def check_clip_result(self, loss, optimizer):
         # if grad is None
-        x = base.dygraph.to_variable(np.array([2, 3]).astype("float32"))
+        x = paddle.to_tensor(np.array([2, 3]).astype("float32"))
         assert len(self.clip([(x, None)])) == 0
         # get params and grads from network
         opt, params_grads = optimizer.minimize(loss)
@@ -839,7 +835,7 @@ def test_gradient_clip(self):
                     'float32'
                 )
                 with paddle.amp.auto_cast(level='O2'):
-                    out = model(base.dygraph.to_variable(inputs))
+                    out = model(paddle.to_tensor(inputs))
                     loss = paddle.mean(out)
                 scaled = scaler.scale(loss)
                 scaled.backward()
@@ -884,7 +880,7 @@ def test_gradient_clip(self):
         with base.dygraph.guard():
             inputs = paddle.uniform([16, 5], min=-10, max=10).astype('float32')
             linear = paddle.nn.Linear(5, 5)
-            out = linear(base.dygraph.to_variable(inputs))
+            out = linear(paddle.to_tensor(inputs))
             loss = paddle.mean(out)
             loss.backward()
             # before clip

test/legacy_test/test_group_norm_op.py

@@ -589,12 +589,12 @@ def test_dygraph_api(self):
         input = np.random.random(self.shape).astype(self.dtype)
 
         with base.dygraph.guard():
-            tensor_1 = base.dygraph.to_variable(input)
+            tensor_1 = paddle.to_tensor(input)
             tensor_1.stop_gradient = False
             groupNorm = paddle.nn.GroupNorm(num_channels=32, num_groups=4)
             ret1 = groupNorm(tensor_1)
             ret1.backward()
-            tensor_eager_1 = base.dygraph.to_variable(input)
+            tensor_eager_1 = paddle.to_tensor(input)
             tensor_eager_1.stop_gradient = False
             groupNorm_eager = paddle.nn.GroupNorm(num_channels=32, num_groups=4)
             ret2 = groupNorm_eager(tensor_eager_1)
@@ -609,12 +609,12 @@ def test_dygraph_api(self):
         input = np.random.random(self.shape).astype(self.dtype)
 
         with base.dygraph.guard():
-            tensor_1 = base.dygraph.to_variable(input)
+            tensor_1 = paddle.to_tensor(input)
             tensor_1.stop_gradient = False
             groupNorm = paddle.nn.GroupNorm(num_channels=32, num_groups=4)
             ret1 = groupNorm(tensor_1)
             ret1.backward()
-            tensor_eager_1 = base.dygraph.to_variable(input)
+            tensor_eager_1 = paddle.to_tensor(input)
             tensor_eager_1.stop_gradient = False
             groupNorm_eager = paddle.nn.GroupNorm(num_channels=32, num_groups=4)
             ret2 = groupNorm_eager(tensor_eager_1)
@@ -635,12 +635,12 @@ def test_dygraph_api(self):
         input = np.random.random(self.shape).astype(self.dtype)
 
         with base.dygraph.guard():
-            tensor_1 = base.dygraph.to_variable(input)
+            tensor_1 = paddle.to_tensor(input)
             tensor_1.stop_gradient = False
             groupNorm = paddle.nn.GroupNorm(num_channels=32, num_groups=4)
             ret1 = groupNorm(tensor_1)
             ret1.backward()
-            tensor_eager_1 = base.dygraph.to_variable(input)
+            tensor_eager_1 = paddle.to_tensor(input)
             tensor_eager_1.stop_gradient = False
             groupNorm_eager = paddle.nn.GroupNorm(num_channels=32, num_groups=4)
             ret2 = groupNorm_eager(tensor_eager_1)