[Clean fluid] Clean fluid elementwise_arithmetic (part2) (#48461)

* clean elem_arithmetic part2 unittest

* fix test_model_cast_to_bf16

* restore test_model_cast_to_bf16

python/paddle/fluid/contrib/slim/tests/test_quantization_pass.py

@@ -72,7 +72,7 @@ def residual_block(num, quant_skip_pattern=None):
     for _ in range(num):
         conv = conv_bn_layer(hidden, 16, 3, 1, 1, act=None, bias_attr=True)
         short = conv_bn_layer(hidden, 16, 1, 1, 0, act=None)
-        hidden = fluid.layers.elementwise_add(x=conv, y=short, act='relu')
+        hidden = paddle.nn.functional.relu(paddle.add(x=conv, y=short))
     matmul_weight = paddle.create_parameter(
         shape=[1, 16, 32, 32], dtype='float32'
     )
@@ -723,7 +723,7 @@ def quant_dequant_residual_block(num, quant_skip_pattern=None):
     for _ in range(num):
         conv = conv_bn_layer(hidden, 16, 3, 1, 1, act=None, bias_attr=True)
         short = conv_bn_layer(hidden, 16, 1, 1, 0, act=None)
-        hidden = fluid.layers.elementwise_add(x=conv, y=short, act='relu')
+        hidden = paddle.nn.functional.relu(paddle.add(x=conv, y=short))
     hidden = fluid.layers.matmul(hidden, data2, True, True)
     if isinstance(quant_skip_pattern, str):
         with fluid.name_scope(quant_skip_pattern):
@@ -733,9 +733,7 @@ def quant_dequant_residual_block(num, quant_skip_pattern=None):
             pool2 = fluid.layers.pool2d(
                 input=hidden, pool_size=2, pool_type='max', pool_stride=2
             )
-            pool_add = fluid.layers.elementwise_add(
-                x=pool1, y=pool2, act='relu'
-            )
+            pool_add = paddle.nn.functional.relu(paddle.add(x=pool1, y=pool2))
     elif isinstance(quant_skip_pattern, list):
         assert (
             len(quant_skip_pattern) > 1
@@ -748,9 +746,7 @@ def quant_dequant_residual_block(num, quant_skip_pattern=None):
                 input=hidden, pool_size=2, pool_type='max', pool_stride=2
             )
         with fluid.name_scope(quant_skip_pattern[1]):
-            pool_add = fluid.layers.elementwise_add(
-                x=pool1, y=pool2, act='relu'
-            )
+            pool_add = paddle.nn.functional.relu(paddle.add(x=pool1, y=pool2))
     else:
         pool1 = fluid.layers.pool2d(
             input=hidden, pool_size=2, pool_type='avg', pool_stride=2
@@ -758,7 +754,7 @@ def quant_dequant_residual_block(num, quant_skip_pattern=None):
         pool2 = fluid.layers.pool2d(
             input=hidden, pool_size=2, pool_type='max', pool_stride=2
         )
-        pool_add = fluid.layers.elementwise_add(x=pool1, y=pool2, act='relu')
+        pool_add = paddle.nn.functional.relu(paddle.add(x=pool1, y=pool2))
     fc = fluid.layers.fc(input=pool_add, size=10)
     loss = fluid.layers.cross_entropy(input=fc, label=label)
     loss = paddle.mean(loss)

python/paddle/fluid/contrib/slim/tests/test_user_defined_quantization.py

@@ -73,12 +73,8 @@ def pact(x, name=None):
         learning_rate=1,
     )
     u_param = helper.create_parameter(attr=u_param_attr, shape=[1], dtype=dtype)
-    x = fluid.layers.elementwise_sub(
-        x, fluid.layers.relu(fluid.layers.elementwise_sub(x, u_param))
-    )
-    x = fluid.layers.elementwise_add(
-        x, fluid.layers.relu(fluid.layers.elementwise_sub(-u_param, x))
-    )
+    x = paddle.subtract(x, fluid.layers.relu(paddle.subtract(x, u_param)))
+    x = paddle.add(x, fluid.layers.relu(paddle.subtract(-u_param, x)))
 
     return x
 

python/paddle/fluid/contrib/tests/test_image_classification_fp16.py

@@ -53,7 +53,7 @@ def resnet_cifar10(input, depth=32):
         tmp = conv_bn_layer(input, ch_out, 3, stride, 1)
         tmp = conv_bn_layer(tmp, ch_out, 3, 1, 1, act=None, bias_attr=True)
         short = shortcut(input, ch_in, ch_out, stride)
-        return fluid.layers.elementwise_add(x=tmp, y=short, act='relu')
+        return paddle.nn.functional.relu(paddle.add(x=tmp, y=short))
 
     def layer_warp(block_func, input, ch_in, ch_out, count, stride):
         tmp = block_func(input, ch_in, ch_out, stride)

python/paddle/fluid/contrib/tests/test_multi_precision_fp16_train.py

@@ -71,7 +71,7 @@ def resnet_cifar10(input, depth=32):
         tmp = conv_bn_layer(input, ch_out, 3, stride, 1)
         tmp = conv_bn_layer(tmp, ch_out, 3, 1, 1, act=None, bias_attr=True)
         short = shortcut(input, ch_in, ch_out, stride)
-        return fluid.layers.elementwise_add(x=tmp, y=short, act='relu')
+        return paddle.nn.functional.relu(paddle.add(x=tmp, y=short))
 
     def layer_warp(block_func, input, ch_in, ch_out, count, stride):
         tmp = block_func(input, ch_in, ch_out, stride)

python/paddle/fluid/contrib/tests/test_quantize_transpiler.py

@@ -56,7 +56,7 @@ def residual_block(num):
     for _ in range(num):
         conv = conv_bn_layer(hidden, 16, 3, 1, 1, act=None, bias_attr=True)
         short = conv_bn_layer(hidden, 16, 1, 1, 0, act=None)
-        hidden = fluid.layers.elementwise_add(x=conv, y=short, act='relu')
+        hidden = paddle.nn.functional.relu(paddle.add(x=conv, y=short))
     fc = fluid.layers.fc(input=hidden, size=10)
     loss = fluid.layers.cross_entropy(input=fc, label=label)
     loss = paddle.mean(loss)

python/paddle/fluid/contrib/tests/test_weight_decay_extend.py

@@ -171,9 +171,7 @@ class TestWeightDecay(unittest.TestCase):
             ]
 
             for params in param_list:
-                updated_p = fluid.layers.elementwise_sub(
-                    x=params[0], y=params[1]
-                )
+                updated_p = paddle.subtract(x=params[0], y=params[1])
                 fluid.layers.assign(input=updated_p, output=params[0])
 
             optimizer.apply_optimize(avg_cost, startup_prog, params_grads)

python/paddle/fluid/tests/book/test_image_classification.py

@@ -52,7 +52,7 @@ def resnet_cifar10(input, depth=32):
         tmp = conv_bn_layer(input, ch_out, 3, stride, 1)
         tmp = conv_bn_layer(tmp, ch_out, 3, 1, 1, act=None, bias_attr=True)
         short = shortcut(input, ch_in, ch_out, stride)
-        return fluid.layers.elementwise_add(x=tmp, y=short, act='relu')
+        return paddle.nn.functional.relu(paddle.add(x=tmp, y=short))
 
     def layer_warp(block_func, input, ch_in, ch_out, count, stride):
         tmp = block_func(input, ch_in, ch_out, stride)

python/paddle/fluid/tests/unittests/collective/fleet/hybrid_parallel_pp_embedding.py

@@ -55,7 +55,7 @@ class SimpleNet(Layer):
     def forward(self, x1, x2, y1):
         x_emb = self.word_embeddings(x1)
         fc = fluid.layers.matmul(x_emb, self.softmax_weight)
-        fc = fluid.layers.elementwise_add(fc, self.softmax_bias)
+        fc = paddle.add(fc, self.softmax_bias)
         projection = paddle.reshape(fc, shape=[-1, vocab_size])
         loss = paddle.nn.functional.softmax_with_cross_entropy(
             logits=projection, label=y1, soft_label=False
@@ -95,7 +95,7 @@ class BiasNet(Layer):
 
     def forward(self, args):
         fc, x2 = args
-        fc = fluid.layers.elementwise_add(fc, self.softmax_bias)
+        fc = paddle.add(fc, self.softmax_bias)
         projection = paddle.reshape(fc, shape=[-1, vocab_size])
         return projection, x2
 

python/paddle/fluid/tests/unittests/collective/fleet/hybrid_parallel_shared_weight.py

@@ -62,7 +62,7 @@ class SimpleNet(Layer):
     def forward(self, x1, x2, y1):
         x_emb = self.word_embeddings(x1)
         fc = fluid.layers.matmul(x_emb, self.softmax_weight)
-        fc = fluid.layers.elementwise_add(fc, self.softmax_bias)
+        fc = paddle.add(fc, self.softmax_bias)
         projection = paddle.reshape(fc, shape=[-1, vocab_size])
 
         projection = paddle.matmul(projection, self.word_embeddings.weight)
@@ -109,7 +109,7 @@ class BiasNet(Layer):
 
     def forward(self, args):
         fc, x2 = args
-        fc = fluid.layers.elementwise_add(fc, self.softmax_bias)
+        fc = paddle.add(fc, self.softmax_bias)
         projection = paddle.reshape(fc, shape=[-1, vocab_size])
         return projection, x2
 

python/paddle/fluid/tests/unittests/collective/fleet/parallel_dygraph_se_resnext.py

@@ -206,7 +206,7 @@ class BottleneckBlock(fluid.dygraph.Layer):
         else:
             short = self.short(inputs)
 
-        y = fluid.layers.elementwise_add(x=short, y=scale, act='relu')
+        y = paddle.nn.functional.relu(paddle.add(x=short, y=scale))
         return y
 
 

python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_elt_act_fuse_pass.py

@@ -53,7 +53,7 @@ class ElementwiseActivationMkldnnFusePassTest(InferencePassTest):
         self.enable_mkldnn = True
 
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = None
 
     def test_check_output(self):
@@ -68,7 +68,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_Relu(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = fluid.layers.relu
 
 
@@ -76,7 +76,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_Tanh(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = paddle.tanh
 
 
@@ -84,7 +84,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_LeakyRelu(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act_alpha = 0.2
         self.act = paddle.nn.functional.leaky_relu
 
@@ -93,7 +93,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_Swish(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = paddle.nn.functional.swish
 
 
@@ -101,7 +101,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_HardSwish(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = fluid.layers.hard_swish
 
 
@@ -109,7 +109,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_SQRT(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = paddle.sqrt
 
 
@@ -117,7 +117,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_ABS(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = paddle.abs
 
 
@@ -125,7 +125,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_Clip(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = fluid.layers.clip
         self.act_alpha = 0.0
         self.act_beta = 10.0
@@ -135,7 +135,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_Gelu(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = paddle.nn.functional.gelu
 
 
@@ -143,7 +143,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_Gelu_Tanh(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = paddle.nn.functional.gelu
         self.act_alpha = True
 
@@ -152,7 +152,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_Relu6(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = paddle.nn.functional.relu6
 
 
@@ -160,7 +160,7 @@ class ElementwiseActivationMkldnnFusePassTest_Add_Sigmoid(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_add
+        self.operand = paddle.add
         self.act = paddle.nn.functional.sigmoid
 
 
@@ -168,7 +168,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_Relu(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = fluid.layers.relu
 
 
@@ -176,7 +176,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_Tanh(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = paddle.tanh
 
 
@@ -184,7 +184,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_LeakyRelu(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act_alpha = 0.2
         self.act = paddle.nn.functional.leaky_relu
 
@@ -193,7 +193,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_Swish(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = paddle.nn.functional.swish
 
 
@@ -201,7 +201,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_HardSwish(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = fluid.layers.hard_swish
 
 
@@ -209,7 +209,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_ABS(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = paddle.abs
 
 
@@ -217,7 +217,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_Clip(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = fluid.layers.clip
         self.act_alpha = 0.0
         self.act_beta = 10.0
@@ -227,7 +227,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_Gelu(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = paddle.nn.functional.gelu
 
 
@@ -235,7 +235,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_Gelu_Tanh(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = paddle.nn.functional.gelu
         self.act_alpha = True
 
@@ -244,7 +244,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_Relu6(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = paddle.nn.functional.relu6
 
 
@@ -252,7 +252,7 @@ class ElementwiseActivationMkldnnFusePassTest_Sub_Sigmoid(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_sub
+        self.operand = paddle.subtract
         self.act = paddle.nn.functional.sigmoid
 
 
@@ -260,7 +260,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_Relu(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = fluid.layers.relu
 
 
@@ -268,7 +268,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_Tanh(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = paddle.tanh
 
 
@@ -276,7 +276,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_LeakyRelu(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act_alpha = 0.2
         self.act = paddle.nn.functional.leaky_relu
 
@@ -285,7 +285,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_Swish(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = paddle.nn.functional.swish
 
 
@@ -293,7 +293,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_HardSwish(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = fluid.layers.hard_swish
 
 
@@ -301,7 +301,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_SQRT(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = paddle.sqrt
 
 
@@ -309,7 +309,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_ABS(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = paddle.abs
 
 
@@ -317,7 +317,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_Clip(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = fluid.layers.clip
         self.act_alpha = 0.0
         self.act_beta = 10.0
@@ -327,7 +327,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_Gelu(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = paddle.nn.functional.gelu
 
 
@@ -335,7 +335,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_Gelu_Tanh(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = paddle.nn.functional.gelu
         self.act_alpha = True
 
@@ -344,7 +344,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_Relu6(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = paddle.nn.functional.relu6
 
 
@@ -352,7 +352,7 @@ class ElementwiseActivationMkldnnFusePassTest_Mul_Sigmoid(
     ElementwiseActivationMkldnnFusePassTest
 ):
     def set_params(self):
-        self.operand = fluid.layers.elementwise_mul
+        self.operand = paddle.multiply
         self.act = paddle.nn.functional.sigmoid
 
 

python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_inplace_fuse_pass.py

@@ -34,9 +34,7 @@ class MkldnnInplacePassTest(InferencePassTest):
             )
             softmax_out = paddle.nn.functional.softmax(conv_out_1)
             relu_out = fluid.layers.relu(conv_out_1)
-            eltwise_out = fluid.layers.elementwise_add(
-                softmax_out, relu_out, axis=-1
-            )
+            eltwise_out = paddle.add(softmax_out, relu_out)
 
         self.pass_name = 'mkldnn_inplace_pass'
         self.feeds = {

python/paddle/fluid/tests/unittests/ir/inference/test_trt_subgraph_pass.py

@@ -378,7 +378,7 @@ class TensorRTSubgraphPassElementwiseTest(InferencePassTest):
         self.fetch_list = [out]
 
     def append_eltwise(self, data1, data2):
-        return fluid.layers.elementwise_add(x=data1, y=data2)
+        return paddle.add(x=data1, y=data2)
 
     def test_check_output(self):
         if core.is_compiled_with_cuda():
@@ -439,7 +439,7 @@ class TensorRTSubgraphPassElementwiseBroadcastDynamicTest(InferencePassTest):
         self.fetch_list = [out]
 
     def append_eltwise(self, data1, data2):
-        return fluid.layers.elementwise_add(x=data1, y=data2)
+        return paddle.add(x=data1, y=data2)
 
     def test_check_output(self):
         if os.path.exists(self.path + "_opt_cache"):

python/paddle/fluid/tests/unittests/ir/test_ir_embedding_eltwise_layernorm_fuse_pass.py

@@ -17,6 +17,7 @@ import unittest
 import numpy as np
 from pass_test import PassTest
 
+import paddle
 import paddle.fluid as fluid
 import paddle.fluid.core as core
 
@@ -51,8 +52,8 @@ class EmbEltwiseLayerNormFusePassTest(PassTest):
             sent_emb = fluid.layers.embedding(
                 input=sent_id, size=(128, 768), dtype='float32'
             )
-            add1 = fluid.layers.elementwise_add(word_emb, pos_emb)
-            add2 = fluid.layers.elementwise_add(add1, sent_emb)
+            add1 = paddle.add(word_emb, pos_emb)
+            add2 = paddle.add(add1, sent_emb)
             hidden1 = fluid.layers.layer_norm(input=add2, begin_norm_axis=2)
 
             id1 = fluid.layers.data(
@@ -91,9 +92,9 @@ class EmbEltwiseLayerNormFusePassTest(PassTest):
             emb4 = fluid.layers.embedding(
                 input=id4, size=(128, 768), dtype='float32'
             )
-            add_1 = fluid.layers.elementwise_add(emb1, emb2)
-            add_2 = fluid.layers.elementwise_add(add_1, emb3)
-            add_3 = fluid.layers.elementwise_add(add_2, emb4)
+            add_1 = paddle.add(emb1, emb2)
+            add_2 = paddle.add(add_1, emb3)
+            add_3 = paddle.add(add_2, emb4)
             hidden_1 = fluid.layers.layer_norm(input=add_3, begin_norm_axis=2)
 
         self.feeds = {

python/paddle/fluid/tests/unittests/ir/test_ir_fusion_group_pass.py

@@ -142,7 +142,7 @@ class FusionGroupPassTestCastAndFP16(FusionGroupPassTest):
             zero = layers.fill_constant(shape=[128], dtype="float16", value=0)
             # TODO(xreki): fix precision problem when using softmax of float16.
             # tmp_2 = layers.softmax(tmp_1)
-            tmp_2 = layers.elementwise_add(tmp_1, zero)
+            tmp_2 = paddle.add(tmp_1, zero)
             tmp_3 = layers.mul(tmp_0, self.feed_vars[2])
             # subgraph with 4 op nodes
             tmp_3 = layers.cast(tmp_2, dtype="float16")
@@ -184,7 +184,7 @@ class FusionGroupPassCastTest(FusionGroupPassTest):
         with fluid.program_guard(self.main_program, self.startup_program):
             self.feed_vars = self._prepare_feed_vars([2, 2], dtype, 2)
 
-            tmp_0 = layers.elementwise_add(self.feed_vars[0], self.feed_vars[1])
+            tmp_0 = paddle.add(self.feed_vars[0], self.feed_vars[1])
             tmp_1 = layers.cast(tmp_0, dtype="float64")
             tmp_2 = layers.cast(tmp_1, dtype="float32")
 
@@ -205,12 +205,12 @@ class FusionGroupPassFillConstantTest(FusionGroupPassTest):
         with fluid.program_guard(self.main_program, self.startup_program):
             self.feed_vars = self._prepare_feed_vars([2, 2], dtype, 2)
 
-            tmp_0 = layers.elementwise_add(self.feed_vars[0], self.feed_vars[1])
+            tmp_0 = paddle.add(self.feed_vars[0], self.feed_vars[1])
             tmp_1 = layers.fill_constant(shape=[2, 2], dtype=dtype, value=2.0)
             tmp_2 = paddle.scale(
                 tmp_1, scale=3.0, bias=1.0, bias_after_scale=True
             )
-            tmp_3 = layers.elementwise_mul(tmp_2, tmp_0)
+            tmp_3 = paddle.multiply(tmp_2, tmp_0)
 
         self.append_gradients(tmp_3)
 

python/paddle/fluid/tests/unittests/ir/test_ir_skip_layernorm_pass.py

@@ -31,7 +31,7 @@ class SkipLayerNormFusePassTest(PassTest):
             y = fluid.data(
                 name="y", shape=[128, 768], dtype="float32", lod_level=0
             )
-            elementwise_out = fluid.layers.elementwise_add(x=x, y=y)
+            elementwise_out = paddle.add(x=x, y=y)
             out = fluid.layers.layer_norm(input=elementwise_out)
 
         self.fetch_list = [out]

python/paddle/fluid/tests/unittests/mkldnn/check_flags_mkldnn_ops_on_off.py

@@ -16,6 +16,7 @@ import os
 
 import numpy as np
 
+import paddle
 import paddle.fluid as fluid
 from paddle.fluid.framework import _enable_legacy_dygraph, _global_flags
 from paddle.fluid.layer_helper import LayerHelper
@@ -48,7 +49,7 @@ def check():
     with fluid.dygraph.guard(fluid.core.CPUPlace()):
         a = fluid.dygraph.to_variable(a_np)
         b = fluid.dygraph.to_variable(b_np)
-        y = fluid.layers.elementwise_add(x=a, y=b)
+        y = paddle.add(x=a, y=b)
         y = fluid.layers.matmul(x=y, y=b, transpose_y=True)
         res1 = func(y)