[cherry-pick]to Release/2.3,modify scale op xpu unittest (#43657)

* modify xpu.cmake,*test=kunlun (#41832)

* modify xpu.cmake,*test=kunlun

* modify xpu.cmake,*test=kunlun

* modify xpu.cmake,*test=kunlun

* modify xpu.cmake,*test=kunlun

* support bilstm,*test=kunlun

* [cherry-pick]support multi_layer of bilstm,*test=kunlun

* [cherry-pick]refactor sum unit test,*test=kunlun (#43561)

python/paddle/fluid/tests/unittests/xpu/test_scale_op_xpu.py

@@ -43,9 +43,7 @@ class XPUTestScaleOp(XPUOpTestWrapper):
             self.place = paddle.XPUPlace(0)
             self.set_inputs()
             self.set_attrs()
-            self.outputs = {
-                'Out': self.inputs['X'] * self.dtype(self.attrs['scale'])
-            }
+            self.set_output()
 
         def set_xpu(self):
             self.__class__.use_xpu = True
@@ -55,6 +53,16 @@ class XPUTestScaleOp(XPUOpTestWrapper):
         def set_inputs(self):
             self.inputs = {'X': np.random.random((10, 10)).astype(self.dtype)}
 
+        def set_output(self):
+            if "float16" == self.in_type:
+                output = self.inputs['X'] * np.float16(self.attrs['scale'])
+            elif "int64" == self.in_type:
+                output = self.inputs['X'] * np.int64(self.attrs['scale'])
+            else:
+                output = self.inputs['X'] * np.float32(self.attrs['scale'])
+
+            self.outputs = {'Out': output}
+
         def init_dtype(self):
             if "float16" == self.in_type:
                 self.dtype = np.float16

python/paddle/fluid/tests/unittests/xpu/test_sum_op_xpu.py

@@ -19,56 +19,70 @@ import unittest
 import numpy as np
 from op_test_xpu import XPUOpTest
 import paddle
-from paddle import enable_static
 import paddle.fluid as fluid
 import paddle.fluid.core as core
 from paddle.fluid.op import Operator
 from paddle.fluid.tests.unittests.op_test import (
     OpTest, convert_float_to_uint16, convert_uint16_to_float)
 from paddle import _C_ops
+import op_test
+from op_test_xpu import XPUOpTest
+from xpu.get_test_cover_info import create_test_class, get_xpu_op_support_types, XPUOpTestWrapper
 
 paddle.enable_static()
 
 
-class TestSumOp(XPUOpTest):
+class XPUTestSumOp(XPUOpTestWrapper):
+    def __init__(self):
+        self.op_name = 'sum'
+        self.use_dynamic_create_class = False
+
+    class TestSumOp(XPUOpTest):
         def setUp(self):
+            self.init_dtype()
+            self.set_xpu()
             self.op_type = "sum"
-        self.init_kernel_type()
-        self.init_kernel_type()
-        x0 = np.random.random((3, 40)).astype(self.dtype)
-        x1 = np.random.random((3, 40)).astype(self.dtype)
-        x2 = np.random.random((3, 40)).astype(self.dtype)
+            self.place = paddle.XPUPlace(0)
+            self.set_shape()
+            x0 = np.random.random(self.shape).astype(self.dtype)
+            x1 = np.random.random(self.shape).astype(self.dtype)
+            x2 = np.random.random(self.shape).astype(self.dtype)
             self.inputs = {"X": [("x0", x0), ("x1", x1), ("x2", x2)]}
             y = x0 + x1 + x2
             self.outputs = {'Out': y}
 
-    def init_kernel_type(self):
-        self.dtype = np.float32
+        def init_dtype(self):
+            self.dtype = self.in_type
+
+        def set_xpu(self):
+            self.__class__.use_xpu = True
+            self.__class__.no_need_check_grad = True
+            self.__class__.op_type = self.dtype
+
+        def set_shape(self):
+            self.shape = (3, 10)
 
         def test_check_output(self):
-        self.check_output()
+            self.check_output_with_place(self.place)
 
         def test_check_grad(self):
-        self.check_grad(['x0'], 'Out')
+            self.check_grad_with_place(self.place, ['x0'], 'Out')
 
+    class TestSumOp1(TestSumOp):
+        def set_shape(self):
+            self.shape = (5)
 
-#----------- test fp16 -----------
-class TestFP16SumOp(TestSumOp):
-    def init_kernel_type(self):
-        self.dtype = np.float16
+    class TestSumOp2(TestSumOp):
+        def set_shape(self):
+            self.shape = (1, 1, 1, 1, 1)
 
-    def test_check_output(self):
-        place = core.XPUPlace(0)
-        # if core.is_float16_supported(place):
-        self.check_output_with_place(place, atol=2e-2)
+    class TestSumOp3(TestSumOp):
+        def set_shape(self):
+            self.shape = (10, 5, 7)
 
-    # FIXME: Because of the precision fp16, max_relative_error
-    # should be 0.15 here.
-    def test_check_grad(self):
-        place = core.XPUPlace(0)
-        # if core.is_float16_supported(place):
-        self.check_grad_with_place(
-            place, ['x0'], 'Out', max_relative_error=0.15)
+    class TestSumOp4(TestSumOp):
+        def set_shape(self):
+            self.shape = (2, 2, 3, 3)
 
 
 def create_test_sum_fp16_class(parent):
@@ -182,6 +196,9 @@ class TestSumOpError(unittest.TestCase):
         self.assertRaises(Exception, test_list_of_none_input)
 
 
+support_types = get_xpu_op_support_types('sum')
+for stype in support_types:
+    create_test_class(globals(), XPUTestSumOp, stype)
+
 if __name__ == "__main__":
-    enable_static()
     unittest.main()