未验证 提交 0993c3b2 编写于 作者: z8hanghuan's avatar z8hanghuan 提交者: GitHub

refactor sum unit test,*test=kunlun (#43561)

上级 4d649893
......@@ -46,9 +46,7 @@ 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
......@@ -58,6 +56,16 @@ 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
......
......@@ -20,7 +20,6 @@ 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
......@@ -28,52 +27,70 @@ 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.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_dtype(self):
self.dtype = self.in_type
def setUp(self):
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.inputs = {"X": [("x0", x0), ("x1", x1), ("x2", x2)]}
y = x0 + x1 + x2
self.outputs = {'Out': y}
def set_xpu(self):
self.__class__.use_xpu = True
self.__class__.no_need_check_grad = True
self.__class__.op_type = self.dtype
def init_kernel_type(self):
self.dtype = np.float32
def set_shape(self):
self.shape = (3, 10)
def test_check_output(self):
self.check_output()
def test_check_output(self):
self.check_output_with_place(self.place)
def test_check_grad(self):
self.check_grad(['x0'], 'Out')
def test_check_grad(self):
self.check_grad_with_place(self.place, ['x0'], 'Out')
class TestSumOp1(TestSumOp):
#----------- test fp16 -----------
class TestFP16SumOp(TestSumOp):
def set_shape(self):
self.shape = (5)
def init_kernel_type(self):
self.dtype = np.float16
class TestSumOp2(TestSumOp):
def test_check_output(self):
place = core.XPUPlace(0)
# if core.is_float16_supported(place):
self.check_output_with_place(place, atol=2e-2)
def set_shape(self):
self.shape = (1, 1, 1, 1, 1)
# 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 TestSumOp3(TestSumOp):
def set_shape(self):
self.shape = (10, 5, 7)
class TestSumOp4(TestSumOp):
def set_shape(self):
self.shape = (2, 2, 3, 3)
def create_test_sum_fp16_class(parent):
......@@ -198,6 +215,9 @@ 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()
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册