1# Copyright 2020-2021 Huawei Technologies Co., Ltd 2# 3# Licensed under the Apache License, Version 2.0 (the "License"); 4# you may not use this file except in compliance with the License. 5# You may obtain a copy of the License at 6# 7# http://www.apache.org/licenses/LICENSE-2.0 8# 9# Unless required by applicable law or agreed to in writing, software 10# distributed under the License is distributed on an "AS IS" BASIS, 11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12# See the License for the specific language governing permissions and 13# limitations under the License. 14# ============================================================================ 15 16import pytest 17import numpy as np 18from mindspore import Tensor 19from mindspore.ops import operations as P 20import mindspore.nn as nn 21import mindspore.context as context 22from mindspore.common.api import ms_function 23 24context.set_context(mode=context.PYNATIVE_MODE, device_target="CPU") 25 26 27class NetReduce(nn.Cell): 28 def __init__(self): 29 super(NetReduce, self).__init__() 30 self.axis0 = 0 31 self.axis1 = 1 32 self.axis2 = -1 33 self.axis3 = (0, 1) 34 self.axis4 = (0, 1, 2) 35 self.axis5 = (-1,) 36 self.axis6 = () 37 self.reduce_mean = P.ReduceMean(False) 38 self.reduce_sum = P.ReduceSum(False) 39 self.reduce_max = P.ReduceMax(False) 40 self.reduce_min = P.ReduceMin(False) 41 42 @ms_function 43 def construct(self, indice): 44 return (self.reduce_mean(indice, self.axis0), 45 self.reduce_mean(indice, self.axis1), 46 self.reduce_mean(indice, self.axis2), 47 self.reduce_mean(indice, self.axis3), 48 self.reduce_mean(indice, self.axis4), 49 self.reduce_sum(indice, self.axis0), 50 self.reduce_sum(indice, self.axis2), 51 self.reduce_max(indice, self.axis0), 52 self.reduce_max(indice, self.axis2), 53 self.reduce_max(indice, self.axis5), 54 self.reduce_max(indice, self.axis6), 55 self.reduce_min(indice, self.axis0), 56 self.reduce_min(indice, self.axis1), 57 self.reduce_min(indice, self.axis2), 58 self.reduce_min(indice, self.axis3), 59 self.reduce_min(indice, self.axis4), 60 self.reduce_min(indice, self.axis5), 61 self.reduce_min(indice, self.axis6)) 62 63 64class NetReduceLogic(nn.Cell): 65 def __init__(self): 66 super(NetReduceLogic, self).__init__() 67 self.axis0 = 0 68 self.axis1 = -1 69 self.axis2 = (0, 1, 2) 70 self.axis3 = () 71 self.reduce_all = P.ReduceAll(False) 72 self.reduce_any = P.ReduceAny(False) 73 74 @ms_function 75 def construct(self, indice): 76 return (self.reduce_all(indice, self.axis0), 77 self.reduce_all(indice, self.axis1), 78 self.reduce_all(indice, self.axis2), 79 self.reduce_all(indice, self.axis3), 80 self.reduce_any(indice, self.axis0), 81 self.reduce_any(indice, self.axis1), 82 self.reduce_any(indice, self.axis2), 83 self.reduce_any(indice, self.axis3),) 84 85 86@pytest.mark.level0 87@pytest.mark.platform_x86_cpu 88@pytest.mark.env_onecard 89def test_reduce(): 90 reduce = NetReduce() 91 indice = Tensor(np.array([ 92 [[0., 2., 1., 4., 0., 2.], [3., 1., 2., 2., 4., 0.]], 93 [[2., 0., 1., 5., 0., 1.], [1., 0., 0., 4., 4., 3.]], 94 [[4., 1., 4., 0., 0., 0.], [2., 5., 1., 0., 1., 3.]] 95 ]).astype(np.float32)) 96 output = reduce(indice) 97 print(output[0]) 98 print(output[1]) 99 print(output[2]) 100 print(output[3]) 101 print(output[4]) 102 print(output[5]) 103 print(output[6]) 104 print(output[7]) 105 print(output[8]) 106 print(output[9]) 107 print(output[10]) 108 print(output[11]) 109 print(output[12]) 110 print(output[13]) 111 print(output[14]) 112 print(output[15]) 113 print(output[16]) 114 print(output[17]) 115 expect_0 = np.array([[2., 1., 2., 3., 0., 1], [2., 2., 1., 2., 3., 2.]]).astype(np.float32) 116 expect_1 = np.array([[1.5, 1.5, 1.5, 3., 2., 1.], [1.5, 0., 0.5, 4.5, 2., 2.], [3., 3., 2.5, 0., 0.5, 1.5]]).astype( 117 np.float32) 118 expect_2 = np.array([[1.5, 2.], [1.5, 2.], [1.5, 2.]]).astype(np.float32) 119 expect_3 = np.array([2, 1.5, 1.5, 2.5, 1.5, 1.5]).astype(np.float32) 120 expect_4 = np.array([1.75]).astype(np.float32) 121 expect_5 = np.array([[6., 3., 6., 9., 0., 3.], [6., 6., 3., 6., 9., 6.]]).astype(np.float32) 122 expect_6 = np.array([[9., 12.], [9., 12.], [9., 12.]]).astype(np.float32) 123 expect_7 = np.array([[4., 2., 4., 5., 0., 2.], [3., 5., 2., 4., 4., 3.]]).astype(np.float32) 124 expect_8 = np.array([[4., 4.], [5., 4.], [4., 5.]]).astype(np.float32) 125 expect_9 = np.array([[0., 0., 1., 0., 0., 0.], [1., 0., 0., 0., 1., 0.]]).astype(np.float32) 126 expect_10 = np.array([[0., 1., 1., 2., 0., 0.], [1., 0., 0., 4., 0., 1.], [2., 1., 1., 0., 0., 0.]]).astype( 127 np.float32) 128 expect_11 = np.array([[0., 0.], [0., 0.], [0., 0.]]).astype(np.float32) 129 expect_12 = np.array([0., 0., 0., 0., 0., 0.]).astype(np.float32) 130 assert (output[0].asnumpy() == expect_0).all() 131 assert (output[1].asnumpy() == expect_1).all() 132 assert (output[2].asnumpy() == expect_2).all() 133 assert (output[3].asnumpy() == expect_3).all() 134 assert (output[4].asnumpy() == expect_4).all() 135 assert (output[5].asnumpy() == expect_5).all() 136 assert (output[6].asnumpy() == expect_6).all() 137 assert (output[7].asnumpy() == expect_7).all() 138 assert (output[8].asnumpy() == expect_8).all() 139 assert (output[9].asnumpy() == expect_8).all() 140 assert (output[10].asnumpy() == 5.0).all() 141 assert (output[11].asnumpy() == expect_9).all() 142 assert (output[12].asnumpy() == expect_10).all() 143 assert (output[13].asnumpy() == expect_11).all() 144 assert (output[14].asnumpy() == expect_12).all() 145 assert (output[15].asnumpy() == 0.0).all() 146 assert (output[16].asnumpy() == expect_11).all() 147 assert (output[17].asnumpy() == 0.0).all() 148 149 150@pytest.mark.level0 151@pytest.mark.platform_x86_cpu 152@pytest.mark.env_onecard 153def test_reduce_logic(): 154 reduce_logic = NetReduceLogic() 155 indice_bool = Tensor([[[False, True, True, True, False, True], 156 [True, True, True, True, True, False]], 157 [[True, False, True, True, False, True], 158 [True, False, False, True, True, True]], 159 [[True, True, True, False, False, False], 160 [True, True, True, False, True, True]]]) 161 output = reduce_logic(indice_bool) 162 expect_all_1 = np.array([[False, False, True, False, False, False], 163 [True, False, False, False, True, False]]) 164 expect_all_2 = np.array([[False, False], [False, False], [False, False]]) 165 expect_all_3 = False 166 expect_all_4 = False 167 expect_any_1 = np.array([[True, True, True, True, False, True], [True, True, True, True, True, True]]) 168 expect_any_2 = np.array([[True, True], [True, True], [True, True]]) 169 expect_any_3 = True 170 expect_any_4 = True 171 172 assert (output[0].asnumpy() == expect_all_1).all() 173 assert (output[1].asnumpy() == expect_all_2).all() 174 assert (output[2].asnumpy() == expect_all_3).all() 175 assert (output[3].asnumpy() == expect_all_4).all() 176 assert (output[4].asnumpy() == expect_any_1).all() 177 assert (output[5].asnumpy() == expect_any_2).all() 178 assert (output[6].asnumpy() == expect_any_3).all() 179 assert (output[7].asnumpy() == expect_any_4).all() 180 181 182test_reduce() 183test_reduce_logic() 184