1# Copyright 2020 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 numpy as np 17import pytest 18 19import mindspore.context as context 20import mindspore.nn as nn 21from mindspore import Tensor 22from mindspore.ops import composite as C 23from mindspore.ops import operations as P 24 25 26@pytest.mark.level0 27@pytest.mark.platform_x86_gpu_training 28@pytest.mark.env_onecard 29def test_logsoftmax(): 30 x = np.array([[-0.08082921, -0.13706027, -0.4711177, -0.05606057], 31 [-0.46082982, 1.1761844, -1.016654, -1.743829], 32 [-1.5062045, 0.6910976, 0.4839723, 1.1502692]]).astype(np.float32) 33 expect = np.array([[-1.2939762, -1.3502073, -1.6842647, -1.2692076], 34 [-1.9445671, -0.3075528, -2.5003912, -3.2275662], 35 [-3.452001, -1.2546989, -1.4618242, -0.79552734]]).astype(np.float32) 36 37 context.set_context(mode=context.GRAPH_MODE, device_target="GPU") 38 logSoftmax = P.LogSoftmax() 39 output = logSoftmax(Tensor(x)) 40 assert np.allclose(output.asnumpy(), expect) 41 42 43class LogSoftmax(nn.Cell): 44 def __init__(self, axis=-1): 45 super(LogSoftmax, self).__init__() 46 self.logsoftmax = P.LogSoftmax(axis) 47 48 def construct(self, x): 49 return self.logsoftmax(x) 50 51 52class Grad(nn.Cell): 53 def __init__(self, network): 54 super(Grad, self).__init__() 55 self.grad = C.GradOperation(get_all=True, sens_param=True) 56 self.network = network 57 58 def construct(self, input_data, sens): 59 gout = self.grad(self.network)(input_data, sens) 60 return gout 61 62 63@pytest.mark.level0 64@pytest.mark.platform_x86_gpu_training 65@pytest.mark.env_onecard 66def test_logsoftmaxgrad(): 67 x = np.array([[-0.47705367, 0.48267725, -1.0453935, 1.574488, 0.20362134, 0.4435456, -0.23984082, -0.43684655, 68 -0.7725506, 1.4481013], 69 [1.1012247, 1.7069651, 0.55062026, 0.3361901, -1.1082426, -0.5001939, -0.3255393, -0.7972024, 70 -0.27965206, -0.702805], 71 [0.19450496, 0.87596166, 0.6467245, -1.044987, 0.5248943, -2.6166635, 1.6719198, 0.06600758, 72 -0.4099178, 1.1861311], 73 [1.1305193, -1.97308, 2.1047623, -1.5105937, 0.93052036, 1.2467804, 0.5310002, 0.7084912, -1.3681422, 74 -0.9686862], 75 [1.871408, 0.14219497, -0.41050452, -0.749807, 1.4900619, -1.8172716, -0.73839617, 0.17565694, 76 -0.4553867, -1.5423119]]).astype(np.float32) 77 dy = np.array([[1.516363, -0.15196544, 0.598733, 0.64357865, 0.16265012, -1.3521105, 0.22621834, 0.7168259, 78 -0.6709239, 0.79757756], 79 [-0.32457778, 1.2831115, 1.1211495, -0.02665559, 1.9170904, -1.3397789, 1.4124829, -1.4298155, 80 0.758519, -0.25322974], 81 [-0.24226122, -1.2555921, 0.6492511, -0.34847677, 0.19916506, 0.628554, -0.19658111, 0.44939864, 82 -0.11677749, -1.2131723], 83 [0.24267715, 0.28106326, 1.1075432, -0.29006946, 0.31335673, 0.8833154, 0.13152207, 1.5482179, 84 0.29770762, -0.16246222], 85 [0.02145994, 0.80424, -0.95061, 1.5875458, -0.00308682, 0.17964548, 0.49912593, 0.46977136, 86 0.2151897, 0.30908248]]).astype(np.float32) 87 expect = np.array([[1.4219905, -0.39837134, 0.5452743, -0.09062839, -0.02375537, -1.5890603, 0.10658137, 0.6185817, 88 -0.7411523, 0.15054005], 89 [-0.94926417, 0.13830578, 0.7609547, -0.31733334, 1.8485254, -1.4657221, 1.2625053, -1.523396, 90 0.601499, -0.35607445], 91 [-0.14447737, -1.0622973, 0.80294746, -0.32016528, 0.33523226, 0.63443416, 0.23186903, 92 0.53539133, -0.0633494, -0.9495847], 93 [-0.36894822, 0.253609, -0.5127511, -0.33366728, -0.18740037, 0.19628316, -0.20430653, 1.1471655, 94 0.24743511, -0.23741922], 95 [-1.2582518, 0.57718843, -1.0812542, 1.4944922, -0.8770549, 0.1476463, 0.40500447, 0.23499368, 96 0.09027944, 0.26695627]]).astype(np.float32) 97 98 context.set_context(mode=context.GRAPH_MODE, device_target="GPU") 99 net = LogSoftmax() 100 dx = Grad(net)(Tensor(x), Tensor(dy)) 101 assert np.allclose(dx[0].asnumpy(), expect) 102 103 104@pytest.mark.level0 105@pytest.mark.platform_x86_gpu_training 106@pytest.mark.env_onecard 107def test_logsoftmaxgrad1(): 108 x = np.array([[-0.47705367, 0.48267725, -1.0453935, 1.574488, 0.20362134, 0.4435456, -0.23984082, -0.43684655, 109 -0.7725506, 1.4481013], 110 [1.1012247, 1.7069651, 0.55062026, 0.3361901, -1.1082426, -0.5001939, -0.3255393, -0.7972024, 111 -0.27965206, -0.702805], 112 [0.19450496, 0.87596166, 0.6467245, -1.044987, 0.5248943, -2.6166635, 1.6719198, 0.06600758, 113 -0.4099178, 1.1861311], 114 [1.1305193, -1.97308, 2.1047623, -1.5105937, 0.93052036, 1.2467804, 0.5310002, 0.7084912, -1.3681422, 115 -0.9686862], 116 [1.871408, 0.14219497, -0.41050452, -0.749807, 1.4900619, -1.8172716, -0.73839617, 0.17565694, 117 -0.4553867, -1.5423119]]).astype(np.float32) 118 dy = np.array([[1.516363, -0.15196544, 0.598733, 0.64357865, 0.16265012, -1.3521105, 0.22621834, 0.7168259, 119 -0.6709239, 0.79757756], 120 [-0.32457778, 1.2831115, 1.1211495, -0.02665559, 1.9170904, -1.3397789, 1.4124829, -1.4298155, 121 0.758519, -0.25322974], 122 [-0.24226122, -1.2555921, 0.6492511, -0.34847677, 0.19916506, 0.628554, -0.19658111, 0.44939864, 123 -0.11677749, -1.2131723], 124 [0.24267715, 0.28106326, 1.1075432, -0.29006946, 0.31335673, 0.8833154, 0.13152207, 1.5482179, 125 0.29770762, -0.16246222], 126 [0.02145994, 0.80424, -0.95061, 1.5875458, -0.00308682, 0.17964548, 0.49912593, 0.46977136, 127 0.2151897, 0.30908248]]).astype(np.float32) 128 expect = np.array([[1.464194, -0.29578894, 0.5296974, -0.39600563, -0.1479242, -1.0869746, 0.04521982, 0.5064515, 129 -0.7515615, 1.0554069], 130 [-0.5774203, 0.793861, 0.7805745, -0.32800734, 1.8334473, -1.236596, 1.2463496, -1.5765365, 131 0.6265108, -0.22322391], 132 [-0.34437084, -1.4687154, 0.27432096, -0.42420125, -0.22908019, 0.640983, -1.4210342, 0.10155854, 133 -0.23266247, -1.0147638], 134 [-0.01768187, 0.26872346, -0.5037259, -0.3376058, -0.3291146, 1.4752979, -0.25972134, 0.8869053, 135 0.25325722, -0.13946185], 136 [-0.5247209, 0.70192003, -1.0808672, 1.4858199, -1.1273282, 0.20728993, 0.38918605, 0.08162117, 137 0.10445589, 0.3220427]],).astype(np.float32) 138 139 context.set_context(mode=context.GRAPH_MODE, device_target="GPU") 140 net = LogSoftmax(0) 141 dx = Grad(net)(Tensor(x), Tensor(dy)) 142 assert np.allclose(dx[0].asnumpy(), expect) 143