• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright 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 numpy as np
17import pytest
18
19import mindspore.context as context
20import mindspore.nn as nn
21from mindspore import Tensor
22from mindspore.ops import operations as P
23from mindspore.ops.operations import _grad_ops as G
24
25
26class NetSigmoid(nn.Cell):
27    def __init__(self):
28        super(NetSigmoid, self).__init__()
29        self.sigmoid = P.Sigmoid()
30
31    def construct(self, x):
32        return self.sigmoid(x)
33
34
35class NetSigmoidGrad(nn.Cell):
36    def __init__(self):
37        super(NetSigmoidGrad, self).__init__()
38        self.sigmoid_grad = G.SigmoidGrad()
39
40    def construct(self, y, dy):
41        return self.sigmoid_grad(y, dy)
42
43
44@pytest.mark.level0
45@pytest.mark.platform_x86_gpu_training
46@pytest.mark.env_onecard
47def test_sigmoid():
48    x = Tensor(np.array([[[[-1, 1, 10],
49                           [1, -1, 1],
50                           [10, 1, -1]]]]).astype(np.float32))
51
52    error = np.ones(shape=[1, 1, 3, 3]) * 1.0e-6
53
54    context.set_context(mode=context.GRAPH_MODE,
55                        enable_graph_kernel=True, device_target="GPU")
56    net = NetSigmoid()
57    result_open_gk = net(x)
58
59    context.set_context(mode=context.GRAPH_MODE,
60                        enable_graph_kernel=False, device_target="GPU")
61    net_beta = NetSigmoid()
62    result_close_gk = net_beta(x)
63    diff = result_open_gk.asnumpy() - result_close_gk.asnumpy()
64    assert np.all(abs(diff) < error)
65
66
67@pytest.mark.level0
68@pytest.mark.platform_x86_gpu_training
69@pytest.mark.env_onecard
70def test_sigmoid_grad():
71    y = Tensor(np.array([[[[-1, 1, 2],
72                           [1, -1, 1],
73                           [2, 1, -1]]]]).astype(np.float32))
74    dy = Tensor(np.array([[[[-11, 2, 4],
75                            [-1, 1, -1],
76                            [-4, 4, -4]]]]).astype(np.float32))
77
78    error = np.ones(shape=[1, 1, 3, 3]) * 1.0e-6
79
80    context.set_context(mode=context.GRAPH_MODE,
81                        enable_graph_kernel=True, device_target="GPU")
82    net = NetSigmoidGrad()
83    result_open_gk = net(y, dy)
84
85    context.set_context(mode=context.GRAPH_MODE,
86                        enable_graph_kernel=False, device_target="GPU")
87    net_beta = NetSigmoidGrad()
88    result_close_gk = net_beta(y, dy)
89    diff = result_open_gk.asnumpy() - result_close_gk.asnumpy()
90    assert np.all(abs(diff) < error)
91