• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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
16"""LeNet test."""
17
18import numpy as np
19
20from lenet import LeNet5
21import mindspore.nn as nn
22import mindspore.ops.composite as C
23from mindspore import Tensor
24from mindspore import context
25from mindspore.common.api import _cell_graph_executor
26
27context.set_context(mode=context.GRAPH_MODE)
28
29
30grad_all_with_sens = C.GradOperation(get_all=True, sens_param=True)
31
32batch_size = 1
33channel = 1
34height = 32
35weight = 32
36num_class = 10
37
38
39class LeNetGrad(nn.Cell):
40    """Backward of LeNet"""
41
42    def __init__(self, network):
43        super(LeNetGrad, self).__init__()
44        self.grad_op = grad_all_with_sens
45        self.network = network
46
47    def construct(self, x, sens):
48        grad_op = self.grad_op(self.network)(x, sens)
49
50        return grad_op
51
52
53def test_compile():
54    """Compile forward graph"""
55    net = LeNet(num_class=num_class)
56    np.random.seed(7)
57    inp = Tensor(np.array(np.random.randn(batch_size,
58                                          channel,
59                                          height,
60                                          weight) * 3, np.float32))
61
62    _cell_graph_executor.compile(net, inp)
63
64
65def test_compile_grad():
66    """Compile forward and backward graph"""
67    net = LeNet5(num_class=num_class)
68    inp = Tensor(np.array(np.random.randn(batch_size,
69                                          channel,
70                                          height,
71                                          weight) * 3, np.float32))
72    sens = Tensor(np.ones([batch_size, num_class]).astype(np.float32))
73    grad_op = LeNetGrad(net)
74
75    _cell_graph_executor.compile(grad_op, inp, sens)
76