• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright 2019 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# ============================================================================
15import numpy as np
16import pytest
17
18import mindspore.nn as nn
19from mindspore import Tensor, Parameter
20from mindspore.common import dtype as mstype
21from mindspore.common.api import _cell_graph_executor
22from mindspore.nn import TrainOneStepCell, WithLossCell, ParameterUpdate
23from mindspore.nn.optim import Momentum
24from mindspore.ops import operations as P
25
26
27class Net(nn.Cell):
28    def __init__(self):
29        super(Net, self).__init__()
30        self.weight = Parameter(Tensor(np.ones([64, 10]).astype(np.float32)), name="weight")
31        self.bias = Parameter(Tensor(np.ones([10]).astype((np.float32))), name="bias")
32        self.matmul = P.MatMul()
33        self.biasAdd = P.BiasAdd()
34
35    def construct(self, x):
36        x = self.biasAdd(self.matmul(x, self.weight), self.bias)
37        return x
38
39
40def test_parameter_update_int32_and_tensor():
41    """ test_parameter_update """
42    net = Net()
43    loss = nn.SoftmaxCrossEntropyWithLogits()
44    optimizer = Momentum(net.get_parameters(), Tensor(np.array([0.1, 0.01, 0.001]), mstype.float32), 0.001)
45
46    net_with_loss = WithLossCell(net, loss)
47    train_network = TrainOneStepCell(net_with_loss, optimizer)
48
49    # compile train graph
50    train_network.set_train()
51    inputs = Tensor(np.ones([1, 64]).astype(np.float32))
52    label = Tensor(np.zeros([1, 10]).astype(np.float32))
53    _cell_graph_executor.compile(train_network, inputs, label)
54
55    # test tensor
56    param_lr = train_network.parameters_dict()['learning_rate']
57    update_network = ParameterUpdate(param_lr)
58    update_network.phase = 'update_param'
59
60    input_lr = Tensor(np.array([0.2, 0.02, 0.002]), mstype.float32)
61    _cell_graph_executor.compile(update_network, input_lr)
62
63    # test int32
64    param_step = train_network.parameters_dict()['global_step']
65    update_global_step = ParameterUpdate(param_step)
66
67    input_step = Tensor(np.array([1000]), mstype.int32)
68    _cell_graph_executor.compile(update_global_step, input_step)
69
70
71def test_parameter_update_float32():
72    """ test_parameter_update """
73    net = Net()
74    loss = nn.SoftmaxCrossEntropyWithLogits()
75    optimizer = Momentum(net.get_parameters(), 0.01, 0.001)
76
77    net_with_loss = WithLossCell(net, loss)
78    train_network = TrainOneStepCell(net_with_loss, optimizer)
79
80    # compile train graph
81    train_network.set_train()
82    inputs = Tensor(np.ones([1, 64]).astype(np.float32))
83    label = Tensor(np.zeros([1, 10]).astype(np.float32))
84    _cell_graph_executor.compile(train_network, inputs, label)
85
86    # construct and compile update graph
87    param_lr = train_network.parameters_dict()['learning_rate']
88    update_network = ParameterUpdate(param_lr)
89    update_network.phase = 'update_param'
90
91    input_lr = Tensor(0.0001, mstype.float32)
92    _cell_graph_executor.compile(update_network, input_lr)
93
94
95def test_parameter_update_error():
96    """ test_parameter_update """
97    input_np = np.array([1])
98
99    with pytest.raises(TypeError):
100        ParameterUpdate(input_np)
101