• 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
16
17import mindspore.nn as nn
18from mindspore import Tensor
19from mindspore import context
20from mindspore.ops import operations as P
21from mindspore.train.model import Model
22
23
24class CrossEntropyLoss(nn.Cell):
25    def __init__(self, reduction='mean'):
26        super(CrossEntropyLoss, self).__init__()
27
28        self.reduce_mean = P.ReduceMean()
29        self.cross_entropy = nn.SoftmaxCrossEntropyWithLogits()
30        self.reduction = reduction
31
32    def construct(self, logits, label):
33        loss = self.cross_entropy(logits, label)
34        if self.reduction == 'mean':
35            loss = self.reduce_mean(loss, (-1,))
36        return loss
37
38
39class DatasetLenet():
40    def __init__(self, predict, label, length=3):
41        self.predict = predict
42        self.label = label
43        self.index = 0
44        self.length = length
45
46    def __iter__(self):
47        return self
48
49    def __next__(self):
50        if self.index >= self.length:
51            raise StopIteration
52        self.index += 1
53        return self.predict, self.label
54
55    def reset(self):
56        self.index = 0
57
58    def get_dataset_size(self):
59        return 32
60
61    def get_repeat_count(self):
62        return 1
63
64    def create_tuple_iterator(self, num_epochs=-1, do_copy=True):
65        return self
66
67
68class Net(nn.Cell):
69    def __init__(self):
70        super().__init__()
71        self.conv = nn.Conv2d(in_channels=3, out_channels=64, kernel_size=1, stride=1, pad_mode='valid',
72                              has_bias=True, weight_init='ones', bias_init='ones')
73        self.conv.conv2d.shard(((8, 1, 1, 1), (1, 1, 1, 1)))
74        self.reduce_mean = P.ReduceMean(keep_dims=False).shard(((1, 1, 1, 8),))
75        self.flat = nn.Flatten()
76
77    def construct(self, inputs):
78        x = self.conv(inputs)
79        x = self.reduce_mean(x, -1)
80        x = self.flat(x)
81        return x
82
83
84def test_bias_add():
85    context.set_context(mode=context.GRAPH_MODE)
86    context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=8)
87    input_np = np.ones([16, 3, 32, 32]).astype(np.float32)
88    label_np = np.zeros([16, 2048]).astype(np.float32)
89    dataset = DatasetLenet(Tensor(input_np), Tensor(label_np), 1)
90    net = Net()
91    loss = CrossEntropyLoss()
92    opt = nn.Momentum(learning_rate=0.01, momentum=0.9, params=net.get_parameters())
93    model = Model(network=net, loss_fn=loss, optimizer=opt)
94    model.train(epoch=1, train_dataset=dataset, dataset_sink_mode=False)
95