• 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"""
16train Conv2dBnFoldQuant Cell
17"""
18
19import pytest
20import numpy as np
21from mindspore import nn
22from mindspore import context
23from mindspore import Tensor
24from mindspore.common import set_seed
25from mindspore.compression.quant import create_quant_config
26
27class Net(nn.Cell):
28    def __init__(self, qconfig):
29        super(Net, self).__init__()
30        self.conv = nn.Conv2dBnFoldQuant(2, 3, kernel_size=(2, 2), stride=(1, 1),
31                                         pad_mode='valid', quant_config=qconfig)
32    def construct(self, x):
33        return self.conv(x)
34
35def test_conv2d_bn_fold_quant():
36    set_seed(1)
37    quant_config = create_quant_config()
38    network = Net(quant_config)
39    inputs = Tensor(np.ones([1, 2, 5, 5]).astype(np.float32))
40    label = Tensor(np.ones([1, 3, 4, 4]).astype(np.int32))
41    opt = nn.Momentum(filter(lambda x: x.requires_grad, network.get_parameters()), learning_rate=0.1, momentum=0.9)
42    loss = nn.MSELoss()
43    net_with_loss = nn.WithLossCell(network, loss)
44    train_network = nn.TrainOneStepCell(net_with_loss, opt)
45    train_network.set_train()
46    out_loss = train_network(inputs, label)
47    expect_loss = np.array([0.940427])
48    error = np.array([0.1])
49    diff = out_loss.asnumpy() - expect_loss
50    assert np.all(abs(diff) < error)
51
52@pytest.mark.level0
53@pytest.mark.platform_arm_ascend_training
54@pytest.mark.platform_x86_ascend_training
55@pytest.mark.env_onecard
56def test_conv2d_bn_fold_quant_ascend():
57    context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
58    test_conv2d_bn_fold_quant()
59