MindSpore ResNet50 模型构建与训练 - 代码详解

python/nimport mindspore.nn as nn/nimport mindspore.ops as ops/nimport mindspore.common.dtype as mstype/nimport os/nimport numpy as np/nfrom mindspore import Tensor/nfrom PIL import Image/nfrom mindspore.dataset import vision/nfrom mindspore import Model # 承载网络结构/nfrom mindspore.nn.metrics import Accuracy # 测试模型用/nfrom mindspore.train.callback import LossMonitor/n/n/nclass ConvBlock(nn.Cell):/n def __init__(self, in_channels, out_channels, kernel_size=3, stride=1, padding=1):/n super(ConvBlock, self).__init__()/n self.conv = nn.Conv2d(in_channels, out_channels, kernel_size, stride, padding, has_bias=False)/n self.bn = nn.BatchNorm2d(out_channels)/n self.relu = nn.ReLU()/n/n def construct(self, x):/n x = self.conv(x)/n x = self.bn(x)/n x = self.relu(x)/n return x/n/nclass ResBlock(nn.Cell):/n def __init__(self, in_channels, out_channels, stride=1):/n super(ResBlock, self).__init__()/n self.conv1 = ConvBlock(in_channels, out_channels, stride=stride)/n self.conv2 = ConvBlock(out_channels, out_channels, kernel_size=3, stride=1, padding=1)/n self.downsample = nn.SequentialCell([nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=stride, has_bias=False), nn.BatchNorm2d(out_channels)])/n self.relu = nn.ReLU()/n/n def construct(self, x):/n identity = x/n x = self.conv1(x)/n x = self.conv2(x)/n if self.downsample is not None:/n identity = self.downsample(identity)/n x = x + identity/n x = self.relu(x)/n return x/n/nclass ResNet(nn.Cell):/n def __init__(self, block, layers, num_classes=1000):/n super(ResNet, self).__init__()/n self.in_channels = 64/n self.conv1 = ConvBlock(3, 64, kernel_size=7, stride=2, padding=3, pad_mode='same') # 更改为'same'/n self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, pad_mode='same')/n self.layer1 = self._make_layer(block, 64, layers[0])/n self.layer2 = self._make_layer(block, 128, layers[1], stride=2)/n self.layer3 = self._make_layer(block, 256, layers[2], stride=2)/n self.layer4 = self._make_layer(block, 512, layers[3], stride=2)/n self.avgpool = nn.AvgPool2d(7, 1)/n self.dropout = nn.Dropout(0.4)/n self.fc = nn.Dense(512 * block.expansion, num_classes)/n/n def _make_layer(self, block, out_channels, blocks, stride=1):/n downsample = None/n if stride != 1 or self.in_channels != out_channels * block.expansion:/n downsample = nn.SequentialCell([nn.Conv2d(self.in_channels, out_channels * block.expansion, kernel_size=1, stride=stride, has_bias=False), nn.BatchNorm2d(out_channels * block.expansion)])/n layers = []/n layers.append(block(self.in_channels, out_channels, stride, downsample))/n self.in_channels = out_channels * block.expansion/n for i in range(1, blocks):/n layers.append(block(self.in_channels, out_channels))/n return nn.SequentialCell(layers)/n/n def construct(self, x):/n x = self.conv1(x)/n x = self.maxpool(x)/n x = self.layer1(x)/n x = self.layer2(x)/n x = self.layer3(x)/n x = self.layer4(x)/n x = self.avgpool(x)/n x = self.dropout(x)/n x = ops.Reshape()(x, (-1, 512 * 1 * 1))/n x = self.fc(x)/n return x/n/nlenet = ResNet()/n/ndef resnet50():/n return ResNet(ResBlock, [3, 4, 6, 3])/n/n# 加载预训练集/n'''''''''/ndef load_dataset(data_path):/n images = []/n labels = []/n for subdir in os.listdir(data_path):/n subpath = os.path.join(data_path, subdir)/n for filename in os.listdir(subpath):/n imgpath = os.path.join(subpath, filename)/n img = Image.open(imgpath)/n img = img.resize((224, 224))/n img = np.array(img).astype(np.float32)/n img = img.transpose((2, 0, 1))/n images.append(img)/n labels.append(int(subdir))/n images = np.array(images)/n labels = np.array(labels)/n return Tensor(images), Tensor(labels)/n'''''''''''/nnet_loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True)/nlr = 0.01/nmomentum = 0.9/nnet_opt = nn.Momentum(lenet.trainable_params(), lr, momentum)/nmodel = Model(lenet, net_loss, net_opt, metrics={'accuracy': Accuracy()})/n/nloss_cb = LossMonitor(per_print_times=train_data.get_dataset_size())/n# 训练模型/nmodel.train(3, train_data, loss_cb) # 训练3个epoch/nmodel.eval(test_data)/n/n# 中文翻译/n/n在ResNet类的构造函数中的第一个卷积层中，您正在将整数值3传递给'pad_mode'参数，但该参数应该是字符串类型。这是因为MindSpore的'Conv2d'操作的'pad_mode'参数需要是字符串类型。/n/n解决此问题，您可以将'pad_mode'参数的值更改为正确的字符串类型。例如，'pad_mode = 'same'' 表示使用填充模式'same'。/n/n', 'content': 'python/nimport mindspore.nn as nn/nimport mindspore.ops as ops/nimport mindspore.common.dtype as mstype/nimport os/nimport numpy as np/nfrom mindspore import Tensor/nfrom PIL import Image/nfrom mindspore.dataset import vision/nfrom mindspore import Model # 承载网络结构/nfrom mindspore.nn.metrics import Accuracy # 测试模型用/nfrom mindspore.train.callback import LossMonitor/n/n/nclass ConvBlock(nn.Cell):/n def __init__(self, in_channels, out_channels, kernel_size=3, stride=1, padding=1):/n super(ConvBlock, self).__init__()/n self.conv = nn.Conv2d(in_channels, out_channels, kernel_size, stride, padding, has_bias=False)/n self.bn = nn.BatchNorm2d(out_channels)/n self.relu = nn.ReLU()/n/n def construct(self, x):/n x = self.conv(x)/n x = self.bn(x)/n x = self.relu(x)/n return x/n/nclass ResBlock(nn.Cell):/n def __init__(self, in_channels, out_channels, stride=1):/n super(ResBlock, self).__init__()/n self.conv1 = ConvBlock(in_channels, out_channels, stride=stride)/n self.conv2 = ConvBlock(out_channels, out_channels, kernel_size=3, stride=1, padding=1)/n self.downsample = nn.SequentialCell([nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=stride, has_bias=False), nn.BatchNorm2d(out_channels)])/n self.relu = nn.ReLU()/n/n def construct(self, x):/n identity = x/n x = self.conv1(x)/n x = self.conv2(x)/n if self.downsample is not None:/n identity = self.downsample(identity)/n x = x + identity/n x = self.relu(x)/n return x/n/nclass ResNet(nn.Cell):/n def __init__(self, block, layers, num_classes=1000):/n super(ResNet, self).__init__()/n self.in_channels = 64/n self.conv1 = ConvBlock(3, 64, kernel_size=7, stride=2, padding=3, pad_mode='same') # 更改为'same'/n self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, pad_mode='same')/n self.layer1 = self._make_layer(block, 64, layers[0])/n self.layer2 = self._make_layer(block, 128, layers[1], stride=2)/n self.layer3 = self._make_layer(block, 256, layers[2], stride=2)/n self.layer4 = self._make_layer(block, 512, layers[3], stride=2)/n self.avgpool = nn.AvgPool2d(7, 1)/n self.dropout = nn.Dropout(0.4)/n self.fc = nn.Dense(512 * block.expansion, num_classes)/n/n def _make_layer(self, block, out_channels, blocks, stride=1):/n downsample = None/n if stride != 1 or self.in_channels != out_channels * block.expansion:/n downsample = nn.SequentialCell([nn.Conv2d(self.in_channels, out_channels * block.expansion, kernel_size=1, stride=stride, has_bias=False), nn.BatchNorm2d(out_channels * block.expansion)])/n layers = []/n layers.append(block(self.in_channels, out_channels, stride, downsample))/n self.in_channels = out_channels * block.expansion/n for i in range(1, blocks):/n layers.append(block(self.in_channels, out_channels))/n return nn.SequentialCell(layers)/n/n def construct(self, x):/n x = self.conv1(x)/n x = self.maxpool(x)/n x = self.layer1(x)/n x = self.layer2(x)/n x = self.layer3(x)/n x = self.layer4(x)/n x = self.avgpool(x)/n x = self.dropout(x)/n x = ops.Reshape()(x, (-1, 512 * 1 * 1))/n x = self.fc(x)/n return x/n/nlenet = ResNet()/n/ndef resnet50():/n return ResNet(ResBlock, [3, 4, 6, 3])/n/n# 加载预训练集/n'''''''''/ndef load_dataset(data_path):/n images = []/n labels = []/n for subdir in os.listdir(data_path):/n subpath = os.path.join(data_path, subdir)/n for filename in os.listdir(subpath):/n imgpath = os.path.join(subpath, filename)/n img = Image.open(imgpath)/n img = img.resize((224, 224))/n img = np.array(img).astype(np.float32)/n img = img.transpose((2, 0, 1))/n images.append(img)/n labels.append(int(subdir))/n images = np.array(images)/n labels = np.array(labels)/n return Tensor(images), Tensor(labels)/n'''''''''''/nnet_loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True)/nlr = 0.01/nmomentum = 0.9/nnet_opt = nn.Momentum(lenet.trainable_params(), lr, momentum)/nmodel = Model(lenet, net_loss, net_opt, metrics={'accuracy': Accuracy()})/n/nloss_cb = LossMonitor(per_print_times=train_data.get_dataset_size())/n# 训练模型/nmodel.train(3, train_data, loss_cb) # 训练3个epoch/nmodel.eval(test_data)/n/n# 中文翻译/n/n在ResNet类的构造函数中的第一个卷积层中，您正在将整数值3传递给'pad_mode'参数，但该参数应该是字符串类型。这是因为MindSpore的'Conv2d'操作的'pad_mode'参数需要是字符串类型。/n/n解决此问题，您可以将'pad_mode'参数的值更改为正确的字符串类型。例如，'pad_mode = 'same'' 表示使用填充模式'same'。/n/