将 Keras NN 转换为 Pytorch NN

Question

我是在 pytorch 和 Keras 中构建神经网络的初学者。 对于 AlexNet model 的变体，我有以下 Keras 代码：

def shallownet(nb_classes):
    global img_size
    model = Sequential()
    model.add(Conv2D(64, (5, 5), input_shape=img_size, data_format='channels_first'))
    model.add(BatchNormalization(axis=1))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(3,3), strides=(2,2), padding='same', data_format='channels_first'))

    model.add(Conv2D(64, (5, 5), padding='same', data_format='channels_first'))
    model.add(BatchNormalization(axis=1))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(3,3), strides=(2,2), padding='same', data_format='channels_first'))

    model.add(Flatten())
    model.add(Dense(384))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(Dense(192))
    model.add(BatchNormalization())
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(Dense(nb_classes, activation='softmax'))
    return model

This is based on the C1/C3 model described on page 12 of this paper: https://arxiv.org/pdf/1609.04836.pdf And I want to convert this to the Pytorch version of the NN. 具体来说，我试图以以下形式获得它：

class AlexNetOWT_BN(nn.Module):

    def __init__(self, num_classes=1000):
        super(AlexNetOWT_BN, self).__init__()
        self.features = nn.Sequential(
            nn.Conv2d(3, 64, kernel_size=11, stride=4, padding=2,
                      bias=False),
            nn.BatchNorm2d(64),
            nn.ReLU(inplace=True),
            nn.MaxPool2d(kernel_size=3, stride=2),

            nn.Conv2d(64, 192, kernel_size=5, padding=2, bias=False),
            nn.BatchNorm2d(192),
            nn.ReLU(inplace=True),
            nn.MaxPool2d(kernel_size=3, stride=2),

            nn.Conv2d(192, 384, kernel_size=3, padding=1, bias=False),
            nn.ReLU(inplace=True),
            nn.BatchNorm2d(384),
            nn.Conv2d(384, 256, kernel_size=3, padding=1, bias=False),
            nn.ReLU(inplace=True),
            nn.BatchNorm2d(256),
            nn.Conv2d(256, 256, kernel_size=3, padding=1, bias=False),
            nn.MaxPool2d(kernel_size=3, stride=2),
            nn.ReLU(inplace=True),
            nn.BatchNorm2d(256)
        )
        self.classifier = nn.Sequential(
            nn.Linear(256 * 6 * 6, 4096, bias=False),
            nn.BatchNorm1d(4096),
            nn.ReLU(inplace=True),
            nn.Dropout(0.5),
            nn.Linear(4096, 4096, bias=False),
            nn.BatchNorm1d(4096),
            nn.ReLU(inplace=True),
            nn.Dropout(0.5),
            nn.Linear(4096, num_classes)
        )

        #self.regime = {
        #    0: {'optimizer': 'SGD', 'lr': 1e-2,
        #        'weight_decay': 5e-4, 'momentum': 0.9},
        #    10: {'lr': 5e-3},
        #    15: {'lr': 1e-3, 'weight_decay': 0},
        #    20: {'lr': 5e-4},
        #    25: {'lr': 1e-4}
        #}
        self.regime = {
            0: {'optimizer': 'SGD', 'lr': 1e-2,
                'weight_decay': 5e-4, 'momentum': 0.9},
            20: {'lr': 1e-3},
            40: {'lr': 1e-4}
        }
        normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                         std=[0.229, 0.224, 0.225])
        self.input_transform = {
            'train': transforms.Compose([
                transforms.Scale(256),
                transforms.RandomCrop(224),
                transforms.RandomHorizontalFlip(),
                transforms.ToTensor(),
                normalize
            ]),
            'eval': transforms.Compose([
                transforms.Scale(256),
                transforms.CenterCrop(224),
                transforms.ToTensor(),
                normalize
            ])
        }

    def forward(self, x):
        x = self.features(x)
        x = x.view(-1, 256 * 6 * 6)
        x = self.classifier(x)
        return x

self.features 的原始 AlexNet 代码在这里（我切换了上面的顺序；不确定这是否允许）：

self.features = nn.Sequential(
            nn.Conv2d(3, 64, kernel_size=11, stride=4, padding=2,
                      bias=False),
            nn.BatchNorm2d(64),
            nn.ReLU(inplace=True),
            nn.MaxPool2d(kernel_size=3, stride=2),

            nn.Conv2d(64, 192, kernel_size=5, padding=2, bias=False),
            nn.BatchNorm2d(192),
            nn.ReLU(inplace=True),
            nn.MaxPool2d(kernel_size=3, stride=2),

            nn.Conv2d(192, 384, kernel_size=3, padding=1, bias=False),
            nn.ReLU(inplace=True),
            nn.BatchNorm2d(384),
            nn.Conv2d(384, 256, kernel_size=3, padding=1, bias=False),
            nn.ReLU(inplace=True),
            nn.BatchNorm2d(256),
            nn.Conv2d(256, 256, kernel_size=3, padding=1, bias=False),
            nn.MaxPool2d(kernel_size=3, stride=2),
            nn.ReLU(inplace=True),
            nn.BatchNorm2d(256)
        )

我对如何更改特征和分类器感到特别困惑。 我想要的 model 的描述在这里： https://arxiv.org/pdf/1609.04836.pdf在第 B 页上。

非常感谢

Answer 1

这是代码：

class AlexNet(nn.Module):

    def __init__(self, num_classes=10):
        super(AlexNet, self).__init__()
        self.features = nn.Sequential(
            nn.Conv2d(3, 64, kernel_size=5, padding=2,
                      bias=False),
            nn.MaxPool2d(kernel_size=3, stride=2),
            nn.BatchNorm2d(64),
            nn.ReLU(inplace=True),
            nn.Conv2d(64, 64, kernel_size=5, padding=2, bias=False),
            nn.MaxPool2d(kernel_size=3, stride=2),
            nn.BatchNorm2d(64),
            nn.ReLU(inplace=True),
        )
        self.classifier = nn.Sequential(
            nn.Linear(64 * 7 * 7, 384, bias=False),
            nn.BatchNorm1d(384),
            nn.ReLU(inplace=True),
            nn.Dropout(0.5),
            nn.Linear(384, 192, bias=False),
            nn.BatchNorm1d(192),
            nn.ReLU(inplace=True),
            nn.Dropout(0.5),
            nn.Linear(192, num_classes)
        )
        self.regime = {
            0: {'optimizer': 'SGD', 'lr': 1e-3,
                'weight_decay': 5e-4, 'momentum': 0.9},
            60: {'lr': 1e-2},
            120: {'lr': 1e-3},
            180: {'lr': 1e-4}
        }

    def forward(self, x):
        x = self.features(x)
        x = x.view(-1, 64 * 7 * 7)
        x = self.classifier(x)
        return F.log_softmax(x)


def cifar10_shallow(**kwargs):
    num_classes = getattr(kwargs, 'num_classes', 10)
    return AlexNet(num_classes)


def cifar100_shallow(**kwargs):
    num_classes = getattr(kwargs, 'num_classes', 100)
    return AlexNet(num_classes)

魏文所著。 平滑存储库