models.py

import torch
import torch.nn as nn
import torch.nn.functional as F
import mlconfig
import torchvision
mlconfig.register(torchvision.models.resnet50)
mlconfig.register(torch.optim.SGD)
mlconfig.register(torch.optim.Adam)
mlconfig.register(torch.optim.lr_scheduler.MultiStepLR)
mlconfig.register(torch.optim.lr_scheduler.CosineAnnealingLR)
mlconfig.register(torch.optim.lr_scheduler.StepLR)
mlconfig.register(torch.optim.lr_scheduler.ExponentialLR)


class ConvBrunch(nn.Module):
    def __init__(self, in_planes, out_planes, kernel_size=3):
        super(ConvBrunch, self).__init__()
        padding = (kernel_size - 1) // 2
        self.out_conv = nn.Sequential(
            nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, padding=padding),
            nn.BatchNorm2d(out_planes),
            nn.ReLU())

    def forward(self, x):
        return self.out_conv(x)


@mlconfig.register
class ToyModel(nn.Module):
    def __init__(self, type='CIFAR10'):
        super(ToyModel, self).__init__()
        self.type = type
        if type == 'CIFAR10':
            self.block1 = nn.Sequential(
                ConvBrunch(3, 64, 3),
                ConvBrunch(64, 64, 3),
                nn.MaxPool2d(kernel_size=2, stride=2))
            self.block2 = nn.Sequential(
                ConvBrunch(64, 128, 3),
                ConvBrunch(128, 128, 3),
                nn.MaxPool2d(kernel_size=2, stride=2))
            self.block3 = nn.Sequential(
                ConvBrunch(128, 196, 3),
                ConvBrunch(196, 196, 3),
                nn.MaxPool2d(kernel_size=2, stride=2))
            # self.global_avg_pool = nn.AdaptiveAvgPool2d(1)
            self.fc1 = nn.Sequential(
                nn.Linear(4*4*196, 256),
                nn.BatchNorm1d(256),
                nn.ReLU())
            self.fc2 = nn.Linear(256, 10)
            self.fc_size = 4*4*196
        elif type == 'MNIST':
            self.block1 = nn.Sequential(
                ConvBrunch(1, 32, 3),
                nn.MaxPool2d(kernel_size=2, stride=2))
            self.block2 = nn.Sequential(
                ConvBrunch(32, 64, 3),
                nn.MaxPool2d(kernel_size=2, stride=2))
            # self.global_avg_pool = nn.AdaptiveAvgPool2d(1)
            self.fc1 = nn.Sequential(
                nn.Linear(64*7*7, 128),
                nn.BatchNorm1d(128),
                nn.ReLU())
            self.fc2 = nn.Linear(128, 10)
            self.fc_size = 64*7*7
        self._reset_prams()

    def _reset_prams(self):
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_uniform_(m.weight, mode='fan_in', nonlinearity='relu')
            elif isinstance(m, nn.Linear):
                nn.init.xavier_uniform_(m.weight)
        return

    def forward(self, x):
        x = self.block1(x)
        x = self.block2(x)
        x = self.block3(x) if self.type == 'CIFAR10' else x
        # x = self.global_avg_pool(x)
        # x = x.view(x.shape[0], -1)
        x = x.view(-1, self.fc_size)
        x = self.fc1(x)
        x = self.fc2(x)
        return x


'''ResNet in PyTorch.
For Pre-activation ResNet, see 'preact_resnet.py'.
Reference:
[1] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
    Deep Residual Learning for Image Recognition. arXiv:1512.03385
'''


class BasicBlock(nn.Module):
    expansion = 1

    def __init__(self, in_planes, planes, stride=1):
        super(BasicBlock, self).__init__()
        self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(planes)

        self.shortcut = nn.Sequential()
        if stride != 1 or in_planes != self.expansion*planes:
            self.shortcut = nn.Sequential(
                nn.Conv2d(in_planes, self.expansion*planes, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(self.expansion*planes)
            )

    def forward(self, x):
        out = F.relu(self.bn1(self.conv1(x)))
        out = self.bn2(self.conv2(out))
        out += self.shortcut(x)
        out = F.relu(out)
        return out


class Bottleneck(nn.Module):
    expansion = 4

    def __init__(self, in_planes, planes, stride=1):
        super(Bottleneck, self).__init__()
        self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=1, bias=False)
        self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(planes)
        self.conv3 = nn.Conv2d(planes, self.expansion*planes, kernel_size=1, bias=False)
        self.bn3 = nn.BatchNorm2d(self.expansion*planes)

        self.shortcut = nn.Sequential()
        if stride != 1 or in_planes != self.expansion*planes:
            self.shortcut = nn.Sequential(
                nn.Conv2d(in_planes, self.expansion*planes, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(self.expansion*planes)
            )

    def forward(self, x):
        out = F.relu(self.bn1(self.conv1(x)))
        out = F.relu(self.bn2(self.conv2(out)))
        out = self.bn3(self.conv3(out))
        out += self.shortcut(x)
        out = F.relu(out)
        return out


class ResNet(nn.Module):
    def __init__(self, block, num_blocks, num_classes=10):
        super(ResNet, self).__init__()
        self.in_planes = 64

        self.conv1 = nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(64)
        self.layer1 = self._make_layer(block, 64, num_blocks[0], stride=1)
        self.layer2 = self._make_layer(block, 128, num_blocks[1], stride=2)
        self.layer3 = self._make_layer(block, 256, num_blocks[2], stride=2)
        self.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2)
        self.linear = nn.Linear(512*block.expansion, num_classes)
        self._reset_prams()

    def _make_layer(self, block, planes, num_blocks, stride):
        strides = [stride] + [1]*(num_blocks-1)
        layers = []
        for stride in strides:
            layers.append(block(self.in_planes, planes, stride))
            self.in_planes = planes * block.expansion
        return nn.Sequential(*layers)

    def forward(self, x):
        out = F.relu(self.bn1(self.conv1(x)))
        out = self.layer1(out)
        out = self.layer2(out)
        out = self.layer3(out)
        out = self.layer4(out)
        out = F.avg_pool2d(out, 4)
        out = out.view(out.size(0), -1)
        out = self.linear(out)
        return out

    def _reset_prams(self):
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_uniform_(m.weight, mode='fan_in', nonlinearity='relu')
            elif isinstance(m, nn.Linear):
                nn.init.xavier_uniform_(m.weight)
        return


@mlconfig.register
def ResNet18(num_classes=10):
    return ResNet(BasicBlock, [2, 2, 2, 2], num_classes=num_classes)


@mlconfig.register
def ResNet34(num_classes=10):
    return ResNet(BasicBlock, [3, 4, 6, 3], num_classes=num_classes)


@mlconfig.register
def ResNet50(num_classes=10):
    return ResNet(Bottleneck, [3, 4, 6, 3], num_classes=num_classes)


@mlconfig.register
def ResNet101(num_classes=10):
    return ResNet(Bottleneck, [3, 4, 23, 3], num_classes=num_classes)


@mlconfig.register
def ResNet152(num_classes=10):
    return ResNet(Bottleneck, [3, 8, 36, 3], num_classes=num_classes)