resnet_CSI.py

'''ResNet in PyTorch.
BasicBlock and Bottleneck module is from the original ResNet paper:
[1] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
    Deep Residual Learning for Image Recognition. arXiv:1512.03385
PreActBlock and PreActBottleneck module is from the later paper:
[2] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
    Identity Mappings in Deep Residual Networks. arXiv:1603.05027
'''
import torch
import torch.nn as nn
import torch.nn.functional as F

from .base_model import BaseModel
from methods.methods_utils.transform_layers import NormalizeLayer

def conv3x3(in_planes, out_planes, stride=1):
    return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, padding=1, bias=False)

def conv1x1(in_planes, out_planes, stride=1):
    """1x1 convolution"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)


class BasicBlock(nn.Module):
    expansion = 1

    def __init__(self, in_planes, planes, stride=1):
        super(BasicBlock, self).__init__()
        self.conv1 = conv3x3(in_planes, planes, stride)
        self.conv2 = conv3x3(planes, planes)
        self.bn1 = nn.BatchNorm2d(planes)
        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 BasicBlock_224(nn.Module):
    expansion = 1

    def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
                 base_width=64, dilation=1, norm_layer=None):
        super(BasicBlock, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        if groups != 1 or base_width != 64:
            raise ValueError('BasicBlock only supports groups=1 and base_width=64')
        if dilation > 1:
            raise NotImplementedError("Dilation > 1 not supported in BasicBlock")
        # Both self.conv1 and self.downsample layers downsample the input when stride != 1
        self.conv1 = conv3x3(inplanes, planes, stride)
        self.bn1 = norm_layer(planes)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = conv3x3(planes, planes)
        self.bn2 = norm_layer(planes)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)

        if self.downsample is not None:
            identity = self.downsample(x)

        out += identity
        out = self.relu(out)

        return out

class PreActBlock(nn.Module):
    '''Pre-activation version of the BasicBlock.'''
    expansion = 1

    def __init__(self, in_planes, planes, stride=1):
        super(PreActBlock, self).__init__()
        self.conv1 = conv3x3(in_planes, planes, stride)
        self.conv2 = conv3x3(planes, planes)
        self.bn1 = nn.BatchNorm2d(in_planes)
        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)
            )

    def forward(self, x):
        out = F.relu(self.bn1(x))
        shortcut = self.shortcut(out)
        out = self.conv1(out)
        out = self.conv2(F.relu(self.bn2(out)))
        out += shortcut
        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.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
        self.conv3 = nn.Conv2d(planes, self.expansion*planes, kernel_size=1, bias=False)
        self.bn1 = nn.BatchNorm2d(planes)
        self.bn2 = nn.BatchNorm2d(planes)
        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 Bottleneck_224(nn.Module):
    # Bottleneck in torchvision places the stride for downsampling at 3x3 convolution(self.conv2)
    # while original implementation places the stride at the first 1x1 convolution(self.conv1)
    # according to "Deep residual learning for image recognition"https://arxiv.org/abs/1512.03385.
    # This variant is also known as ResNet V1.5 and improves accuracy according to
    # https://ngc.nvidia.com/catalog/model-scripts/nvidia:resnet_50_v1_5_for_pytorch.

    expansion = 4

    def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
                 base_width=64, dilation=1, norm_layer=None):
        super(Bottleneck, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        width = int(planes * (base_width / 64.)) * groups
        # Both self.conv2 and self.downsample layers downsample the input when stride != 1
        self.conv1 = conv1x1(inplanes, width)
        self.bn1 = norm_layer(width)
        self.conv2 = conv3x3(width, width, stride, groups, dilation)
        self.bn2 = norm_layer(width)
        self.conv3 = conv1x1(width, planes * self.expansion)
        self.bn3 = norm_layer(planes * self.expansion)
        self.relu = nn.ReLU(inplace=True)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)
        out = self.relu(out)

        out = self.conv3(out)
        out = self.bn3(out)

        if self.downsample is not None:
            identity = self.downsample(x)

        out += identity
        out = self.relu(out)

        return out

class PreActBottleneck(nn.Module):
    '''Pre-activation version of the original Bottleneck module.'''
    expansion = 4

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

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


class ResNet_32x32(BaseModel):
    def __init__(self, block, num_blocks, channel=3, num_classes=10, record_embedding: bool = False,
                 no_grad: bool = False):
        last_dim = 512 * block.expansion
        super(ResNet_32x32, self).__init__(last_dim, num_classes)

        self.in_planes = 64
        self.last_dim = last_dim

        self.normalize = NormalizeLayer()

        self.conv1 = conv3x3(3, 64)
        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)

    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 penultimate(self, x, all_features=False):
        out_list = []

        out = self.normalize(x)
        out = self.conv1(out)
        out = self.bn1(out)
        out = F.relu(out)
        out_list.append(out)

        out = self.layer1(out)
        out_list.append(out)
        out = self.layer2(out)
        out_list.append(out)
        out = self.layer3(out)
        out_list.append(out)
        out = self.layer4(out)
        out_list.append(out)

        out = F.avg_pool2d(out, 4)
        out = out.view(out.size(0), -1)

        if all_features:
            return out, out_list
        else:
            return out

class ResNet_224x224(BaseModel):
    def __init__(self, block, layers, channel, num_classes=10, record_embedding: bool = False,
                 no_grad: bool = False, zero_init_residual=False, groups=1, width_per_group=64,
                 replace_stride_with_dilation=None, norm_layer=None):
        last_dim = 512 * block.expansion
        super(ResNet_224x224, self).__init__(last_dim, num_classes)
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        self._norm_layer = norm_layer

        self.inplanes = 64
        self.dilation = 1
        if replace_stride_with_dilation is None:
            # each element in the tuple indicates if we should replace
            # the 2x2 stride with a dilated convolution instead
            replace_stride_with_dilation = [False, False, False]
        if len(replace_stride_with_dilation) != 3:
            raise ValueError("replace_stride_with_dilation should be None "
                             "or a 3-element tuple, got {}".format(replace_stride_with_dilation))
        self.groups = groups
        self.base_width = width_per_group
        self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3,
                               bias=False)
        self.bn1 = norm_layer(self.inplanes)
        self.relu = nn.ReLU(inplace=True)
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        self.layer1 = self._make_layer(block, 64, layers[0])
        self.layer2 = self._make_layer(block, 128, layers[1], stride=2,
                                       dilate=replace_stride_with_dilation[0])
        self.layer3 = self._make_layer(block, 256, layers[2], stride=2,
                                       dilate=replace_stride_with_dilation[1])
        self.layer4 = self._make_layer(block, 512, layers[3], stride=2,
                                       dilate=replace_stride_with_dilation[2])
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
        self.normalize = NormalizeLayer()
        self.last_dim = 512 * block.expansion

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)

        # Zero-initialize the last BN in each residual branch,
        # so that the residual branch starts with zeros, and each residual block behaves like an identity.
        # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677
        if zero_init_residual:
            for m in self.modules():
                if isinstance(m, Bottleneck):
                    nn.init.constant_(m.bn3.weight, 0)
                elif isinstance(m, BasicBlock):
                    nn.init.constant_(m.bn2.weight, 0)

    def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
        norm_layer = self._norm_layer
        downsample = None
        previous_dilation = self.dilation
        if dilate:
            self.dilation *= stride
            stride = 1
        if stride != 1 or self.inplanes != planes * block.expansion:
            downsample = nn.Sequential(
                conv1x1(self.inplanes, planes * block.expansion, stride),
                norm_layer(planes * block.expansion),
            )

        layers = []
        layers.append(block(self.inplanes, planes, stride, downsample, self.groups,
                            self.base_width, previous_dilation, norm_layer))
        self.inplanes = planes * block.expansion
        for _ in range(1, blocks):
            layers.append(block(self.inplanes, planes, groups=self.groups,
                                base_width=self.base_width, dilation=self.dilation,
                                norm_layer=norm_layer))

        return nn.Sequential(*layers)

    def penultimate(self, x, all_features=False):
        # See note [TorchScript super()]
        out_list = []

        x = self.normalize(x)
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.maxpool(x)
        out_list.append(x)

        x = self.layer1(x)
        out_list.append(x)
        x = self.layer2(x)
        out_list.append(x)
        x = self.layer3(x)
        out_list.append(x)
        x = self.layer4(x)
        out_list.append(x)

        x = self.avgpool(x)
        x = torch.flatten(x, 1)

        if all_features:
            return x, out_list
        else:
            return x


def ResNet(arch: str, channel: int, num_classes: int, im_size, record_embedding: bool = False, no_grad: bool = False):
    arch = arch.lower()
    if im_size[0] == 224 and im_size[1] == 224:
        if arch == "resnet18":
            net = ResNet_224x224(BasicBlock_224, [2, 2, 2, 2], channel=channel, num_classes=num_classes,
                                 record_embedding=record_embedding, no_grad=no_grad)
        elif arch == "resnet34":
            net = ResNet_224x224(BasicBlock_224, [3, 4, 6, 3], channel=channel, num_classes=num_classes,
                                 record_embedding=record_embedding, no_grad=no_grad)
        elif arch == "resnet50":
            net = ResNet_224x224(Bottleneck_224, [3, 4, 6, 3], channel=channel, num_classes=num_classes,
                                 record_embedding=record_embedding, no_grad=no_grad)
        elif arch == "resnet101":
            net = ResNet_224x224(Bottleneck_224, [3, 4, 23, 3], channel=channel, num_classes=num_classes,
                                 record_embedding=record_embedding, no_grad=no_grad)
        elif arch == "resnet152":
            net = ResNet_224x224(Bottleneck_224, [3, 8, 36, 3], channel=channel, num_classes=num_classes,
                                 record_embedding=record_embedding, no_grad=no_grad)
        else:
            raise ValueError("Model architecture not found.")
    elif (channel == 1 and im_size[0] == 28 and im_size[1] == 28) or (
            channel == 3 and im_size[0] == 32 and im_size[1] == 32):
        if arch == "resnet18":
            net = ResNet_32x32(BasicBlock, [2, 2, 2, 2], channel=channel, num_classes=num_classes,
                               record_embedding=record_embedding, no_grad=no_grad)
        elif arch == "resnet34":
            net = ResNet_32x32(BasicBlock, [3, 4, 6, 3], channel=channel, num_classes=num_classes,
                               record_embedding=record_embedding, no_grad=no_grad)
        elif arch == "resnet50":
            net = ResNet_32x32(Bottleneck, [3, 4, 6, 3], channel=channel, num_classes=num_classes,
                               record_embedding=record_embedding, no_grad=no_grad)
        elif arch == "resnet101":
            net = ResNet_32x32(Bottleneck, [3, 4, 23, 3], channel=channel, num_classes=num_classes,
                               record_embedding=record_embedding, no_grad=no_grad)
        elif arch == "resnet152":
            net = ResNet_32x32(Bottleneck, [3, 8, 36, 3], channel=channel, num_classes=num_classes,
                               record_embedding=record_embedding, no_grad=no_grad)
        else:
            raise ValueError("Model architecture not found.")
    else:
        raise NotImplementedError("Network Architecture for current dataset has not been implemented.")
    return net

def ResNet18_CSI(channel: int, num_classes: int, im_size, record_embedding: bool = False, no_grad: bool = False):
    return ResNet("resnet18", channel, num_classes, im_size, record_embedding, no_grad)

def ResNet34_CSI(channel: int, num_classes: int, im_size, record_embedding: bool = False, no_grad: bool = False):
    return ResNet("resnet34", channel, num_classes, im_size, record_embedding, no_grad)

def ResNet50_CSI(channel: int, num_classes: int, im_size, record_embedding: bool = False, no_grad: bool = False):
    return ResNet("resnet50", channel, num_classes, im_size, record_embedding, no_grad)