add base models

models/__init__.py

@@ -0,0 +1,2 @@
+from .resnet_cifar import *
+from .resnet import *

models/resnet.py

@@ -0,0 +1,162 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+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 = BN(planes)
+        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
+        self.bn2 = BN(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),
+                BN(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 = BN(planes)
+        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
+        self.bn2 = BN(planes)
+        self.conv3 = nn.Conv2d(planes, self.expansion * planes, kernel_size=1, bias=False)
+        self.bn3 = BN(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),
+                BN(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, use_norm=False, return_features=False):
+        super(ResNet, self).__init__()
+
+        global BN
+        BN = nn.BatchNorm2d
+
+        self.in_planes = 64
+        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False)
+        self.layer_one = self.conv1
+        self.bn1 = BN(64)
+        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
+        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.return_encoding = return_features
+
+    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):
+        self.layer_one_out = self.conv1(x)
+        self.layer_one_out.requires_grad_()
+        self.layer_one_out.retain_grad()
+        out = F.relu(self.bn1(self.layer_one_out))
+        out = self.maxpool(out)
+        out = self.layer1(out)
+        out = self.layer2(out)
+        out = self.layer3(out)
+        out = self.layer4(out)
+        out = F.avg_pool2d(out, 7)
+        encoding = out.view(out.size(0), -1)
+        out = self.linear(encoding)
+        if self.return_encoding:
+            return out, encoding
+        else:
+            return out
+
+
+class ResNetCIFAR(nn.Module):
+    def __init__(self, block, num_blocks, num_classes=10):
+        super(ResNetCIFAR, self).__init__()
+
+        global BN
+        BN = nn.BatchNorm2d
+
+        self.in_planes = 16
+        self.conv1 = nn.Conv2d(3, 16, kernel_size=3, stride=1, padding=1, bias=False)
+        self.bn1 = BN(16)
+        self.layer1 = self._make_layer(block, 16, num_blocks[0], stride=1)
+        self.layer2 = self._make_layer(block, 32, num_blocks[1], stride=2)
+        self.layer3 = self._make_layer(block, 64, num_blocks[2], stride=2)
+        self.linear = nn.Linear(64 * block.expansion, num_classes)
+
+    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 = F.avg_pool2d(out, 8)
+        out = out.view(out.size(0), -1)
+        out = self.linear(out)
+        return out
+
+
+def resnet18():
+    return ResNet(BasicBlock, [2, 2, 2, 2])
+
+
+def resnet34():
+    return ResNet(BasicBlock, [3, 4, 6, 3])
+
+
+def resnet50(**kwargs):
+    return ResNet(Bottleneck, [3, 4, 6, 3], **kwargs)
+
+
+def resnet101():
+    return ResNet(Bottleneck, [3, 4, 23, 3])
+
+
+def resnet152():
+    return ResNet(Bottleneck, [3, 8, 36, 3])
+
+
+if __name__ == '__main__':
+    net = resnet50()
+    y = net(torch.randn(1, 3, 32, 32))
+    print(y.size())

models/resnet_cifar.py

@@ -0,0 +1,150 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.nn.init as init
+from torch.nn import Parameter
+
+__all__ = ['ResNet_s', 'resnet20', 'resnet32', 'resnet44', 'resnet56', 'resnet110', 'resnet1202']
+
+
+def _weights_init(m):
+    classname = m.__class__.__name__
+    if isinstance(m, nn.Linear) or isinstance(m, nn.Conv2d):
+        init.kaiming_normal_(m.weight)
+
+
+class NormedLinear(nn.Module):
+
+    def __init__(self, in_features, out_features):
+        super(NormedLinear, self).__init__()
+        self.weight = Parameter(torch.Tensor(in_features, out_features))
+        self.weight.data.uniform_(-1, 1).renorm_(2, 1, 1e-5).mul_(1e5)
+
+    def forward(self, x):
+        out = F.normalize(x, dim=1).mm(F.normalize(self.weight, dim=0))
+        return out
+
+
+class LambdaLayer(nn.Module):
+
+    def __init__(self, lambd):
+        super(LambdaLayer, self).__init__()
+        self.lambd = lambd
+
+    def forward(self, x):
+        return self.lambd(x)
+
+
+class BasicBlock(nn.Module):
+    expansion = 1
+
+    def __init__(self, in_planes, planes, stride=1, option='A'):
+        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 != planes:
+            if option == 'A':
+                """
+                For CIFAR10 ResNet paper uses option A.
+                """
+                self.shortcut = LambdaLayer(lambda x:
+                                            F.pad(x[:, :, ::2, ::2], (0, 0, 0, 0, planes//4, planes//4), "constant", 0))
+            elif option == 'B':
+                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 ResNet_s(nn.Module):
+
+    def __init__(self, block, num_blocks, num_classes=10, use_norm=False, return_features=False):
+        super(ResNet_s, self).__init__()
+        self.in_planes = 16
+
+        self.conv1 = nn.Conv2d(3, 16, kernel_size=3, stride=1, padding=1, bias=False)
+        self.bn1 = nn.BatchNorm2d(16)
+        self.layer1 = self._make_layer(block, 16, num_blocks[0], stride=1)
+        self.layer2 = self._make_layer(block, 32, num_blocks[1], stride=2)
+        self.layer3 = self._make_layer(block, 64, num_blocks[2], stride=2)
+        if use_norm:
+            self.fc = NormedLinear(64, num_classes)
+        else:
+            self.fc = nn.Linear(64, num_classes)
+        self.apply(_weights_init)
+        self.return_encoding = return_features
+
+    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 = F.avg_pool2d(out, out.size()[3])
+        encoding = out.view(out.size(0), -1)
+        out = self.fc(encoding)
+        if self.return_encoding:
+            return out, encoding
+        else:
+            return out
+
+
+def resnet20():
+    return ResNet_s(BasicBlock, [3, 3, 3])
+
+
+def resnet32(num_classes=10, use_norm=False, return_features=False):
+    return ResNet_s(BasicBlock, [5, 5, 5], num_classes=num_classes, use_norm=use_norm, return_features=return_features)
+
+
+def resnet44():
+    return ResNet_s(BasicBlock, [7, 7, 7])
+
+
+def resnet56():
+    return ResNet_s(BasicBlock, [9, 9, 9])
+
+
+def resnet110():
+    return ResNet_s(BasicBlock, [18, 18, 18])
+
+
+def resnet1202():
+    return ResNet_s(BasicBlock, [200, 200, 200])
+
+
+def test(net):
+    import numpy as np
+    total_params = 0
+
+    for x in filter(lambda p: p.requires_grad, net.parameters()):
+        total_params += np.prod(x.data.numpy().shape)
+    print("Total number of params", total_params)
+    print("Total layers", len(list(filter(lambda p: p.requires_grad and len(p.data.size()) > 1, net.parameters()))))
+
+
+if __name__ == "__main__":
+    for net_name in __all__:
+        if net_name.startswith('resnet'):
+            print(net_name)
+            test(globals()[net_name]())
+            print()