commit

Author: z-hXu

cifar/evaluate.sh

@@ -1,7 +1,7 @@
 python -u main.py \
 --gpus 0 \
 -e [best_model_path] \
---model resnet20_1w1a \
+--model resnet20_bireal_1w1a \
 --data_path [DATA_PATH] \
 --dataset cifar10 \
 -bt 128 \

cifar/models_bnn/__init__.py

@@ -1,3 +1,4 @@
 from .resnet import *
+from .resnet_bireal import *
 from .resnet2 import *
 from .vgg import * 

cifar/models_bnn/resnet.py

@@ -104,11 +104,9 @@ def __init__(self, in_planes, planes, stride=1, option='A'):
 
     def forward(self, x):
         out = self.bn1(self.conv1(x))
-        out += self.shortcut(x)
         out = F.hardtanh(out)
-        x1 = out
         out = self.bn2(self.conv2(out))
-        out += x1
+        out += self.shortcut(x)
         out = F.hardtanh(out)
         return out
 

cifar/models_bnn/resnet_bireal.py

@@ -0,0 +1,200 @@
+'''
+Properly implemented ResNet-s for CIFAR10 as described in paper [1].
+
+The implementation and structure of this file is hugely influenced by [2]
+which is implemented for ImageNet and doesn't have option A for identity.
+Moreover, most of the implementations on the web is copy-paste from
+torchvision's resnet and has wrong number of params.
+
+Proper ResNet-s for CIFAR10 (for fair comparision and etc.) has following
+number of layers and parameters:
+
+name      | layers | params
+ResNet20  |    20  | 0.27M
+ResNet32  |    32  | 0.46M
+ResNet44  |    44  | 0.66M
+ResNet56  |    56  | 0.85M
+ResNet110 |   110  |  1.7M
+ResNet1202|  1202  | 19.4m
+
+which this implementation indeed has.
+
+Reference:
+[1] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
+    Deep Residual Learning for Image Recognition. arXiv:1512.03385
+[2] https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py
+
+If you use this implementation in you work, please don't forget to mention the
+author, Yerlan Idelbayev.
+'''
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.nn.init as init
+from modules import *
+
+from torch.autograd import Variable
+
+__all__ = ['resnet20_bireal_1w1a']
+
+
+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 BasicBlock_1w1a(nn.Module):
+    expansion = 1
+
+    def __init__(self, in_planes, planes, stride=1, option='A'):
+        super(BasicBlock_1w1a, self).__init__()
+        self.conv1 = BinarizeConv2d(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
+        self.bn1 = nn.BatchNorm2d(planes)
+        self.conv2 = BinarizeConv2d(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(
+                     BinarizeConv2d(in_planes, self.expansion * planes, kernel_size=1, stride=stride, bias=False),
+                     nn.BatchNorm2d(self.expansion * planes)
+                )
+
+    def forward(self, x):
+        out = self.bn1(self.conv1(x))
+        out += self.shortcut(x)
+        out = F.hardtanh(out)
+        x1 = out
+        out = self.bn2(self.conv2(out))
+        out += x1
+        out = F.hardtanh(out)
+        return out
+
+
+class ResNet(nn.Module):
+    def __init__(self, block, num_blocks, num_classes=10):
+        super(ResNet, 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)
+        self.bn2 = nn.BatchNorm1d(64)
+        self.linear = nn.Linear(64, num_classes)
+
+        for m in self.modules():
+            if isinstance(m, nn.BatchNorm2d):
+                m.weight.data.fill_(1e-8)
+                m.bias.data.zero_()
+            if isinstance(m, nn.Linear) or isinstance(m, nn.Conv2d):
+                init.kaiming_normal_(m.weight)
+
+    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.hardtanh(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])
+        out = out.view(out.size(0), -1) 
+        out = self.bn2(out)
+        out = self.linear(out)
+
+        return out
+
+
+def resnet20_bireal_1w1a(**kwargs):
+    return ResNet(BasicBlock_1w1a, [3, 3, 3],**kwargs)
+
+
+def resnet20():
+    return ResNet(BasicBlock, [3, 3, 3])
+
+
+def resnet32():
+    return ResNet(BasicBlock, [5, 5, 5])
+
+
+def resnet44():
+    return ResNet(BasicBlock, [7, 7, 7])
+
+
+def resnet56():
+    return ResNet(BasicBlock, [9, 9, 9])
+
+
+def resnet110():
+    return ResNet(BasicBlock, [18, 18, 18])
+
+
+def resnet1202():
+    return ResNet(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()

cifar/run.sh

@@ -1,6 +1,6 @@
 python -u main.py \
 --gpus 0 \
---model resnet20_1w1a \
+--model resnet20_bireal_1w1a \
 --results_dir [DIR] \
 --data_path [DATA_PATH] \
 --dataset cifar10 \

cifar/utils/options.py

@@ -43,7 +43,7 @@
     '--model',
     '-a',
     metavar='MODEL',
-    default='resnet20_1w1a',
+    default='resnet20_bireal_1w1a',
     help='model architecture ')
 
 parser.add_argument(
@@ -55,7 +55,7 @@
 parser.add_argument(
     '--data_path',
     type=str,
-    default='/home/xuzihan/data',
+    default='/home/data',
     help='The dictionary where the dataset is stored.')
 
 parser.add_argument(

imagenet/models_bnn/__init__.py

@@ -1,3 +1,4 @@
 from .resnet import *
+from .resnet_bireal import *
 from .resnet2 import *
 from .vgg import *