-
Notifications
You must be signed in to change notification settings - Fork 20
/
wrn_mixup_model.py
187 lines (147 loc) · 6.51 KB
/
wrn_mixup_model.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
### dropout has been removed in this code. original code had dropout#####
import torch
import torch.nn as nn
import torch.nn.init as init
import torch.nn.functional as F
from torch.autograd import Variable
import sys, os
import numpy as np
import random
act = torch.nn.ReLU()
import backbone
import math
from torch.nn.utils.weight_norm import WeightNorm
class BasicBlock(nn.Module):
def __init__(self, in_planes, out_planes, stride, dropRate=0.0):
super(BasicBlock, self).__init__()
self.bn1 = nn.BatchNorm2d(in_planes)
self.relu1 = nn.ReLU(inplace=True)
self.conv1 = nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
padding=1, bias=False)
self.bn2 = nn.BatchNorm2d(out_planes)
self.relu2 = nn.ReLU(inplace=True)
self.conv2 = nn.Conv2d(out_planes, out_planes, kernel_size=3, stride=1,
padding=1, bias=False)
self.droprate = dropRate
self.equalInOut = (in_planes == out_planes)
self.convShortcut = (not self.equalInOut) and nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride,
padding=0, bias=False) or None
def forward(self, x):
if not self.equalInOut:
x = self.relu1(self.bn1(x))
else:
out = self.relu1(self.bn1(x))
out = self.relu2(self.bn2(self.conv1(out if self.equalInOut else x)))
if self.droprate > 0:
out = F.dropout(out, p=self.droprate, training=self.training)
out = self.conv2(out)
return torch.add(x if self.equalInOut else self.convShortcut(x), out)
class NetworkBlock(nn.Module):
def __init__(self, nb_layers, in_planes, out_planes, block, stride, dropRate=0.0):
super(NetworkBlock, self).__init__()
self.layer = self._make_layer(block, in_planes, out_planes, nb_layers, stride, dropRate)
def _make_layer(self, block, in_planes, out_planes, nb_layers, stride, dropRate):
layers = []
for i in range(int(nb_layers)):
layers.append(block(i == 0 and in_planes or out_planes, out_planes, i == 0 and stride or 1, dropRate))
return nn.Sequential(*layers)
def forward(self, x):
return self.layer(x)
def to_one_hot(inp,num_classes):
y_onehot = torch.FloatTensor(inp.size(0), num_classes)
if torch.cuda.is_available():
y_onehot = y_onehot.cuda()
y_onehot.zero_()
x = inp.type(torch.LongTensor)
if torch.cuda.is_available():
x = x.cuda()
x = torch.unsqueeze(x , 1)
y_onehot.scatter_(1, x , 1)
return Variable(y_onehot,requires_grad=False)
# return y_onehot
def mixup_data(x, y, lam):
'''Compute the mixup data. Return mixed inputs, pairs of targets, and lambda'''
batch_size = x.size()[0]
index = torch.randperm(batch_size)
if torch.cuda.is_available():
index = index.cuda()
mixed_x = lam * x + (1 - lam) * x[index,:]
y_a, y_b = y, y[index]
return mixed_x, y_a, y_b, lam
class WideResNet(nn.Module):
def __init__(self, depth=28, widen_factor=10, num_classes= 200 , loss_type = 'dist', per_img_std = False, stride = 1 , dropRate=0.5 ):
flatten = True
super(WideResNet, self).__init__()
nChannels = [16, 16*widen_factor, 32*widen_factor, 64*widen_factor]
assert((depth - 4) % 6 == 0)
n = (depth - 4) / 6
block = BasicBlock
# 1st conv before any network block
self.conv1 = nn.Conv2d(3, nChannels[0], kernel_size=3, stride=1,
padding=1, bias=False)
# 1st block
self.block1 = NetworkBlock(n, nChannels[0], nChannels[1], block, stride, dropRate)
# 2nd block
self.block2 = NetworkBlock(n, nChannels[1], nChannels[2], block, 2, dropRate)
# 3rd block
self.block3 = NetworkBlock(n, nChannels[2], nChannels[3], block, 2, dropRate)
# global average pooling and linear
self.bn1 = nn.BatchNorm2d(nChannels[3])
self.relu = nn.ReLU(inplace=True)
self.nChannels = nChannels[3]
if loss_type == 'softmax':
self.linear = nn.Linear(nChannels[3], int(num_classes))
self.linear.bias.data.fill_(0)
else:
self.linear = backbone.distLinear(nChannels[3], int(num_classes))
self.num_classes = num_classes
if flatten:
self.final_feat_dim = 640
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2. / n))
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
def forward(self, x, target= None, mixup=False, mixup_hidden=True, mixup_alpha=None , lam = 0.4):
if target is not None:
if mixup_hidden:
layer_mix = random.randint(0,3)
elif mixup:
layer_mix = 0
else:
layer_mix = None
out = x
target_a = target_b = target
if layer_mix == 0:
out, target_a , target_b , lam = mixup_data(out, target, lam=lam)
out = self.conv1(out)
out = self.block1(out)
if layer_mix == 1:
out, target_a , target_b , lam = mixup_data(out, target, lam=lam)
out = self.block2(out)
if layer_mix == 2:
out, target_a , target_b , lam = mixup_data(out, target, lam=lam)
out = self.block3(out)
if layer_mix == 3:
out, target_a , target_b , lam = mixup_data(out, target, lam=lam)
out = self.relu(self.bn1(out))
out = F.avg_pool2d(out, out.size()[2:])
out = out.view(out.size(0), -1)
out1 = self.linear(out)
return out , out1 , target_a , target_b
else:
out = x
out = self.conv1(out)
out = self.block1(out)
out = self.block2(out)
out = self.block3(out)
out = self.relu(self.bn1(out))
out = F.avg_pool2d(out, out.size()[2:])
out = out.view(out.size(0), -1)
out1 = self.linear(out)
return out , out1
def wrn28_10(num_classes=10 , drop_rate= 0.5, loss_type = 'dist'):
model = WideResNet(depth=28, widen_factor=10, num_classes=num_classes, loss_type = loss_type , per_img_std = False, stride = 1 ,dropRate = drop_rate)
return model