-
Notifications
You must be signed in to change notification settings - Fork 0
/
models.py
152 lines (120 loc) · 6.53 KB
/
models.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
import numpy as np
import torch.nn as nn
import torch.nn.functional as F
import torch
from collections import OrderedDict
def get_model(name, n_classes, n_time, embedding_dim=128, n_layer=32, channels=1):
if name == "resnet":
return ResNet(n_classes, n_time, embedding_dim=embedding_dim, n_layer1=n_layer, n_layer2=n_layer, n_layer3=n_layer, n_layer4=n_layer)
elif name == "resnet_256":
return ResNet(n_classes, n_time, downpool_strides=[(2,1), (2,1), (2,1), (2,1)], embedding_dim=embedding_dim, n_layer1=n_layer, n_layer2=n_layer, n_layer3=n_layer, n_layer4=n_layer, channels=channels)
elif name == "resnet_512":
return ResNet(n_classes, n_time, downpool_strides=[(2,2), (2,1), (2,1), (2,1)], embedding_dim=embedding_dim, n_layer1=n_layer, n_layer2=n_layer, n_layer3=n_layer, n_layer4=n_layer, channels=channels)
elif name == "resnet_1024":
return ResNet(n_classes, n_time, downpool_strides=[(2,2), (2,2), (2,1), (2,1)], embedding_dim=embedding_dim, n_layer1=n_layer, n_layer2=n_layer, n_layer3=n_layer, n_layer4=n_layer, channels=channels)
elif name == "resnet_2048":
return ResNet(n_classes, n_time, downpool_strides=[(2,2), (2,2), (2,2), (2,1)], embedding_dim=embedding_dim, n_layer1=n_layer, n_layer2=n_layer, n_layer3=n_layer, n_layer4=n_layer, channels=channels)
elif name == "resnet_4096":
return ResNet(n_classes, n_time, downpool_strides=[(2,2), (2,2), (2,2), (2,2)], embedding_dim=embedding_dim, n_layer1=n_layer, n_layer2=n_layer, n_layer3=n_layer, n_layer4=n_layer, channels=channels)
elif name == "resnet_8192":
return ResNet(n_classes, n_time, downpool_strides=[(2,2), (2,2), (2,2), (2,2)], embedding_dim=embedding_dim, n_layer1=n_layer, n_layer2=n_layer, n_layer3=n_layer, n_layer4=n_layer, channels=channels)
elif name == "resnet_16384":
return ResNet(n_classes, n_time, downpool_strides=[(2,2), (2,2), (2,2), (2,2)], embedding_dim=embedding_dim, n_layer1=n_layer, n_layer2=n_layer, n_layer3=n_layer, n_layer4=n_layer, channels=channels)
elif name == "resnet_big":
return ResNet(n_classes, n_time, n_layer1=64, n_layer2=128, n_layer3=256)
else:
raise ValueError("model with name {} not defined ... ")
def conv3x3(in_planes, out_planes, stride=1):
"""3x3 convolution with padding"""
return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
padding=1, bias=False)
class BasicBlock(nn.Module):
expansion = 1
def __init__(self, inplanes, planes, stride=1, downsample=None, drop_rate=0.0, drop_block=False, block_size=1):
super(BasicBlock, self).__init__()
self.conv1 = conv3x3(inplanes, planes)
self.bn1 = nn.BatchNorm2d(planes)
self.relu = nn.LeakyReLU(0.1)
self.conv2 = conv3x3(planes, planes)
self.bn2 = nn.BatchNorm2d(planes)
self.conv3 = conv3x3(planes, planes)
self.bn3 = nn.BatchNorm2d(planes)
self.maxpool = nn.MaxPool2d(stride)
self.downsample = downsample
self.stride = stride
self.drop_rate = drop_rate
self.num_batches_tracked = 0
self.drop_block = drop_block
self.block_size = block_size
def forward(self, x):
self.num_batches_tracked += 1
residual = 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:
residual = self.downsample(x)
out += residual
out = self.relu(out)
out = self.maxpool(out)
out = F.dropout(out, p=self.drop_rate, training=self.training, inplace=True)
return out
class ResNet(nn.Module):
def __init__(self, n_classes, n_time, block=BasicBlock, keep_prob=1.0, avg_pool=True, drop_rate=0.1, dropblock_size=5, embedding_dim=128, n_layer1 = 32, n_layer2 = 32, n_layer3 = 32, n_layer4 = 32, downpool_strides=[(2,2), (2,2), (2,2), (2,2)], channels=1):
self.inplanes = channels
super(ResNet, self).__init__()
# settings
pooling_size = (4,2)
self.layer1 = self._make_layer(block, n_layer1, stride=downpool_strides[0], drop_rate=drop_rate)
self.layer2 = self._make_layer(block, n_layer2, stride=downpool_strides[1], drop_rate=drop_rate)
self.layer3 = self._make_layer(block, n_layer3, stride=downpool_strides[2], drop_rate=drop_rate, drop_block=True, block_size=dropblock_size)
self.layer4 = self._make_layer(block, n_layer4, stride=downpool_strides[3], drop_rate=drop_rate, drop_block=True, block_size=dropblock_size)
if avg_pool:
self.avgpool = nn.AvgPool2d(5, stride=1)
self.keep_prob = keep_prob
self.keep_avg_pool = avg_pool
self.dropout = nn.Dropout(p=1 - self.keep_prob, inplace=False)
self.drop_rate = drop_rate
self.pool = nn.AdaptiveAvgPool2d(pooling_size)
self.fc1 = nn.Linear(np.prod(pooling_size)*n_layer4, embedding_dim)
self.dropout = nn.Dropout(0.3)
self.fc2 = nn.Linear(embedding_dim, n_classes*n_time)
self.n_classes = n_classes
self.n_time = n_time
for m in self.modules():
if isinstance(m, nn.Conv2d):
nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='leaky_relu')
elif isinstance(m, nn.BatchNorm2d):
nn.init.constant_(m.weight, 1)
nn.init.constant_(m.bias, 0)
def _make_layer(self, block, planes, stride=1, drop_rate=0.0, drop_block=False, block_size=1):
downsample = None
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
nn.Conv2d(self.inplanes, planes * block.expansion,
kernel_size=1, stride=1, bias=False),
nn.BatchNorm2d(planes * block.expansion),
)
layers = []
layers.append(block(self.inplanes, planes, stride, downsample, drop_rate, drop_block, block_size))
self.inplanes = planes * block.expansion
return nn.Sequential(*layers)
def forward(self, x):
x = self.layer1(x)
x = self.layer2(x)
x = self.layer3(x)
x = self.layer4(x)
x = self.pool(x)
# flatten
x = x.view(x.size(0), -1)
x = self.fc1(x)
x_rep = F.relu(x)
x = self.dropout(x_rep)
x = self.fc2(x)
y_pred = x.view((-1, self.n_classes, self.n_time))
return y_pred, x_rep