-
Notifications
You must be signed in to change notification settings - Fork 0
/
models.py
161 lines (113 loc) · 4.5 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
153
154
155
156
157
158
159
160
import os
from settings import cfg
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision.models as models
class Ensemble(nn.Module):
def __init__(self, modelA, modelB, modelC, nb_classes=13):
super(Ensemble, self).__init__()
self.modelA = modelA
self.modelB = modelB
self.modelC = modelC
# self.modelD = modelD
# self.modelE = modelE
self.modelA.fc = nn.Identity()
self.modelB.classifier = nn.Identity()
self.modelC.fc = nn.Identity()
# self.modelD.fc = nn.Identity()
# self.modelE.fc = nn.Identity()
self.classifier = nn.Linear(2048+2048+1024, cfg.num_classes)
def forward(self, x):
x1 = self.modelA(x.clone())
x1 = x1.view(x1.size(0), -1)
x2 = self.modelB(x.clone())
x2 = x2.view(x2.size(0), -1)
if self.modelC.training:
x3, aux = self.modelC(x.clone())
else:
x3 = self.modelC(x.clone())
x3 = x3.view(x3.size(0), -1)
# x4 = self.modelD(x.clone())
# x4 = x4.view(x4.size(0), -1)
# x5 = self.modelE(x.clone())
# x5 = x5.view(x5.size(0), -1)
x = torch.cat((x1, x2, x3), dim=1)
x = self.classifier(F.relu(x))
return x
def trainable_parameters(model):
for name, param in model.named_parameters():
if param.requires_grad:
print(name)
def pretrained_net():
# modelA = models.densenet121(pretrained=True, progress=True)
# modelB = models.densenet161(pretrained=True, progress=True)
# modelC = models.densenet169(pretrained=True, progress=True)
# modelD = models.densenet121(pretrained=True, progress=True)
# modelE = models.resnet34(pretrained=True, progress=True)
# for name, param in modelD.named_parameters():
# if param.requires_grad:
# print(name)
modelA = models.resnext50_32x4d(pretrained=True)
modelB = models.densenet121(pretrained=True)
modelC = models.inception_v3(pretrained=True)
# modelD = models.resnet18(pretrained=True)
# modelE = models.wide_resnet50_2(pretrained=True)
# modelC = models.resnext50_32x4d(pretrained=True, progress=True)
# modelD = models.resnext101_32x8d(pretrained=True, progress=True)
# modelE = models.resnet34(pretrained=True, progress=True)
# for name, param in modelD.named_parameters():
# print(name)
for param in modelA.parameters():
param.requires_grad_(False)
for param in modelB.parameters():
param.requires_grad_(False)
for param in modelC.parameters():
param.requires_grad_(False)
# for param in modelD.parameters():
# param.requires_grad_(False)
# for param in modelE.parameters():
# param.requires_grad_(False)
layers = [
modelA.layer4,
modelB.features.denseblock4.denselayer16,
]
for layer in layers:
for param in layer.parameters():
param.requires_grad = True
model = Ensemble(modelA, modelB, modelC)
# model_path = os.path.join(tc.model_dir, tc.model_name)
# pretrained_model_path = os.path.join(tc.root_dir,
# tc.model_dir,
# tc.model_name + '_pretrained.pt')
# try:
# model = torch.load(pretrained_model_path)
# except FileNotFoundError:
# os.environ['TORCH_HOME'] = tc.root_dir
# model = models.densenet121(pretrained=True, progress=True)
# torch.save(model, pretrained_model_path)
# model = torch.load(pretrained_model_path)
return model
def save_model(model, device):
model_dir = os.path.join(cfg.root_dir, cfg.model_dir)
model_path = os.path.join(model_dir, cfg.description)
if not os.path.exists(model_dir):
os.makedirs(model_dir)
if device == 'cuda':
model.to('cpu')
torch.save(model.state_dict(), model_path + '.pt' )
if device == 'cuda':
model.to('cuda')
return
def load_model(model):
model_path = os.path.join(cfg.root_dir, cfg.model_dir, cfg.description)
model.load_state_dict(torch.load(model_path + '.pt'))
return model
def load_multiple_models(model):
model_1 = load_model(model=model, fold=0)
model_2 = load_model(model=model, fold=1)
model_3 = load_model(model=model, fold=2)
model_4 = load_model(model=model, fold=3)
model_5 = load_model(model=model, fold=4)
models = [model_1, model_2, model_3, model_4, model_5]
return models