-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain.py
226 lines (194 loc) · 9.17 KB
/
main.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
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
import os
import os.path
import sys
import random
import numpy as np
import logging
import argparse
import torch
import torchvision
import pytorch_ssim
from torch.utils.tensorboard import SummaryWriter
import global_v as glv
from network_parser import parse
from datasets import load_dataset_snn
from utils import *
import fsvae_models.fsvae as fsvae
from fsvae_models.snn_layers import LIFSpike
from scipy.spatial.distance import cdist
from itertools import combinations
from dmsvdd import kmeans_plus_plus, update_c, update_r, get_c_min_max, KMeansPlusPlus, distance
# 定义前置超参数
init_device = torch.device("cuda:0")
device = init_device
max_accuracy = 0
min_loss = 1000
nu = 0.1
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--name', type=str)
parser.add_argument('--checkpoint', action='store', dest='checkpoint',
help='The path of checkpoint, if use checkpoint')
parser.add_argument('--config', action='store', dest='config', help='The path of config file')
parser.add_argument('--device', type=int)
# 读取一些信息，并新建一些文件夹
try:
args = parser.parse_args()
except:
parser.print_help()
exit(0)
if args.config is None:
raise Exception('Unrecognized config file.')
if args.device is None:
init_device = torch.device("cuda:0")
else:
init_device = torch.device(f"cuda:{args.device}")
os.makedirs(f'checkpoint/{args.name}', exist_ok=True)
writer = SummaryWriter(log_dir=f'checkpoint/{args.name}/tb')
logging.basicConfig(filename=f'checkpoint/{args.name}/{args.name}.log', level=logging.INFO)
logging.info("start parsing settings")
params = parse(args.config)
network_config = params['Network']
logging.info("finish parsing settings")
logging.info(network_config)
print(network_config)
glv.init(network_config, [args.device])
dataset_name = glv.network_config['dataset']
data_path = glv.network_config['data_path']
logging.info("dataset loading...")
# 数据集加载
train_loader_id_list,test_loader_id_list = load_data_all(glv.network_config['dataset'],1)
test_loader_ood_list = load_data_all(glv.network_config['ood_data'],0)
logging.info("dataset loaded")
# 加载现有模型_待完成
if args.checkpoint is not None:
checkpoint_path = args.checkpoint
checkpoint = torch.load(checkpoint_path)
if "Letters" in checkpoint_path:
# 加载网络
if network_config['model'] == 'FSVAE':
net = fsvae.FSVAE(class_num=26)
elif network_config['model'] == 'FSVAE_large':
net = fsvae.FSVAELarge()
else:
raise Exception('not defined model')
net = net.to(init_device)
net.load_state_dict(checkpoint['net'])
R = checkpoint['R']
c = checkpoint['c']
print("n_class = 26")
print('R = ',R)
print('c = ',torch.mean(c))
else:
# 加载网络
if network_config['model'] == 'FSVAE':
net = fsvae.FSVAE(class_num=10)
elif network_config['model'] == 'FSVAE_large':
net = fsvae.FSVAELarge()
else:
raise Exception('not defined model')
net = net.to(init_device)
net.load_state_dict(checkpoint['net'])
R = checkpoint['R']
c = checkpoint['c']
print("n_class = 10")
print('R = ',R)
print('c = ',torch.mean(c))
# 开始测试异常检测性能
print('test start!')
auc_list = []
aupr_list = []
if 'KMNIST' in checkpoint_path:
train_loader_id_list,test_loader_id_list = load_data_all(['KMNIST'],1)
test_loader = test_loader_id_list[0]
dataset_name = 'KMNIST'
elif 'Letters' in checkpoint_path:
train_loader_id_list,test_loader_id_list = load_data_all(['Letters'],1)
test_loader = test_loader_id_list[0]
dataset_name = 'Letters'
elif 'MNIST' in checkpoint_path:
train_loader_id_list,test_loader_id_list = load_data_all(['MNIST'],1)
test_loader = test_loader_id_list[0]
dataset_name = 'MNIST'
elif 'FashionMNIST' in checkpoint_path:
train_loader_id_list,test_loader_id_list = load_data_all(['FashionMNIST'],1)
test_loader = test_loader_id_list[0]
dataset_name = 'FashionMNIST'
for _ood, test_loader_ood in zip(glv.network_config['ood_data'], test_loader_ood_list):
if _ood != dataset_name:
print("ood-data = ", _ood)
anoscores_in = get_anoscore(net, test_loader, c,network_config,R)
anoscores_ood = get_anoscore(net, test_loader_ood, c,network_config,R)
labels_test = [0] * len(anoscores_in) + [1] * len(anoscores_ood)
scores = anoscores_in + anoscores_ood
auc, aupr = auc_and_aupr(labels_test, scores)
auc_list.append(round(auc, 3))
aupr_list.append(round(aupr, 3))
print("auc_list = ", auc_list)
print("aupr_list = ", aupr_list)
sys.exit()
best_loss = 1e8
latent_dim = glv.network_config['latent_dim']
for train_data, test_data, id_data_name,class_num in \
zip(train_loader_id_list,test_loader_id_list,network_config['dataset'],network_config['class_num']):
# 加载网络
if network_config['model'] == 'FSVAE':
net = fsvae.FSVAE(class_num)
elif network_config['model'] == 'FSVAE_large':
net = fsvae.FSVAELarge()
else:
raise Exception('not defined model')
net = net.to(init_device)
optimizer = torch.optim.AdamW(net.parameters(),
lr=glv.network_config['lr'],
betas=(0.9, 0.999),
weight_decay=0.001)
print("id_data = ", id_data_name)
for seed in network_config['seed']:
print("seed = ",seed)
setup_seed(seed)
# 随机初始化R和c
c = torch.randn(size=(class_num, latent_dim), device=device)
net.update_c(c)
R = torch.ones(size=(class_num,), device=device) * 5
for e in range(glv.network_config['epochs']):
write_weight_hist(net, e,writer)
if network_config['scheduled']:
net.update_p(e, glv.network_config['epochs'])
logging.info("update p")
latent_dim = glv.network_config['latent_dim']
if e<network_config['pre_epochs']:
train_loss = train(net, train_data, optimizer, c, R, e,glv,writer,network_config,name=args.name,class_num=class_num)
#test_loss = test(net, test_loader, e,ood=0,c=c,R=R,writer=writer)
#test_loss_ood = test(net,test_loader_ood_list[4],e,ood=1,c=c,R=R,writer=writer)
elif e == network_config['pre_epochs']:
print("epoch reaches the threshold, update c and R !")
# 更新R和c，其中c只更新一次
c = update_c(device, train_data, net, class_num, seed)
net.update_c(c)
R = update_r(device, train_data, net, c, nu,class_num)
print("Update R and C successfully!")
train_loss = train(net, train_data, optimizer, c, R, e, glv,writer,network_config,name=args.name,class_num=class_num)
#test_loss = test(net, test_loader, e, ood=0, c=c, R=R,writer=writer)
#test_loss_ood = test(net, test_loader_ood_list[4], e, ood=1, c=c, R=R,writer=writer)
elif e>network_config['pre_epochs']:
# 仅在外面更新R
#c = update_c(device, train_data, net, class_num, seed)
if e%10 == 0:
R = update_r(device, train_data, net, c, nu,class_num)
train_loss = train(net, train_data, optimizer, c, R, e, glv,writer,network_config,name=args.name,class_num=class_num)
#test_loss = test(net, test_loader, e, ood=0, c=c, R=R)
#test_loss_ood = test(net, test_loader_ood_list[4], e, ood=1, c=c, R=R)
# 保存模型,R和c
if e == glv.network_config['epochs']-1:
R = update_r(device, train_data, net, c, nu,class_num)
net_params = {'net':net.state_dict(),'R':R,'c':c}
torch.save(net_params, f'checkpoint/{args.name}/checkpoint'+'_'+id_data_name+'_'+str(seed)+'.pth')
#if test_loss < best_loss:
# best_loss = test_loss
# torch.save(net.state_dict(), f'checkpoint/{args.name}/best.pth')
# 采样模型
#sample(net, e, batch_size=128)
# 开始测试异常检测性能
print('test start!')
get_auc_and_aupr_list(network_config['ood_data'],test_loader_ood_list,c,id_data_name,net,test_data,network_config,writer,R)