forked from multirobustbench/multirobustbench
-
Notifications
You must be signed in to change notification settings - Fork 0
/
adv_train.py
201 lines (172 loc) · 8.54 KB
/
adv_train.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
# script for adversarial training for obtaining scores for plotting ideal curve
import argparse
import torch
import torch.nn as nn
import numpy as np
import models
from torchvision import datasets, transforms
import utils
import attacks
import logging
import os
import time
def lr_schedule(t, epochs, max_lr):
if t / epochs < 0.5:
return max_lr
elif t / epochs < 0.75:
return max_lr / 10.
else:
return max_lr / 100.
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--data_dir', type=str)
parser.add_argument('--metrics', nargs='+')
parser.add_argument('--arch', type=str, help='model architecture', default='resnet18')
parser.add_argument('--normalize', action='store_true', help='whether data is normalized before passing into model or not')
parser.add_argument('--attack', type=str, default='NoAttack')
parser.add_argument('--eps', type=float)
parser.add_argument('--model_dir', type=str, help='path to directory of ckpts', default='adv_train_resnet')
parser.add_argument('--epochs', type=int, default=100)
parser.add_argument('--lr_max', type=float, default=0.1)
parser.add_argument('--resume', type=int, help='epoch to resume training from if corresponding ckpt file exists')
parser.add_argument('--num_iters', type=int, default=20)
parser.add_argument('--chkpt_iters', type=int, default=10)
parser.add_argument('--batch_size', type=int, default=150)
parser.add_argument('--use_syn', action='store_true')
parser.add_argument('--batch_size_syn', type=int, default=350)
parser.add_argument('--syn_data_path', type=str, default='data/cifar10_ddpm.npz')
args = parser.parse_args()
if not os.path.exists(args.model_dir):
os.makedirs(args.model_dir)
if not os.path.exists(args.data_dir):
os.makedirs(args.data_dir)
# load data
mean = [0.4914, 0.4822, 0.4465]
std = [0.2023, 0.1994, 0.2010]
transform = transforms.Compose([transforms.RandomCrop(32, padding=4), transforms.RandomHorizontalFlip(),
transforms.ToTensor()])
train_data = datasets.CIFAR10(root=args.data_dir, train=True, transform=transform, download=True)
num_classes = 10
if args.use_syn:
from data import SynData, combine_dataloaders
syn_data = SynData(args.syn_data_path, transform=transforms.ToTensor())
syn_loader = torch.utils.data.DataLoader(dataset=syn_data, batch_size=args.batch_size_syn, shuffle=True, num_workers=4)
train_loader = torch.utils.data.DataLoader(dataset=train_data, batch_size=args.batch_size, shuffle=True, num_workers=4)
train_loader = combine_dataloaders(train_loader, syn_loader)
test_transform = transforms.ToTensor()
test_data = datasets.CIFAR10(root=args.data_dir, train=False, transform=test_transform, download=True)
test_loader = torch.utils.data.DataLoader(dataset=test_data, batch_size=args.batch_size, shuffle=False, num_workers=4)
# set up model
if args.arch in models.__dict__:
model = models.__dict__[args.arch](num_classes)
if args.normalize:
model = models.apply_normalization(model, mean, std)
model = nn.DataParallel(model).cuda()
else:
raise ValueError('unsupported architecture')
# initialize adversary
if args.attack == 'NoAttack':
atk = attacks.__dict__[args.attack]()
test_atk = attacks.__dict__[args.attack]()
args.eps = 0
else:
if args.attack == 'FastLagrangePerceptualAttack':
atk = attacks.__dict__[args.attack](model, bound=args.eps, lpips_model='alexnet_cifar', num_iterations=args.num_iters)
test_atk = attacks.__dict__['LPIPSAttack'](model, bound=args.eps, lpips_model='alexnet_cifar', num_iterations=args.num_iters)
else:
try:
atk = attacks.__dict__[args.attack](model, dataset_name='cifar',bound=args.eps, num_iterations=args.num_iters)
except:
atk = attacks.__dict__[args.attack](model, bound=args.eps, num_iterations=args.num_iters)
test_atk = atk
# run training
logger = logging.getLogger(__name__)
logging.basicConfig(
format='[%(asctime)s] - %(message)s',
datefmt='%Y/%m/%d %H:%M:%S',
level=logging.DEBUG,
handlers=[
logging.FileHandler(os.path.join(args.model_dir, 'at_{}_{}_output.log'.format(args.attack, args.eps))),
logging.StreamHandler()
])
logger.info(args)
opt = torch.optim.SGD(model.parameters(), lr=args.lr_max, momentum=0.9, weight_decay=5e-4)
best_test_robust_acc = 0
if args.resume:
start_epoch = args.resume
model.load_state_dict(torch.load(os.path.join(args.model_dir, f'{args.arch}_{args.attack}_{args.eps}_model_{start_epoch-1}.pth')))
opt.load_state_dict(torch.load(os.path.join(args.model_dir, f'{args.arch}_{args.attack}_{args.eps}_opt_{start_epoch-1}.pth')))
logger.info(f'Resuming at epoch {start_epoch}')
if os.path.exists(os.path.join(args.model_dir, f'model_best.pth')):
best_test_robust_acc = torch.load(os.path.join(args.model_dir, f'{args.arch}_{args.attack}_{args.eps}_model_best.pth'))['test_robust_acc']
else:
start_epoch = 0
criterion = nn.CrossEntropyLoss()
logger.info('Epoch \t Train Time \t Test Time \t LR \t \t Train Loss \t Train Acc \t Train Robust Loss \t Train Robust Acc \t Test Loss \t Test Acc \t Test Robust Loss \t Test Robust Acc')
for epoch in range(start_epoch, args.epochs):
start_time = time.time()
train_loss = 0
train_acc = 0
train_robust_loss = 0
train_robust_acc = 0
train_n = 0
for i, (X, y) in enumerate(train_loader):
model.eval()
X, y = X.cuda(), y.cuda()
lr = lr_schedule(epoch + (i + 1) / len(train_loader), args.epochs, max_lr=args.lr_max)
opt.param_groups[0].update(lr=lr)
X_adv = atk(X, y)
model.train()
robust_output = model(X_adv)
robust_loss = criterion(robust_output, y)
opt.zero_grad()
robust_loss.backward()
opt.step()
output = model(X)
loss = criterion(output, y)
train_robust_loss += robust_loss.item() * y.size(0)
train_robust_acc += (robust_output.max(1)[1] == y).sum().item()
train_loss += loss.item() * y.size(0)
train_acc += (output.max(1)[1] == y).sum().item()
train_n += y.size(0)
train_time = time.time()
model.eval()
test_loss = 0
test_acc = 0
test_robust_loss = 0
test_robust_acc = 0
test_n = 0
for i, (X, y) in enumerate(test_loader):
X, y = X.cuda(), y.cuda()
# Random initialization
X_adv = test_atk(X, y)
robust_output = model(X_adv)
robust_loss = criterion(robust_output, y)
output = model(X)
loss = criterion(output, y)
test_robust_loss += robust_loss.item() * y.size(0)
test_robust_acc += (robust_output.max(1)[1] == y).sum().item()
test_loss += loss.item() * y.size(0)
test_acc += (output.max(1)[1] == y).sum().item()
test_n += y.size(0)
test_time = time.time()
logger.info('%d \t %.1f \t \t %.1f \t \t %.4f \t %.4f \t %.4f \t %.4f \t \t %.4f \t \t %.4f \t %.4f \t %.4f \t \t %.4f',
epoch, train_time - start_time, test_time - train_time, lr,
train_loss/train_n, train_acc/train_n, train_robust_loss/train_n, train_robust_acc/train_n,
test_loss/test_n, test_acc/test_n, test_robust_loss/test_n, test_robust_acc/test_n)
# save checkpoint
if (epoch+1) % args.chkpt_iters == 0 or epoch+1 == args.epochs:
torch.save(model.state_dict(), os.path.join(args.model_dir, f'{args.arch}_{args.attack}_{args.eps}_model_{epoch}.pth'))
torch.save(opt.state_dict(), os.path.join(args.model_dir, f'{args.arch}_{args.attack}_{args.eps}_opt_{epoch}.pth'))
# save best
if test_robust_acc/test_n > best_test_robust_acc:
torch.save({
'state_dict':model.state_dict(),
'test_robust_acc':test_robust_acc/test_n,
'test_robust_loss':test_robust_loss/test_n,
'test_loss':test_loss/test_n,
'test_acc':test_acc/test_n,
}, os.path.join(args.model_dir, f'{args.arch}_{args.attack}_{args.eps}_model_best.pth'))
best_test_robust_acc = test_robust_acc/test_n
if __name__== '__main__':
main()