-
Notifications
You must be signed in to change notification settings - Fork 8
/
pretrain.py
162 lines (146 loc) · 7.8 KB
/
pretrain.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
import argparse
import json
import numpy as np
import os
import pickle
import shutil
import random
import torch
import torch.backends.cudnn as cudnn
from sklearn.model_selection import train_test_split
from torch.utils.data import ConcatDataset, DataLoader, Subset
from base_model import BaseModel
from datasets import get_dataset
from utils import seed_worker
parser = argparse.ArgumentParser()
parser.add_argument('device', default=0, type=int, help="GPU id to use")
parser.add_argument('config_path', default=0, type=str, help="config file path")
parser.add_argument('--dataset_name', default='mnist', type=str)
parser.add_argument('--model_name', default='mnist', type=str)
parser.add_argument('--num_cls', default=10, type=int)
parser.add_argument('--input_dim', default=1, type=int)
parser.add_argument('--image_size', default=28, type=int)
parser.add_argument('--hidden_size', default=128, type=int)
parser.add_argument('--seed', default=7, type=int)
parser.add_argument('--batch_size', default=128, type=int)
parser.add_argument('--epochs', default=50, type=int)
parser.add_argument('--early_stop', default=5, type=int, help="patience for early stopping")
parser.add_argument('--lr', default=0.001, type=float)
parser.add_argument('--weight_decay', default=5e-4, type=float)
parser.add_argument('--optimizer', default="adam", type=str)
parser.add_argument('--shadow_num', default=5, type=int)
def main(args):
torch.manual_seed(args.seed)
random.seed(args.seed)
np.random.seed(args.seed)
device = f"cuda:{args.device}"
cudnn.benchmark = True
save_folder = f"results/{args.dataset_name}_{args.model_name}"
print(f"Save Folder: {save_folder}")
trainset = get_dataset(args.dataset_name, train=True)
testset = get_dataset(args.dataset_name, train=False)
if testset is None:
total_dataset = trainset
else:
total_dataset = ConcatDataset([trainset, testset])
total_size = len(total_dataset)
data_path = f"{save_folder}/data_index.pkl"
# split the dataset into victim dataset and shadow dataset, then split each into train, val, test
if not os.path.exists(save_folder):
os.mkdir(save_folder)
victim_list, attack_list = train_test_split(list(range(total_size)), test_size=0.5, random_state=args.seed)
victim_train_list, victim_test_list = train_test_split(victim_list, test_size=0.45, random_state=args.seed)
victim_train_list, victim_dev_list = train_test_split(
victim_train_list, test_size=0.1818, random_state=args.seed)
attack_split_list = []
for i in range(args.shadow_num):
attack_train_list, attack_test_list = train_test_split(
attack_list, test_size=0.45, random_state=args.seed + i)
attack_train_list, attack_dev_list = train_test_split(
attack_train_list, test_size=0.1818, random_state=args.seed + i)
attack_split_list.append([attack_train_list, attack_dev_list, attack_test_list])
with open(data_path, 'wb') as f:
pickle.dump([victim_train_list, victim_dev_list, victim_test_list, attack_split_list], f)
# train and prune the victim model
victim_train_dataset = Subset(total_dataset, victim_train_list)
victim_dev_dataset = Subset(total_dataset, victim_dev_list)
victim_test_dataset = Subset(total_dataset, victim_test_list)
print(f"Total Data Size: {total_size}, "
f"Victim Train Size: {len(victim_train_list)}, "
f"Victim Dev Size: {len(victim_dev_list)}, "
f"Victim Test Size: {len(victim_test_list)}")
victim_train_loader = DataLoader(victim_train_dataset, batch_size=args.batch_size, shuffle=True, num_workers=4,
pin_memory=True, worker_init_fn=seed_worker)
victim_dev_loader = DataLoader(victim_dev_dataset, batch_size=args.batch_size, shuffle=False, num_workers=4,
pin_memory=True, worker_init_fn=seed_worker)
victim_test_loader = DataLoader(victim_test_dataset, batch_size=args.batch_size, shuffle=False, num_workers=4,
pin_memory=True, worker_init_fn=seed_worker)
victim_model_save_folder = save_folder + "/victim_model"
print("Train Victim Model")
if not os.path.exists(victim_model_save_folder):
os.makedirs(victim_model_save_folder)
victim_model = BaseModel(
args.model_name, num_cls=args.num_cls, input_dim=args.input_dim, save_folder=victim_model_save_folder,
device=device, optimizer=args.optimizer, lr=args.lr, weight_decay=args.weight_decay)
best_acc = 0
count = 0
for epoch in range(args.epochs):
train_acc, train_loss = victim_model.train(victim_train_loader, f"Epoch {epoch} Train")
dev_acc, dev_loss = victim_model.test(victim_dev_loader, f"Epoch {epoch} Dev")
test_acc, test_loss = victim_model.test(victim_test_loader, f"Epoch {epoch} Test")
if dev_acc > best_acc:
best_acc = dev_acc
save_path = victim_model.save(epoch, test_acc, test_loss)
best_path = save_path
count = 0
elif args.early_stop > 0:
count += 1
if count > args.early_stop:
print(f"Early Stop at Epoch {epoch}")
break
shutil.copyfile(best_path, f"{victim_model_save_folder}/best.pth")
# Train shadow models
for shadow_ind in range(args.shadow_num):
attack_train_list, attack_dev_list, attack_test_list = attack_split_list[shadow_ind]
attack_train_dataset = Subset(total_dataset, attack_train_list)
attack_dev_dataset = Subset(total_dataset, attack_dev_list)
attack_test_dataset = Subset(total_dataset, attack_test_list)
attack_train_loader = DataLoader(attack_train_dataset, batch_size=args.batch_size, shuffle=True, num_workers=4,
pin_memory=True, worker_init_fn=seed_worker)
attack_dev_loader = DataLoader(attack_dev_dataset, batch_size=args.batch_size, shuffle=False, num_workers=4,
pin_memory=True, worker_init_fn=seed_worker)
attack_test_loader = DataLoader(attack_test_dataset, batch_size=args.batch_size, shuffle=False, num_workers=4,
pin_memory=True, worker_init_fn=seed_worker)
print(f"Train Shadow Model {shadow_ind}")
shadow_model_save_folder = f"{save_folder}/shadow_model_{shadow_ind}"
if not os.path.exists(shadow_model_save_folder):
os.makedirs(shadow_model_save_folder)
shadow_model = BaseModel(
args.model_name, num_cls=args.num_cls, input_dim=args.input_dim, save_folder=shadow_model_save_folder,
device=device, optimizer=args.optimizer, lr=args.lr, weight_decay=args.weight_decay)
best_acc = 0
count = 0
for epoch in range(args.epochs):
train_acc, train_loss = shadow_model.train(attack_train_loader, f"Epoch {epoch} Shadow Train")
dev_acc, dev_loss = shadow_model.test(attack_dev_loader, f"Epoch {epoch} Shadow Dev")
test_acc, test_loss = shadow_model.test(attack_test_loader, f"Epoch {epoch} Shadow Test")
if dev_acc > best_acc:
best_acc = dev_acc
save_path = shadow_model.save(epoch, test_acc, test_loss)
best_path = save_path
count = 0
elif args.early_stop > 0:
count += 1
if count > args.early_stop:
print(f"Early Stop at Epoch {epoch}")
break
shutil.copyfile(best_path, f"{shadow_model_save_folder}/best.pth")
if __name__ == '__main__':
args = parser.parse_args()
with open(args.config_path) as f:
t_args = argparse.Namespace()
t_args.__dict__.update(json.load(f))
args = parser.parse_args(namespace=t_args)
args.prune_epochs = int(args.epochs) // 2
print(args)
main(args)