-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathload_data.py
156 lines (126 loc) · 6.33 KB
/
load_data.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
import torchvision.transforms as transforms
from dataset import *
from torch.autograd import Variable
import torch.nn.functional as F
from petrel_client.utils.data import DataLoader
def load_cifar10_data():
transform = transforms.Compose([transforms.ToTensor()])
cifar10_train_ds = CIFAR10(train=True, transform=transform)
cifar10_test_ds = CIFAR10(train=False, transform=transform)
X_train, y_train = cifar10_train_ds.data, cifar10_train_ds.targets
X_test, y_test = cifar10_test_ds.data, cifar10_test_ds.targets
return (X_train, y_train, X_test, y_test)
def load_mnist_data():
transform = transforms.Compose([transforms.ToTensor()])
mnist_train_ds = MNIST_truncated(train=True, transform=transform)
mnist_test_ds = MNIST_truncated(train=False, transform=transform)
X_train, y_train = mnist_train_ds.data, mnist_train_ds.target
X_test, y_test = mnist_test_ds.data, mnist_test_ds.target
X_train = X_train.data.numpy()
y_train = y_train.data.numpy()
X_test = X_test.data.numpy()
y_test = y_test.data.numpy()
return (X_train, y_train, X_test, y_test)
def load_svhn_data():
transform = transforms.Compose([transforms.ToTensor()])
svhn_train_ds = SVHN_custom( train=True, transform=transform)
svhn_test_ds = SVHN_custom( train=False, transform=transform)
X_train, y_train = svhn_train_ds.data, svhn_train_ds.target
X_test, y_test = svhn_test_ds.data, svhn_test_ds.target
return (X_train, y_train, X_test, y_test)
def load_fmnist_data():
transform = transforms.Compose([transforms.ToTensor()])
fmnist_train_ds = FashionMNIST_truncated(train=True, transform=transform)
fmnist_test_ds = FashionMNIST_truncated(train=False, transform=transform)
X_train, y_train = fmnist_train_ds.data, fmnist_train_ds.target
X_test, y_test = fmnist_test_ds.data, fmnist_test_ds.target
X_train = X_train.data.numpy()
y_train = y_train.data.numpy()
X_test = X_test.data.numpy()
y_test = y_test.data.numpy()
return (X_train, y_train, X_test, y_test)
def load_cifar100_data():
transform = transforms.Compose([transforms.ToTensor()])
cifar100_train_ds = CIFAR100(train=True, transform=transform)
cifar100_test_ds = CIFAR100(train=False, transform=transform)
X_train, y_train = cifar100_train_ds.data, cifar100_train_ds.targets
X_test, y_test = cifar100_test_ds.data, cifar100_test_ds.targets
return (X_train, y_train, X_test, y_test)
def get_dataloader(args, dataidxs=None,identity=None, noise_level=0):
if args.dataset in ('cifar10', 'cifar100'):
if args.dataset == 'cifar10':
dl_obj = CIFAR10
normalize = transforms.Normalize((0.4914, 0.4822, 0.4465),
(0.2470, 0.2435, 0.2615))
transform_train = transforms.Compose([
transforms.ToTensor(),
transforms.Lambda(lambda x: F.pad(
Variable(x.unsqueeze(0), requires_grad=False),
(4, 4, 4, 4), mode='reflect').data.squeeze()),
transforms.ToPILImage(),
transforms.ColorJitter(brightness=noise_level),
transforms.RandomCrop(32),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize
])
transform_test = transforms.Compose([
transforms.ToTensor(),
normalize])
elif args.dataset == 'cifar100':
dl_obj = CIFAR100
normalize = transforms.Normalize(mean=[0.5070751592371323, 0.48654887331495095, 0.4409178433670343],
std=[0.2673342858792401, 0.2564384629170883, 0.27615047132568404])
transform_train = transforms.Compose([
# transforms.ToPILImage(),
transforms.RandomCrop(32, padding=4),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(15),
transforms.ToTensor(),
normalize
])
# data prep for test set
transform_test = transforms.Compose([
transforms.ToTensor(),
normalize])
train_ds = dl_obj(dataidxs=dataidxs, train=True, transform=transform_train)
test_ds = dl_obj(train=False, transform=transform_test)
elif args.dataset=="mnist":
transform_train = transforms.Compose([
transforms.ToTensor(),
])
transform_test = transforms.Compose([
transforms.ToTensor(),
])
train_ds = MNIST_truncated(dataidxs=dataidxs, train=True, transform=transform_train)
test_ds = MNIST_truncated(train=False, transform=transform_test)
elif args.dataset=="fmnist":
#transforms.Normalize((0.5,), (0.5,)),
transform_train = transforms.Compose([
transforms.ToTensor(),
])
transform_test = transforms.Compose([
transforms.ToTensor(),
])
train_ds = FashionMNIST_truncated(dataidxs=dataidxs, train=True, transform=transform_train)
test_ds = FashionMNIST_truncated(train=False, transform=transform_test)
elif args.dataset=="SVHN":
transform_train = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(
(0.4376821, 0.4437697, 0.47280442), (0.19803012, 0.20101562, 0.19703614),
),
])
transform_test = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(
(0.4376821, 0.4437697, 0.47280442), (0.19803012, 0.20101562, 0.19703614),
),
])
train_ds = SVHN_custom(dataidxs=dataidxs, train=True, transform=transform_train)
test_ds = SVHN_custom(train=False, transform=transform_test)
train_dl = DataLoader(dataset=train_ds, batch_size=args.train_batchsize, drop_last=True, shuffle=True,
prefetch_factor=4, persistent_workers=True,num_workers=4)
test_dl = DataLoader(dataset=test_ds, batch_size=args.test_batchsize, shuffle=False, drop_last=False,
prefetch_factor=4, persistent_workers=True,num_workers=4)
return train_dl, test_dl, train_ds, test_ds