-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataset.py
58 lines (49 loc) · 2.9 KB
/
dataset.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
import torchvision.transforms as transforms
import torchvision
import torch
#num_training = 55000
#num_validation = 5000
batch_size = 100
#-------------------------------------------------
# Load the MNIST dataset
#-------------------------------------------------
norm_transform = transforms.Compose([transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])
mnist_dataset = torchvision.datasets.MNIST(root='datasets/',
train=True,
transform=norm_transform,
download=True)
test_dataset = torchvision.datasets.MNIST(root='datasets/',
train=False,
transform=norm_transform
)
def init_data_mask_base_expt(n_train, n_valid):
# Prepare the training and validation splits
train_data_mask = list(range(n_train))
val_data_mask = list(range(n_train, n_train + n_valid))
train_dataset = torch.utils.data.Subset(mnist_dataset, train_data_mask)
val_dataset = torch.utils.data.Subset(mnist_dataset, val_data_mask)
# Data loader
train_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=batch_size, shuffle=True)
val_loader = torch.utils.data.DataLoader(dataset=val_dataset, batch_size=batch_size, shuffle=False)
return train_loader, val_loader
def init_data_mask_split_data_expt(n_train1, n_valid1, n_train2, n_valid2):
# Prepare the training and validation splits
train_data_mask1 = list(range(n_train1))
val_data_mask1 = list(range(n_train1, n_train1 + n_valid1))
train_data_mask2 = list(range(n_train1 + n_valid1, n_train1 + n_valid1 + n_train2))
val_data_mask2 = list(range(n_train1 + n_valid1 + n_train2, n_train1 + n_valid1 + n_train2 + n_valid2))
train_dataset1 = torch.utils.data.Subset(mnist_dataset, train_data_mask1)
val_dataset1 = torch.utils.data.Subset(mnist_dataset, val_data_mask1)
train_dataset2 = torch.utils.data.Subset(mnist_dataset, train_data_mask2)
val_dataset2 = torch.utils.data.Subset(mnist_dataset, val_data_mask2)
# Data loader
train_loader1 = torch.utils.data.DataLoader(dataset=train_dataset1, batch_size=batch_size, shuffle=True)
val_loader1 = torch.utils.data.DataLoader(dataset=val_dataset1, batch_size=batch_size, shuffle=False)
train_loader2 = torch.utils.data.DataLoader(dataset=train_dataset2, batch_size=batch_size, shuffle=True)
val_loader2 = torch.utils.data.DataLoader(dataset=val_dataset2, batch_size=batch_size, shuffle=False)
return train_loader1, val_loader1, train_loader2, val_loader2
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
batch_size=batch_size,
shuffle=False)