vae-ddf/datasets.py at main · miro-code/vae-ddf · GitHub

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
from torchvision.datasets import CIFAR10, CIFAR100, MNIST
from torch.utils.data import DataLoader
from torchvision import transforms
import pytorch_lightning as pl
from torch.utils.data import random_split
import torch

#I CALCULATED NORMALIZATION VALUES FOR ALL THE DATASETS
CIFAR10_TRANSFORMS = transforms.Compose([
        transforms.ToTensor(),
        transforms.Normalize((0.49139967861519745, 0.4821584083946076, 0.44653091444546616), (0.2470322324632823, 0.24348512800005553, 0.2615878417279641))
])

MNIST_TRANSFORMS = transforms.Compose([
    transforms.ToTensor(),
    #increase the image size to 32x32
    transforms.Pad(2),
    transforms.Normalize((0.1307,), (0.3081,))
])

CIFAR100_TRANSFORMS = transforms.Compose([
        transforms.ToTensor(),
        transforms.Normalize((0.5070751592371322, 0.4865488733149497, 0.44091784336703466), (0.26733428587924063, 0.25643846291708833, 0.27615047132568393))
])


class MNISTDataModule(pl.LightningDataModule):
    def __init__(self, data_dir: str = "data", batch_size: int = 512):
        super().__init__()
        self.data_dir = data_dir
        self.batch_size = batch_size

    def setup(self, stage: str):
        self.mnist_test = MNIST(self.data_dir, train=False, transform=MNIST_TRANSFORMS, download=True)
        mnist_full = MNIST(self.data_dir, train=True, transform=MNIST_TRANSFORMS, download=True)
        self.mnist_train, self.mnist_val = random_split(
            mnist_full, [55000, 5000], generator=torch.Generator().manual_seed(42)
        )

    def train_dataloader(self):
        return DataLoader(self.mnist_train, batch_size=self.batch_size, shuffle=True)

    def val_dataloader(self):
        return DataLoader(self.mnist_val, batch_size=self.batch_size)

    def test_dataloader(self):
        return DataLoader(self.mnist_test, batch_size=self.batch_size)

class MNISTDebugDataModule(MNISTDataModule):
    def __init__(self, data_dir: str = "data", batch_size: int = 512):
        super().__init__(data_dir, batch_size)

    def setup(self, stage: str):
        mnist_full = MNIST(self.data_dir, train=True, transform=MNIST_TRANSFORMS, download=True)
        self.mnist_train, self.mnist_val, _ = random_split(
            mnist_full, [400, 100, 59500], generator=torch.Generator().manual_seed(42)
        )
        self.mnist_test = MNIST(self.data_dir, train=False, transform=MNIST_TRANSFORMS, download=True)
        self.mnist_test, _ = random_split(self.mnist_test, [100, 9900], generator=torch.Generator().manual_seed(42))

class CIFAR10DataModule(pl.LightningDataModule):
    def __init__(self, data_dir: str = "data", batch_size: int = 512):
        super().__init__()
        self.data_dir = data_dir
        self.batch_size = batch_size

    def setup(self, stage: str):
        self.cifar10_test = CIFAR10(self.data_dir, train=False, transform=CIFAR10_TRANSFORMS, download=True)
        cifar10_full = CIFAR10(self.data_dir, train=True, transform=CIFAR10_TRANSFORMS, download=True)
        self.cifar10_train, self.cifar10_val = random_split(
            cifar10_full, [45000, 5000], generator=torch.Generator().manual_seed(42)
        )

    def train_dataloader(self):
        return DataLoader(self.cifar10_train, batch_size=self.batch_size, shuffle=True)

    def val_dataloader(self):
        return DataLoader(self.cifar10_val, batch_size=self.batch_size)

    def test_dataloader(self):
        return DataLoader(self.cifar10_test, batch_size=self.batch_size)

class CIFAR10DebugDataModule(CIFAR10DataModule):
    def __init__(self, data_dir: str = "data", batch_size: int = 512):
        super().__init__(data_dir, batch_size)

    def setup(self, stage: str):
        cifar10_full = CIFAR10(self.data_dir, train=True, transform=CIFAR10_TRANSFORMS, download=True)
        self.cifar10_train, self.cifar10_val, _ = random_split(
            cifar10_full, [400, 100, 59500], generator=torch.Generator().manual_seed(42)
        )
        self.cifar10_test = CIFAR10(self.data_dir, train=False, transform=CIFAR10_TRANSFORMS, download=True)
        self.cifar10_test, _ = random_split(self.cifar10_test, [100, 9900], generator=torch.Generator().manual_seed(42))

class CIFAR100DataModule(pl.LightningDataModule):
    def __init__(self, data_dir: str = "data", batch_size: int = 512):
        super().__init__()
        self.data_dir = data_dir
        self.batch_size = batch_size

    def setup(self, stage: str):
        self.cifar100_test = CIFAR100(self.data_dir, train=False, transform=CIFAR100_TRANSFORMS, download=True)
        cifar100_full = CIFAR100(self.data_dir, train=True, transform=CIFAR100_TRANSFORMS, download=True)
        self.cifar100_train, self.cifar100_val = random_split(
            cifar100_full, [45000, 5000], generator=torch.Generator().manual_seed(42)
        )

    def train_dataloader(self):
        return DataLoader(self.cifar100_train, batch_size=self.batch_size, shuffle=True)

    def val_dataloader(self):
        return DataLoader(self.cifar100_val, batch_size=self.batch_size)

    def test_dataloader(self):
        return DataLoader(self.cifar100_test, batch_size=self.batch_size)

class CIFAR100DebugDataModule(CIFAR10DataModule):
    def __init__(self, data_dir: str = "data", batch_size: int = 512):
        super().__init__(data_dir, batch_size)

    def setup(self, stage: str):
        cifar100_full = CIFAR10(self.data_dir, train=True, transform=CIFAR10_TRANSFORMS, download=True)
        self.cifar100_train, self.cifar100_val, _ = random_split(
            cifar100_full, [400, 100, 59500], generator=torch.Generator().manual_seed(42)
        )
        self.cifar100_test = CIFAR10(self.data_dir, train=False, transform=CIFAR10_TRANSFORMS, download=True)
        self.cifar100_test, _ = random_split(self.cifar100_test, [100, 9900], generator=torch.Generator().manual_seed(42))

DATAMODULES = {
    "mnist": MNISTDataModule,
    "cifar10": CIFAR10DataModule,
    "cifar100": CIFAR100DataModule,
}
DEBUGDATAMODULES = {
    "mnist": MNISTDebugDataModule,
    "cifar10": CIFAR10DebugDataModule,
    "cifar100": CIFAR100DebugDataModule,
}