added SVHN

utils/data.py

@@ -3,7 +3,7 @@
 import torchvision.transforms as transforms
 from torch.utils.data import DataLoader
 from torchvision import datasets
-from torchvision.datasets import CIFAR10, CIFAR100, MNIST, FashionMNIST
+from torchvision.datasets import CIFAR10, CIFAR100, MNIST, FashionMNIST, SVHN
 
 
 def load_dataset(configs):
@@ -12,6 +12,8 @@ def load_dataset(configs):
  return load_mnist(configs)
  elif dataset == "fashionmnist":
  return load_fashionmnist(configs)
+ elif dataset == "svhn":
+ return load_svhn(configs)
  elif dataset == "cifar-10":
  return load_cifar10(configs)
  elif dataset == "cifar-100":
@@ -89,6 +91,42 @@ def load_cifar100(configs):
 
  return {'train': train_loader, 'test': test_loader}
 
+def load_cifar100(configs):
+ # transform for the training data
+ train_transforms = transforms.Compose([
+ transforms.RandomCrop(32),
+ transforms.RandomHorizontalFlip(),
+ transforms.ToTensor(),
+ transforms.Normalize(mean=[0.485, 0.456, 0.406],
+ std=[0.229, 0.224, 0.225]),
+ ])
+
+ test_transforms = transforms.Compose([
+ transforms.ToTensor(),
+ transforms.Normalize(mean=[0.485, 0.456, 0.406],
+ std=[0.229, 0.224, 0.225]),
+ ])
+
+ # load datasets, downloading if needed
+ train_set = SVHN('./data/svhn', split="train", download=True,
+ transform=train_transforms)
+ test_set = SVHN('./data/svhn', split="test", download=True,
+ transform=test_transforms)
+
+ train_loader = torch.utils.data.DataLoader(
+ train_set, batch_size=configs.batch_size, num_workers=0)
+ test_loader = torch.utils.data.DataLoader(
+ test_set, batch_size=configs.batch_size, num_workers=0)
+
+ print('Number of iterations required to get through training data of length {}: {}'.format(
+ len(train_set), len(train_loader)))
+
+ print(train_set.data.shape)
+ print(test_set.data.shape)
+
+ return {'train': train_loader, 'test': test_loader}
+
+
 def load_fashionmnist(configs):
  # transform for the training data
  train_transforms = transforms.Compose([