Test on CINIC-10 dataset

[ashentears]

CINIC-10 dataset: https://datashare.is.ed.ac.uk/handle/10283/3192

There are certain changes on main.py file. The first change is on data loading directory to CINIC-10 dataset folder

cinic_dir = 'CINIC-10'
trainset = torchvision.datasets.ImageFolder(cinic_dir + '/train', transform=transform_train)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=128, shuffle=True, num_workers=5)

testset = torchvision.datasets.ImageFolder(cinic_dir + '/test', transform=transform_test)
testloader = torch.utils.data.DataLoader(testset, batch_size=80, shuffle=False, num_workers=5)

classes = ('airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck')

Next, to adjust with CINIC-10 dataset, transform normalize value has been changed as

cinic_mean = [0.47889522, 0.47227842, 0.43047404]
cinic_std = [0.24205776, 0.23828046, 0.25874835]

transform_train = transforms.Compose([
    transforms.RandomCrop(32, padding=1),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor(),
    transforms.Normalize(cinic_mean, cinic_std),
])

transform_test = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize(cinic_mean, cinic_std),
])

Lastly, the learning rate which is used in this training is lr = 0.01