def __init__(self, shape, nhid=16):
super(VAEMLPDecoder, self).__init__()
flattened_size = torch.Size(shape).numel()
self.shape = shape
self.decode = nn.Sequential(
MLP([nhid, 64, 128, 256, flattened_size], last_activation=False),
nn.Sigmoid(),
)
self.invTrans = transforms.Compose(
[transforms.Normalize((0.1307, ), (0.3081, ))])
def get_data():
# 将像素点转换到[-1, 1]之间，使得输入变成一个比较对称分布，训练容易收敛
data_tf = transforms.Compose(
[transforms.Totensor(),
transforms.Normalize([0.5, 0.5])])
train_dataset = datasets.MNIST(root='./data',
train=True,
transform=data_tf,
download=True)
train_loder = DataLoader(train_dataset,
shuffle=True,
batch_size=batch_size,
drop_last=True)
return train_loder