main_cnn.py

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from matplotlib import pyplot as plt

# Ref: https://github.com/lyeoni/pytorch-mnist-VAE/blob/master/pytorch-mnist-VAE.ipynb
# Ref: https://towardsdatascience.com/building-a-convolutional-vae-in-pytorch-a0f54c947f71

class VAE(nn.Module):
    def __init__(self, imgChannels=1, featureDim=32*20*20, zDim=256):
        super(VAE, self).__init__()

        # Initializing the 2 convolutional layers and 2 full-connected layers for the encoder
        self.encConv1 = nn.Conv2d(imgChannels, 16, 5)
        self.encConv2 = nn.Conv2d(16, 32, 5)
        self.encFC1 = nn.Linear(featureDim, zDim)
        self.encFC2 = nn.Linear(featureDim, zDim)

        # Initializing the fully-connected layer and 2 convolutional layers for decoder
        self.decFC1 = nn.Linear(zDim, featureDim)
        self.decConv1 = nn.ConvTranspose2d(32, 16, 5)
        self.decConv2 = nn.ConvTranspose2d(16, imgChannels, 5)


    def encoder(self, x):
        # Input is fed into 2 convolutional layers sequentially
        # The output feature map are fed into 2 fully-connected layers to predict mean (mu) and variance (logVar)
        # Mu and logVar are used for generating middle representation z and KL divergence loss
        x = F.relu(self.encConv1(x))
        x = F.relu(self.encConv2(x))
        x = x.view(-1, 32 * 20 * 20)
        mu = self.encFC1(x)
        logVar = self.encFC2(x)
        return mu, logVar

    def sampling(self, mu, log_var):
        # Reparameterization takes in the input mu and logVar and sample the mu + std * eps
        std = torch.exp(log_var / 2)
        eps = torch.randn_like(std)
        return mu + std * eps

    def decoder(self, z):
        # z is fed back into a fully-connected layers and then into two transpose convolutional layers
        # The generated output is the same size of the original input
        x = F.relu(self.decFC1(z))
        x = x.view(-1, 32, 20, 20)
        x = F.relu(self.decConv1(x))
        x = torch.sigmoid(self.decConv2(x))
        return x

    def forward(self, x):
        # The entire pipeline of the VAE: encoder -> reparameterization -> decoder
        # output, mu, and logVar are returned for loss computation
        mu, log_var = self.encoder(x)
        z = self.sampling(mu, log_var)
        out = self.decoder(z)
        return out, mu, log_var

def loss_function(recon_x, x, mu, log_var):
    BCE = F.binary_cross_entropy(recon_x, x, reduction='sum')
    KLD = -0.5 * torch.sum(1 + log_var - mu.pow(2) - log_var.exp())
    return BCE + KLD


def train(epoch, vae, train_loader, optimizer):
    vae.train()
    train_loss = 0
    for batch_idx, (data, _) in enumerate(train_loader):
        data = data.cuda()
        optimizer.zero_grad()

        recon_batch, mu, log_var = vae(data)
        loss = loss_function(recon_batch, data, mu, log_var)

        loss.backward()
        train_loss += loss.item()
        optimizer.step()

        if batch_idx % 100 == 0:
            print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
                epoch, batch_idx * len(data), len(train_loader.dataset),
                       100. * batch_idx / len(train_loader), loss.item() / len(data)))
    print('====> Epoch: {} Average loss: {:.4f}'.format(epoch, train_loss / len(train_loader.dataset)))


def test(vae, test_loader):
    vae.eval()
    test_loss = 0
    with torch.no_grad():
        for data, _ in test_loader:
            data = data.cuda()
            recon, mu, log_var = vae(data)

            # sum up batch loss
            test_loss += loss_function(recon, data, mu, log_var).item()

    test_loss /= len(test_loader.dataset)
    print('====> Test set loss: {:.4f}'.format(test_loss))

def main():

    print('this is hw3')
    device = 'cuda' if torch.cuda.is_available() else 'cpu'
    print(device)

    bs = 100
    # MNIST Dataset
    train_dataset = datasets.MNIST(root='./mnist_data/', train=True, transform=transforms.ToTensor(), download=True)
    test_dataset = datasets.MNIST(root='./mnist_data/', train=False, transform=transforms.ToTensor(), download=False)

    # Data Loader (Input Pipeline)
    train_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=bs, shuffle=True)
    test_loader = torch.utils.data.DataLoader(dataset=test_dataset, batch_size=bs, shuffle=False)

    vae = VAE()
    if torch.cuda.is_available():
        vae.cuda()

    optimizer = optim.Adam(vae.parameters())

    # Training:
    for epoch in range(1, 2):
        train(epoch, vae, train_loader, optimizer)
        test(vae, test_loader)

    images, labels = iter(test_loader).next()
    sample_pic = images[0]
    print(sample_pic.shape)
    plt.imshow(sample_pic[0].reshape(28, 28), cmap="gray")
    plt.show()

    with torch.no_grad():
        sample_pic = sample_pic.cuda()
        sample_pic = sample_pic.unsqueeze(0)
        print(sample_pic.shape)
        result = vae.forward(sample_pic)
        print('Got result')
        result = result[0].cpu()
        plt.imshow(result.reshape(28, 28), cmap="gray")
        plt.show()

if __name__ == '__main__':
    main()