main.py

from __future__ import print_function
import argparse
import torch
import torch.utils.data
from torch import nn, optim
from torch.nn import functional as F
from torchvision import datasets, transforms
from torchvision.utils import save_image
import time
import os
import uuid
from tensorboardX import SummaryWriter
from itertools import chain

parser = argparse.ArgumentParser(description='VAE MNIST Example')
parser.add_argument('--batch-size', type=int, default=128, metavar='N',
                    help='input batch size for training (default: 128)')
parser.add_argument('--epochs', type=int, default=20, metavar='N',
                    help='number of epochs to train (default: 20)')
parser.add_argument('--no-cuda', action='store_true', default=False,
                    help='enables CUDA training')
parser.add_argument('--seed', type=int, default=1, metavar='S',
                    help='random seed (default: 1)')
parser.add_argument('--log-interval', type=int, default=100, metavar='N',
                    help='how many batches to wait before logging training status')
parser.add_argument('--log-folder', type=str, default='./tensorboard',
                    help='tensorboard root folder')
parser.add_argument('--kl-scale', type=float, default=1,
                    help='kl penalty strength')
parser.add_argument('--no-aiqn', action='store_true', default=False,
                    help='enables random normal instead of AIQN')
parser.add_argument('--conditioned', action='store_true', default=False,
                    help='AIQN conditioning')
args = parser.parse_args()
args.cuda = not args.no_cuda and torch.cuda.is_available()

torch.manual_seed(args.seed)

device = torch.device("cuda" if args.cuda else "cpu")

kwargs = {'num_workers': 1, 'pin_memory': True} if args.cuda else {}
train_loader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=True, download=True,
                   transform=transforms.ToTensor()),
    batch_size=args.batch_size, shuffle=True, **kwargs)
test_loader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=False, transform=transforms.ToTensor()),
    batch_size=args.batch_size, shuffle=True, **kwargs)

timestr = time.strftime("%Y%m%d-%H%M%S")
summary_name = f"VAE_{timestr}_{uuid.uuid4()}"
summary_path = os.path.join(args.log_folder, summary_name)


class VAE(nn.Module):
    def __init__(self):
        super(VAE, self).__init__()

        self.fc1 = nn.Linear(784, 400)
        self.fc21 = nn.Linear(400, 20)
        self.fc22 = nn.Linear(400, 20)
        self.fc3 = nn.Linear(20, 400)
        self.fc4 = nn.Linear(400, 784)

    def encode(self, x):
        h1 = F.relu(self.fc1(x))
        return self.fc21(h1), self.fc22(h1)

    def reparameterize(self, mu, logvar):
        return reparameterize(mu, logvar) if self.training else mu

    def decode(self, z):
        h3 = F.relu(self.fc3(z))
        return torch.sigmoid(self.fc4(h3))

    def forward(self, x):
        mu, logvar = self.encode(x.view(-1, 784))
        z = self.reparameterize(mu, logvar)
        return self.decode(z), mu, logvar


class QNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.tau_fc = nn.Linear(20, 256)
        self.classes_fc = nn.Linear(10, 256)
        self.q_net = nn.Sequential(
            nn.ReLU(True),
            nn.Linear(256, 256),
            nn.ReLU(True),
            nn.Linear(256, 20),
        )

    def forward(self, tau, classes):
        net_input = self.tau_fc(tau * 2 - 1)
        if args.conditioned:
            with torch.no_grad():
                onehot = torch.zeros((classes.shape[0], 10), device=device)
                onehot.scatter_(dim=-1, index=classes.view(classes.shape[0], 1), value=1)
                onehot -= 1 / 10
            net_input += self.classes_fc(onehot)
        return self.q_net(net_input)
    

def reparameterize(mu, logvar):
    std = torch.exp(0.5 * logvar)
    eps = torch.randn_like(std)
    return eps.mul(std).add_(mu)


model = VAE().to(device)
q_net = QNet().to(device)

optimizer = optim.Adam(chain(model.parameters(), q_net.parameters()), lr=1e-3, weight_decay=1e-3)


# Reconstruction + KL divergence losses summed over all elements and batch
def loss_function(recon_x, x, mu, logvar):
    BCE = F.binary_cross_entropy(recon_x, x.view(-1, 784), reduction='sum')

    # see Appendix B from VAE paper:
    # Kingma and Welling. Auto-Encoding Variational Bayes. ICLR, 2014
    # https://arxiv.org/abs/1312.6114
    # 0.5 * sum(1 + log(sigma^2) - mu^2 - sigma^2)
    KLD = -0.5 * torch.sum(1 + logvar - mu.pow(2) - logvar.exp())

    return BCE + args.kl_scale * KLD


def l1_quantile_loss(output, target, tau, reduce=True):
    u = target - output
    loss = (tau - (u.detach() <= 0).float()).mul_(u)
    return loss.mean() if reduce else loss


def huber_quantile_loss(output, target, tau, k=0.02, reduce=True):
    u = target - output
    loss = (tau - (u.detach() <= 0).float()).mul_(u.detach().abs().clamp(max=k).div_(k)).mul_(u)
    return loss.mean() if reduce else loss


def train(epoch):
    model.train()
    train_loss = 0
    for batch_idx, (data, classes) in enumerate(train_loader):
        data, classes = data.to(device), classes.to(device)
        optimizer.zero_grad()
        recon_batch, mu, logvar = model(data)
        vae_loss = loss_function(recon_batch, data, mu, logvar)

        tau = torch.rand(mu.shape[0], 20, device=device)
        Q = q_net(tau, classes)
        q_loss = huber_quantile_loss(Q, mu.detach(), tau)

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

        if batch_idx % args.log_interval == 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(epoch):
    model.eval()
    test_loss = 0
    with torch.no_grad():
        for i, (data, _) in enumerate(test_loader):
            data = data.to(device)
            recon_batch, mu, logvar = model(data)
            test_loss += loss_function(recon_batch, data, mu, logvar).item()
            if i == 0:
                n = min(data.size(0), 8)
                comparison = torch.cat([data[:n],
                                        recon_batch.view(args.batch_size, 1, 28, 28)[:n]])
                logger.add_image('reconstruction', comparison.view(-1, 1, 28, 28), epoch)

    test_loss /= len(test_loader.dataset)
    logger.add_scalar('test_loss', test_loss, epoch)


fixed_rand = torch.rand(80 if args.conditioned else 64, 20, device=device)
fixed_randn = torch.randn(80 if args.conditioned else 64, 20, device=device)
fixed_classes = torch.arange(10, device=device, dtype=torch.long)
fixed_classes = fixed_classes.expand(8, 10).transpose(0, 1).reshape(-1)
with SummaryWriter(summary_path) as logger:
    for epoch in range(1, args.epochs + 1):
        train(epoch)
        test(epoch)

        with torch.no_grad():
            z = fixed_randn if args.no_aiqn else q_net(fixed_rand, fixed_classes)
            sample = model.decode(z).cpu()
            logger.add_image('sample', sample.view(-1, 1, 28, 28), epoch)