Why doesnt reduce='mean' work for official VAE implementation in Pytorch?

In the official VAE implementation which is given below, reduction='sum' is used in the BCE loss and if someone uses ‘mean’ instead, in the backward pass, the autograd will fail with the error :

RuntimeError: Function ‘AddmmBackward’ returned nan values in its 2th output.

This happens, while all the weight norms are positive and none of them have inf values in them. the input to the layer which fails in the backward pass with the mentioned error, also has no inf values.

I believe we should be able to use ‘mean’ as well, so what is stopping us from using that?
I remember the initial version of the VAE, used the default behavior (i.e.reduction='mean')

BCE = F.binary_cross_entropy(recon_x, x.view(-1, 784))

and everything was fine. but after 0.4 it seems, this was changed to sum .

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

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
kwargs = {'num_workers': 1, 'pin_memory': True} 

train_loader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=True, download=True,
    batch_size=batch_size, shuffle=True, **kwargs)
test_loader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=False, transform=transforms.ToTensor()),
    batch_size=batch_size, shuffle=True, **kwargs)

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):
        std = torch.exp(0.5*logvar)
        eps = torch.randn_like(std)
        return mu + eps*std

    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

model = VAE().to(device)
optimizer = optim.Adam(model.parameters(), lr=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 + KLD

def train(epoch):
    train_loss = 0
    for batch_idx, (data, _) in enumerate(train_loader):
        data = data.to(device)
        recon_batch, mu, logvar = model(data)
        loss = loss_function(recon_batch, data, mu, logvar)
        train_loss += loss.item()
        if batch_idx % 1000 == 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):
    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(batch_size, 1, 28, 28)[:n]])
                         'vae_results/reconstruction_' + str(epoch) + '.png', nrow=n)

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

for epoch in range(1, epochs + 1):
    with torch.no_grad():
        sample = torch.randn(64, 20).to(device)
        sample = model.decode(sample).cpu()
        save_image(sample.view(64, 1, 28, 28),
                    'vae_results/sample_' + str(epoch) + '.png')

Any help is greatly appreciated