Generator and discriminator in PyTorch - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to define a simple generator class in PyTorch.

PyTorch

import torch.nn as nn

class Generator(nn.Module):
    def __init__(self):
        super(Generator, self).__init__()
        self.main = nn.Sequential(
            nn.Linear(100, 256),
            nn.ReLU(True),
            nn.Linear(256, 784),
            nn.Tanh()
        )

    def forward(self, input):
        return self.main([1])

Drag options to blanks, or click blank then click option'

Ainput

Bself

Coutput

Dx

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'self' instead of 'input' inside the forward method.

Using a variable not defined in the method like 'x' or 'output'.

The forward method takes the input tensor and passes it through the model layers. So, the input variable should be used.

2fill in blank

medium

Complete the code to define a simple discriminator class in PyTorch.

PyTorch

import torch.nn as nn

class Discriminator(nn.Module):
    def __init__(self):
        super(Discriminator, self).__init__()
        self.main = nn.Sequential(
            nn.Linear(784, 256),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Linear(256, 1),
            nn.Sigmoid()
        )

    def forward(self, input):
        return self.main([1])

Drag options to blanks, or click blank then click option'

Ainput

Bself

Coutput

Dx

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'self' or undefined variables inside the forward method.

Confusing the input variable name.

The forward method receives the input tensor and passes it through the model layers. So, 'input' should be used.

3fill in blank

hard

Fix the error in the training loop where the generator's output is passed to the discriminator.

PyTorch

for epoch in range(num_epochs):
    for real_data in dataloader:
        noise = torch.randn(batch_size, 100)
        fake_data = generator([1])
        output = discriminator(fake_data)
        # rest of training code...

Drag options to blanks, or click blank then click option'

Aoutput

Breal_data

Cnoise

Dgenerator

Attempts:

3 left

💡 Hint

Common Mistakes

Passing real_data instead of noise to the generator.

Passing the generator object itself.

The generator expects noise as input to produce fake data. Passing 'noise' is correct.

4fill in blank

hard

Fill both blanks to complete the code for calculating the discriminator loss using binary cross entropy.

PyTorch

criterion = nn.BCELoss()
real_labels = torch.ones(batch_size, 1)
fake_labels = torch.zeros(batch_size, 1)

output_real = discriminator(real_data)
loss_real = criterion(output_real, [1])

output_fake = discriminator(fake_data.detach())
loss_fake = criterion(output_fake, [2])

d_loss = loss_real + loss_fake

Drag options to blanks, or click blank then click option'

Areal_labels

Bfake_labels

Coutput_real

Doutput_fake

Attempts:

3 left

💡 Hint

Common Mistakes

Swapping real_labels and fake_labels.

Using output tensors as labels.

For real data, the target labels are ones (real_labels). For fake data, the target labels are zeros (fake_labels).

5fill in blank

hard

Fill all three blanks to complete the generator training step with loss calculation and backpropagation.

PyTorch

optimizer_g.zero_grad()
noise = torch.randn(batch_size, 100)
fake_data = generator([1])
output = discriminator(fake_data)
loss_g = criterion(output, [2])
loss_g.[3]()
optimizer_g.step()

Drag options to blanks, or click blank then click option'

Anoise

Breal_labels

Cbackward

Dfake_labels

Attempts:

3 left

💡 Hint

Common Mistakes

Using fake_labels instead of real_labels for generator loss.

Forgetting to call backward() before optimizer step.

Passing wrong input to generator.

The generator input is noise. The generator tries to fool the discriminator, so target labels are real_labels (ones). Then call backward() to compute gradients.