GAN training loop in PyTorch - Model Metrics & Evaluation

Metrics & Evaluation - GAN training loop

Which metric matters for GAN training loop and WHY

In GANs, we want the generator to create data that looks real, and the discriminator to tell real from fake. The key metrics are discriminator loss and generator loss. These losses show how well each part is learning. Lower loss means better performance. We also look at Inception Score or FID to measure how real the generated images look overall. These metrics help us know if the GAN is improving or stuck.

Confusion matrix or equivalent visualization

Discriminator predictions:

           | Real Data | Fake Data
-----------|-----------|----------
Pred Real  |    TP     |    FP    
Pred Fake  |    FN     |    TN    

Where:
- TP (True Positive): Discriminator correctly says real data is real
- FP (False Positive): Discriminator wrongly says fake data is real
- FN (False Negative): Discriminator wrongly says real data is fake
- TN (True Negative): Discriminator correctly says fake data is fake

Losses are computed from these predictions to update both networks.
    

Precision vs Recall tradeoff in GANs with examples

GAN training balances two goals:

• Discriminator precision: How often it correctly identifies fake images. High precision means few fake images are mistaken as real.
• Generator recall: How well the generator fools the discriminator. High recall means many generated images look real enough to fool the discriminator.

If the discriminator is too strong (high precision), the generator struggles and produces poor images (low recall). If the generator is too strong (high recall), the discriminator fails to learn (low precision). Good GAN training finds a balance where both improve together.

What "good" vs "bad" metric values look like for GAN training

Good:

• Discriminator loss and generator loss both decrease steadily and stabilize.
• Generated images look realistic and diverse.
• Inception Score or FID improves over time (higher IS, lower FID).

Bad:

• Discriminator loss quickly goes to zero while generator loss stays high (discriminator too strong).
• Generator loss goes to zero but discriminator loss stays high (generator collapses to few outputs).
• Generated images are blurry, repetitive, or obviously fake.

Common pitfalls in GAN metrics

• Mode collapse: Generator produces limited variety of outputs. Losses may look good but diversity is poor.
• Overfitting discriminator: Discriminator becomes too perfect, causing generator to fail learning.
• Ignoring qualitative checks: Only looking at loss numbers without checking generated samples can mislead.
• Data leakage: Using test data in training can inflate metrics falsely.

Self-check question

Your GAN training shows discriminator loss near zero but generator loss remains high. Generated images look poor and repetitive. Is this good?

Answer: No. This means the discriminator is too strong and easily spots fakes. The generator is not learning well and may be stuck producing limited outputs. You should adjust training to balance both networks better.

Key Result

In GAN training, balanced discriminator and generator losses plus improved image quality metrics indicate good progress.

Practice

1. In a GAN training loop, what is the main role of the discriminator?

easy

A. To calculate the learning rate

B. To generate new fake data samples

C. To distinguish between real and fake data

D. To optimize the generator's weights directly

Solution

1. Step 1: Understand the discriminator's purpose

   The discriminator's job is to tell apart real data from fake data created by the generator.

2. Step 2: Differentiate roles in GAN

   The generator creates fake data, while the discriminator judges it. It does not generate data or adjust learning rates.

3. Final Answer:

   To distinguish between real and fake data -> Option C

4. Quick Check:

   Discriminator = spot fake data [OK]

Hint: Discriminator judges data authenticity, generator creates data [OK]

Common Mistakes:

• Confusing discriminator with generator
• Thinking discriminator generates data
• Assuming discriminator sets learning rate

2. Which of the following is the correct way to zero the gradients of both generator and discriminator optimizers in PyTorch during a GAN training loop?

easy

A. optimizer_G.clear(); optimizer_D.clear()

B. optimizer_G.zero_grad(); optimizer_D.zero_grad()

C. optimizer_G.reset_grad(); optimizer_D.reset_grad()

D. optimizer_G.zero(); optimizer_D.zero()

Solution

1. Step 1: Recall PyTorch optimizer method

   In PyTorch, the method to clear gradients before backpropagation is zero_grad().

2. Step 2: Check each option's validity

   Only optimizer_G.zero_grad(); optimizer_D.zero_grad() uses zero_grad(), which is the correct method. Other options use non-existent methods.

3. Final Answer:

   optimizer_G.zero_grad(); optimizer_D.zero_grad() -> Option B

4. Quick Check:

   Clear gradients = zero_grad() [OK]

Hint: Use zero_grad() to clear gradients before backward pass [OK]

Common Mistakes:

• Using non-existent optimizer methods
• Forgetting to zero gradients before backward
• Confusing zero_grad() with optimizer step()

3. Given the following PyTorch GAN training snippet, what will be the printed output for print(loss_D.item(), loss_G.item()) after one iteration?

real_data = torch.ones(4)
fake_data = torch.zeros(4)
criterion = torch.nn.BCELoss()

# Discriminator output
output_real = torch.tensor([0.9, 0.8, 0.95, 0.85])
output_fake = torch.tensor([0.1, 0.2, 0.05, 0.15])

loss_real = criterion(output_real, real_data)
loss_fake = criterion(output_fake, fake_data)
loss_D = (loss_real + loss_fake) / 2

# Generator tries to fool discriminator
output_fake_for_G = torch.tensor([0.4, 0.3, 0.5, 0.6])
loss_G = criterion(output_fake_for_G, real_data)

print(loss_D.item(), loss_G.item())

medium

A. Approximately 0.693 and 0.105

B. Approximately 0.0 and 1.0

C. Approximately 0.5 and 0.5

D. Approximately 0.136 and 0.831

Solution

1. Step 1: Calculate discriminator loss

   Loss_real = BCELoss([0.9,0.8,0.95,0.85], [1,1,1,1]) ≈ 0.136; Loss_fake = BCELoss([0.1,0.2,0.05,0.15], [0,0,0,0]) ≈ 0.136; Average loss_D ≈ 0.136.

2. Step 2: Calculate generator loss

   Generator tries to fool discriminator, so loss_G = BCELoss([0.4,0.3,0.5,0.6], [1,1,1,1]) ≈ 0.831.

3. Final Answer:

   Approximately 0.136 and 0.831 -> Option D

4. Quick Check:

   Discriminator loss < Generator loss [OK]

Hint: Discriminator loss low near 0, generator loss higher when fooled less [OK]

Common Mistakes:

• Mixing up real and fake labels in loss
• Confusing which loss belongs to generator
• Ignoring averaging of discriminator losses

4. Identify the error in this GAN training loop snippet and select the fix:

for real_data in dataloader:
    optimizer_D.zero_grad()
    output_real = discriminator(real_data)
    loss_real = criterion(output_real, torch.ones_like(output_real))

    fake_data = generator(torch.randn(batch_size, noise_dim))
    output_fake = discriminator(fake_data.detach())
    loss_fake = criterion(output_fake, torch.zeros_like(output_fake))

    loss_D = loss_real + loss_fake
    loss_D.backward()
    optimizer_D.step()

    optimizer_G.zero_grad()
    output_fake_for_G = discriminator(fake_data)
    loss_G = criterion(output_fake_for_G, torch.ones_like(output_fake_for_G))
    loss_G.backward()
    optimizer_D.step()

medium

A. Replace last optimizer_D.step() with optimizer_G.step()

B. Add optimizer_G.zero_grad() before loss_G.backward()

C. Detach fake_data before passing to generator

D. Use torch.zeros_like for output_fake_for_G labels

Solution

1. Step 1: Analyze optimizer steps

   After computing generator loss and backward, the generator's optimizer should update weights, not discriminator's.

2. Step 2: Identify incorrect optimizer call

   The code calls optimizer_D.step() twice; the second should be optimizer_G.step() to update generator.

3. Final Answer:

   Replace last optimizer_D.step() with optimizer_G.step() -> Option A

4. Quick Check:

   Generator update uses optimizer_G.step() [OK]

Hint: Generator updates need optimizer_G.step(), not optimizer_D.step() [OK]

Common Mistakes:

• Calling discriminator optimizer twice
• Forgetting to update generator weights
• Misplacing zero_grad calls

5. You want to improve GAN training stability by modifying the training loop. Which change below best helps prevent the discriminator from overpowering the generator?

hard

A. Train the discriminator less frequently than the generator

B. Increase the learning rate of the discriminator only

C. Remove the generator's loss calculation

D. Use the same optimizer instance for both models

Solution

1. Step 1: Understand discriminator overpowering

   If the discriminator learns too fast, the generator receives poor feedback and training stalls.

2. Step 2: Evaluate options to balance training

   Training discriminator less often than generator slows its learning, giving generator a chance to improve.

3. Final Answer:

   Train the discriminator less frequently than the generator -> Option A

4. Quick Check:

   Balance training frequency to stabilize GAN [OK]

Hint: Train discriminator less often to balance GAN training [OK]

Common Mistakes:

• Increasing discriminator learning rate worsens imbalance
• Removing generator loss stops generator learning
• Sharing optimizer causes parameter update conflicts