Diffusion model concept in Prompt Engineering / GenAI - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to add noise to the data in a diffusion model step.

Prompt Engineering / GenAI

noisy_data = data + [1] * noise

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

A0.1

B1.0

C0.5

D2.0

Attempts:

3 left

💡 Hint

Common Mistakes

Using noise scale too small like 0.1 which barely changes data

Using noise scale too large like 2.0 which overwhelms data

In diffusion models, noise is added gradually. A common noise scale is 0.5 for a moderate noise level.

2fill in blank

medium

Complete the code to reverse the diffusion process by denoising the noisy data.

Prompt Engineering / GenAI

denoised_data = model.predict([1])

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

Alabels

Bdata

Cnoise

Dnoisy_data

Attempts:

3 left

💡 Hint

Common Mistakes

Passing original data instead of noisy data

Passing noise directly to the model

The model predicts the clean data from the noisy input, so it takes noisy_data as input.

3fill in blank

hard

Fix the error in the training loss calculation for the diffusion model.

Prompt Engineering / GenAI

loss = loss_fn([1], predicted_noise)

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

Anoise

Bdenoised_data

Cnoisy_data

Ddata

Attempts:

3 left

💡 Hint

Common Mistakes

Using original data instead of noise in loss

Using noisy data instead of noise in loss

The loss compares the predicted noise to the actual noise added, so the first argument is noise.

4fill in blank

hard

Fill both blanks to create a dictionary comprehension that maps each step to its noise level if the noise level is less than 0.5.

Prompt Engineering / GenAI

noise_levels = {step: [1] for step, noise in steps.items() if noise [2] 0.5}

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

Anoise

B>

C<

Dstep

Attempts:

3 left

💡 Hint

Common Mistakes

Using step as value instead of noise

Using > instead of < in condition

The dictionary maps step to noise only if noise is less than 0.5, so we use noise and <.

5fill in blank

hard

Fill all three blanks to create a dictionary comprehension that maps each step's name in uppercase to its noise value if noise is greater than 0.3.

Prompt Engineering / GenAI

filtered_noise = { [1]: [2] for step, [3] in steps.items() if noise > 0.3 }

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

Astep.upper()

Bnoise

Dstep

Attempts:

3 left

💡 Hint

Common Mistakes

Using step instead of step.upper() as key

Using step instead of noise as value

Swapping variable names in the loop

The key is step.upper(), the value is noise, and the loop unpacks step and noise.