Learning rate selection in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to set the learning rate for the optimizer.

Computer Vision

optimizer = torch.optim.SGD(model.parameters(), lr=[1])

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

A0

B0.01

C-0.1

D'fast'

Attempts:

3 left

💡 Hint

Common Mistakes

Using zero or negative values for learning rate.

Using strings instead of numbers.

The learning rate must be a positive float. 0.01 is a common starting value.

2fill in blank

medium

Complete the code to reduce the learning rate by a factor of 10.

Computer Vision

new_lr = old_lr [1] 10

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

A*

B+

C/

D-

Attempts:

3 left

💡 Hint

Common Mistakes

Multiplying instead of dividing.

Adding or subtracting instead of dividing.

Dividing the old learning rate by 10 reduces it by a factor of 10.

3fill in blank

hard

Fix the error in the learning rate scheduler code.

Computer Vision

scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=5, gamma=[1])

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

A0.1

B0

C-0.5

D1.5

Attempts:

3 left

💡 Hint

Common Mistakes

Using gamma greater than 1 or negative.

Using zero which stops decay.

The gamma parameter must be a positive number less than 1 to reduce the learning rate.

4fill in blank

hard

Fill both blanks to create a dictionary of learning rates for different layers.

Computer Vision

lr_dict = {'conv': [1], 'fc': [2]

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

A0.001

B0.01

C0.0001

D0.1

Attempts:

3 left

💡 Hint

Common Mistakes

Using the same learning rate for both layers.

Using too large learning rates for conv layers.

Typically, convolutional layers use a smaller learning rate (0.001) and fully connected layers a larger one (0.01).

5fill in blank

hard

Fill all three blanks to create a learning rate scheduler with warmup and decay.

Computer Vision

scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda=lambda epoch: [1] if epoch < 5 else [2] ** (epoch - 5) * [3])

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

Aepoch / 5

B0.9

C1

Depoch

Attempts:

3 left

💡 Hint

Common Mistakes

Using wrong decay base or multiplier.

Confusing warmup formula.

This code linearly increases learning rate during warmup (epoch/5), then decays it by 0.9 power times 1.