Why pre-trained models save time in Computer Vision - Test Your Understanding

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a pre-trained model from torchvision.

Computer Vision

import torchvision.models as models
model = models.[1](pretrained=True)

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

Aload

Btrain

Cresnet18

Dfit

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'train' or 'fit' instead of a model name.

Trying to call 'load' which is not a model constructor.

The function 'resnet18' loads the ResNet-18 pre-trained model from torchvision. 'train', 'load', and 'fit' are not model names.

2fill in blank

medium

Complete the code to freeze all layers of the pre-trained model to save training time.

Computer Vision

for param in model.[1]():
    param.requires_grad = False

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

Aparameters

Bchildren

Clayers

Dmodules

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'children()' which returns submodules, not parameters.

Using 'layers()' which is not a valid method.

The 'parameters()' method returns all parameters of the model. Setting requires_grad to False freezes them.

3fill in blank

hard

Fix the error in the code to replace the last layer for transfer learning.

Computer Vision

import torch.nn as nn
model.fc = nn.[1](512, 10)

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

AConv2d

BDropout

CReLU

DLinear

Attempts:

3 left

💡 Hint

Common Mistakes

Using Conv2d which is for convolutional layers.

Using activation layers like ReLU or Dropout instead of Linear.

The last layer of ResNet is a fully connected (Linear) layer. We replace it with a Linear layer with output size 10.

4fill in blank

hard

Fill both blanks to set the model to evaluation mode and disable gradient calculation.

Computer Vision

model.[1]()
with torch.[2]():
    outputs = model(inputs)

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

Aeval

Btrain

Cno_grad

Denable_grad

Attempts:

3 left

💡 Hint

Common Mistakes

Using model.train() instead of model.eval().

Using torch.enable_grad() which enables gradients.

Calling model.eval() sets the model to evaluation mode. Using torch.no_grad() disables gradient calculation to save time.

5fill in blank

hard

Fill all three blanks to create a dictionary of layer names and their requires_grad status.

Computer Vision

grad_status = {name: param.[1] for name, param in model.[2]() if name.[3]('fc') == False}

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

Arequires_grad

Bnamed_parameters

Cendswith

Dstartswith

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'parameters()' instead of 'named_parameters()' which lacks names.

Using 'endswith' instead of 'startswith' to filter layer names.

Checking 'requires_grad' incorrectly.

We check requires_grad for each parameter from named_parameters(). We exclude layers starting with 'fc' to keep them trainable.