You are writing a PyTorch training script on a machine with a CUDA-enabled GPU. You have correctly identified the device and moved your model to it. However, when you run your script, it crashes on the forward pass with a RuntimeException, indicating that the model's weights and the input data are on different devices. Review the following code snippet and determine the single line of code that must be added to the ---BLANK --- to fix this specific error. import torch import torch.nn as nn # 1. Setup device and move the model to it device = torch.device("cuda" if torch.cuda.is_available() else "cpu") model = nn.Linear(in_features=64, out_features=10).to(device) optimizer = torch.optim.Adam(model.parameters(), lr=0.001) # 2. Create a batch of input data (defaults to CPU) # This simulates data coming from a DataLoader input_batch = torch.randn(32, 64) # [batch_size, feature_size] # 3. Perform a training step optimizer.zero_grad() # --- BLANK --- # The following line causes the error: # "RuntimeException: Expected all tensors to be on the same device, # but found at least two devices, cuda:0 and cpu!" outputs = model(input_batch) # ... loss calculation and backward pass would follow What line of code should be placed in the --- BLANK --- to resolve this device mismatch?

(D) input_batch = input_batch.to(device)

(C) input_batch = input_batch.requires_grad_(True)

(E) torch.set_default_tensor_type('torch.cuda.FloatTensor')

Figure from the original question paper

(C) It disables the creation of the computational graph, saving memory and speeding up execution.

(A) It prevents parameters from being changed by the optimizer during the forward pass.

(B) It deallocates the model's parameters from the GPU to save memory.

(D) It ensures that the model operates entirely on the CPU, preventing unnecessary GPU transfers.

Figure from the original question paper

(D) Option figure from the original question paper

(A) Option figure from the original question paper

(B) Option figure from the original question paper

(C) Option figure from the original question paper

Consider the following code: import torch import torch.nn as nn model = nn.Sequential( nn.Linear(10, 10), nn.BatchNorm1d(10), nn.ReLU(), nn.Dropout(p=0.5), nn.Linear(10, 1) ) X = torch.randn(4, 10) # Run in training mode model.train() out_train = model(X) # Run in evaluation mode model.eval() out_eval = model(X) print(out_train) print(out_eval) Removing which of the following layers will make out_train and out_eval identical for the same input?

(C) Option figure from the original question paper

(A) Option figure from the original question paper

(B) Option figure from the original question paper

(D) Option figure from the original question paper

(E) Option figure from the original question paper

Simple network with three input nodes (input in R^3) connected to a single output node that outputs the sigmoid of the weighted input, trained with binary cross entropy loss Based on the above data, answer the given subquestions. Figure from the original question paper

(D) Option figure from the original question paper

(A) Option figure from the original question paper

(B) Option figure from the original question paper

(C) Option figure from the original question paper

Neural network with inputs x1, x2, x3: w1 connects x1 to the hidden neuron, w3 connects x3 to the hidden neuron, w2 connects x2 directly to the output neuron, w4 connects the hidden neuron to the output; ReLU activation, biases 0 Select the equation representing the forward pass of the given network:

(B) Option figure from the original question paper

(A) Option figure from the original question paper

(C) Option figure from the original question paper

(D) Option figure from the original question paper

September 2025 term · Introduction to Deep Learning and Generative AI · BSDA2001

Introduction to Deep Learning and Generative AI Quiz 1: 26 October 2025 (September 2025 term)

The IIT Madras BS Introduction to Deep Learning and Generative AI (Deep Learning and GenAI) Quiz 1 paper sat on 26 Oct 2025, in the September 2025 term: 20 questions for 50 marks in 120 minutes. Every question is below with its answer. Take it as a timed mock test to be marked, or read it through first.

Questions: 20
Marks: 50
Duration: 120 min
MCQ: 10
Numerical: 10

Take as mock test Practise with answers

Updated 25 Sept 2026

Official paper: IIT M DIPLOMA AN EXAM QDD2 26 Oct 2025 · No negative marking.

Question 1

+2 marksOne correct option

A
torch.tensor([1, 4, 7])
B
torch.tensor([2, 5, 8])
C
torch.tensor([3, 6, 9])
D
torch.tensor([4, 5, 6])
E
torch.tensor([5])

Show answer

Correct answer

B
torch.tensor([2, 5, 8])

Question 1 on its own page

Question 2

+2 marksOne correct option

You have a tensor x = torch.tensor([1, 2, 3, 4]) with shape (4,). You perform the following operations:

python

1stacked_x = torch.stack([x, x, x], dim=0)
2stacked_y = torch.stack([x, x, x], dim=1)

What are the shapes of stacked_x and stacked_y respectively?

A
(3, 4) and (3, 4)
B
(3, 4) and (4, 3)
C
(4, 3) and (4, 3)
D
(4, 3) and (3, 4)

Show answer

Correct answer

B
(3, 4) and (4, 3)

Question 2 on its own page

Question 3

+2 marksOne correct option

A
(1, 2, 2)
B
(2, 2)
C
(2,)
D
(1, 1, 2, 2)
E
(1, 2, 2, 1)

Show answer

Correct answer

B
(2, 2)

Question 3 on its own page

Question 4

+2 marksOne correct option

You are writing a PyTorch training script on a machine with a CUDA-enabled GPU. You have correctly identified the device and moved your model to it. However, when you run your script, it crashes on the forward pass with a RuntimeException, indicating that the model's weights and the input data are on different devices.

Review the following code snippet and determine the single line of code that must be added to the ---BLANK --- to fix this specific error.

python

1import torch
2import torch.nn as nn
3
4# 1. Setup device and move the model to it
5device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
6model = nn.Linear(in_features=64, out_features=10).to(device)
7optimizer = torch.optim.Adam(model.parameters(), lr=0.001)
8
9# 2. Create a batch of input data (defaults to CPU)
10#    This simulates data coming from a DataLoader
11input_batch = torch.randn(32, 64) # [batch_size, feature_size]
12
13# 3. Perform a training step
14optimizer.zero_grad()
15
16# --- BLANK ---
17
18# The following line causes the error:
19# "RuntimeException: Expected all tensors to be on the same device,
20# but found at least two devices, cuda:0 and cpu!"
21outputs = model(input_batch)
22# ... loss calculation and backward pass would follow

What line of code should be placed in the --- BLANK --- to resolve this device mismatch?

A
model.cuda()
B
input_batch.cuda()
C
input_batch = input_batch.requires_grad_(True)
D
input_batch = input_batch.to(device)
E
torch.set_default_tensor_type('torch.cuda.FloatTensor')

Show answer

Correct answer

D
input_batch = input_batch.to(device)

Question 4 on its own page

Question 5

+2 marksOne correct option

A
It prevents parameters from being changed by the optimizer during the forward pass.
B
It deallocates the model's parameters from the GPU to save memory.
C
It disables the creation of the computational graph, saving memory and speeding up execution.
D
It ensures that the model operates entirely on the CPU, preventing unnecessary GPU transfers.

Show answer

Correct answer

C
It disables the creation of the computational graph, saving memory and speeding up execution.

Question 5 on its own page

Question 6

+2 marksOne correct option

Show answer

Correct answer

Question 6 on its own page

Question 7

+2 marksNumerical answer

Consider the following PyTorch code that defines a small 3-layer fully connected neural network

python

1import torch
2import torch.nn as nn
3
4class SmallNet(nn.Module):
5    def __init__(self):
6        super(SmallNet, self).__init__()
7        self.fc1 = nn.Linear(4, 6)   # input -> hidden1
8        self.fc2 = nn.Linear(6, 3)   # hidden1 -> hidden2
9        self.fc3 = nn.Linear(3, 2)   # hidden2 -> output
10    def forward(self, x):
11        x = torch.relu(self.fc1(x))
12        x = torch.relu(self.fc2(x))
13        x = self.fc3(x)
14        return x

How many trainable parameters (weights + biases) does this network have in total?

Show answer

Correct answer: 59

Question 7 on its own page

Question 8

+3 marksOne correct option

Consider the following code:

python

1import torch
2import torch.nn as nn
3
4model = nn.Sequential(
5    nn.Linear(10, 10),
6    nn.BatchNorm1d(10),
7    nn.ReLU(),
8    nn.Dropout(p=0.5),
9    nn.Linear(10, 1)
10)
11X = torch.randn(4, 10)
12
13# Run in training mode
14model.train()
15out_train = model(X)
16
17# Run in evaluation mode
18model.eval()
19out_eval = model(X)
20
21print(out_train)
22print(out_eval)

Removing which of the following layers will make out_train and out_eval identical for the same input?

Show answer

Correct answer

Question 8 on its own page

Question 9

+3 marksOne correct option

Based on the above data, answer the given subquestions.

If all the weights are initalized to 0 and the input to the network is also a zero vector, then what should be the value of bias so that the output of the network is 0.5 ?

A
0.5
B
ln(5)
C
ln(1)
D
1

Show answer

Correct answer

C
ln(1)

Question 9 on its own page

Question 10

+4 marksOne correct option

Based on the above data, answer the given subquestions.

Show answer

Correct answer

Question 10 on its own page

Question 11

+3 marksOne correct option

Select the equation representing the forward pass of the given network:

Show answer

Correct answer

Question 11 on its own page

Question 12

+6 marksNumerical answer

Show answer

Correct answer: 1

Question 12 on its own page

Question 13

+2 marksNumerical answer

Show answer

Correct answer: 1

Question 13 on its own page

Question 14

+3 marksNumerical answer

Consider the following CNN defined in PyTorch:

python

1import torch
2import torch.nn as nn
3import torch.nn.functional as F
4
5class SimpleCNN(nn.Module):
6    def __init__(self):
7        super(SimpleCNN, self).__init__()
8        # Conv Layer 1: in_channels=1 (grayscale), out_channels=8
9        self.conv1 = nn.Conv2d(in_channels=1, out_channels=8,
10        kernel_size=3, stride=1, padding=1)
11        # Pool Layer 1
12        self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2)
13
14        # Conv Layer 2: in_channels=8, out_channels=16
15        self.conv2 = nn.Conv2d(in_channels=8, out_channels=16,
16        kernel_size=3, stride=1, padding=1)
17        # Pool Layer 2
18        self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2)
19
20        # Fully Connected Layers
21        self.fc1 = nn.Linear(16 * 7 * 7, 128)
22        # Flattened from conv2 output
23        self.fc2 = nn.Linear(128, 10)  # Output for 10 classes
24
25    def forward(self, x):
26        x = self.pool1(F.relu(self.conv1(x)))   # Conv1 + Pool1
27        x = self.pool2(F.relu(self.conv2(x)))   # Conv2 + Pool2
28        x = x.view(-1, 16 * 7 * 7)              # Flatten
29        x = F.relu(self.fc1(x))                 # FC1
30        x = self.fc2(x)                         # FC2
31        return x
32
33# Example input
34model = SimpleCNN()
35print(model)

Based on the above data, answer the given subquestions.

How many trainable parameters are there in the first convolutional layer including bias (conv1)?

Show answer

Correct answer: 80

Question 14 on its own page

Question 15

+1 markNumerical answer

Consider the following CNN defined in PyTorch:

python

1import torch
2import torch.nn as nn
3import torch.nn.functional as F
4
5class SimpleCNN(nn.Module):
6    def __init__(self):
7        super(SimpleCNN, self).__init__()
8        # Conv Layer 1: in_channels=1 (grayscale), out_channels=8
9        self.conv1 = nn.Conv2d(in_channels=1, out_channels=8,
10        kernel_size=3, stride=1, padding=1)
11        # Pool Layer 1
12        self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2)
13
14        # Conv Layer 2: in_channels=8, out_channels=16
15        self.conv2 = nn.Conv2d(in_channels=8, out_channels=16,
16        kernel_size=3, stride=1, padding=1)
17        # Pool Layer 2
18        self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2)
19
20        # Fully Connected Layers
21        self.fc1 = nn.Linear(16 * 7 * 7, 128)
22        # Flattened from conv2 output
23        self.fc2 = nn.Linear(128, 10)  # Output for 10 classes
24
25    def forward(self, x):
26        x = self.pool1(F.relu(self.conv1(x)))   # Conv1 + Pool1
27        x = self.pool2(F.relu(self.conv2(x)))   # Conv2 + Pool2
28        x = x.view(-1, 16 * 7 * 7)              # Flatten
29        x = F.relu(self.fc1(x))                 # FC1
30        x = self.fc2(x)                         # FC2
31        return x
32
33# Example input
34model = SimpleCNN()
35print(model)

Based on the above data, answer the given subquestions.

What is the total number of trainable parameters in the pooling layers(pool1 and pool2)?

Show answer

Correct answer: 0

Question 15 on its own page

Question 16

+1 markNumerical answer

Consider the following CNN defined in PyTorch:

python

1import torch
2import torch.nn as nn
3import torch.nn.functional as F
4
5class SimpleCNN(nn.Module):
6    def __init__(self):
7        super(SimpleCNN, self).__init__()
8        # Conv Layer 1: in_channels=1 (grayscale), out_channels=8
9        self.conv1 = nn.Conv2d(in_channels=1, out_channels=8,
10        kernel_size=3, stride=1, padding=1)
11        # Pool Layer 1
12        self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2)
13
14        # Conv Layer 2: in_channels=8, out_channels=16
15        self.conv2 = nn.Conv2d(in_channels=8, out_channels=16,
16        kernel_size=3, stride=1, padding=1)
17        # Pool Layer 2
18        self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2)
19
20        # Fully Connected Layers
21        self.fc1 = nn.Linear(16 * 7 * 7, 128)
22        # Flattened from conv2 output
23        self.fc2 = nn.Linear(128, 10)  # Output for 10 classes
24
25    def forward(self, x):
26        x = self.pool1(F.relu(self.conv1(x)))   # Conv1 + Pool1
27        x = self.pool2(F.relu(self.conv2(x)))   # Conv2 + Pool2
28        x = x.view(-1, 16 * 7 * 7)              # Flatten
29        x = F.relu(self.fc1(x))                 # FC1
30        x = self.fc2(x)                         # FC2
31        return x
32
33# Example input
34model = SimpleCNN()
35print(model)

Based on the above data, answer the given subquestions.

How many parameters (weights and biases) does the first fully connected layer (fc1) have?

Show answer

Correct answer: 100480

Question 16 on its own page

Question 17

+1 markNumerical answer

Consider the following CNN defined in PyTorch:

python

1import torch
2import torch.nn as nn
3import torch.nn.functional as F
4
5class SimpleCNN(nn.Module):
6    def __init__(self):
7        super(SimpleCNN, self).__init__()
8        # Conv Layer 1: in_channels=1 (grayscale), out_channels=8
9        self.conv1 = nn.Conv2d(in_channels=1, out_channels=8,
10        kernel_size=3, stride=1, padding=1)
11        # Pool Layer 1
12        self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2)
13
14        # Conv Layer 2: in_channels=8, out_channels=16
15        self.conv2 = nn.Conv2d(in_channels=8, out_channels=16,
16        kernel_size=3, stride=1, padding=1)
17        # Pool Layer 2
18        self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2)
19
20        # Fully Connected Layers
21        self.fc1 = nn.Linear(16 * 7 * 7, 128)
22        # Flattened from conv2 output
23        self.fc2 = nn.Linear(128, 10)  # Output for 10 classes
24
25    def forward(self, x):
26        x = self.pool1(F.relu(self.conv1(x)))   # Conv1 + Pool1
27        x = self.pool2(F.relu(self.conv2(x)))   # Conv2 + Pool2
28        x = x.view(-1, 16 * 7 * 7)              # Flatten
29        x = F.relu(self.fc1(x))                 # FC1
30        x = self.fc2(x)                         # FC2
31        return x
32
33# Example input
34model = SimpleCNN()
35print(model)

Based on the above data, answer the given subquestions.

What is the total number of convolution filters used in this CNN?

Show answer

Correct answer: 24

Question 17 on its own page

Question 18

+3 marksNumerical answer

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 1

Question 18 on its own page

Question 19

+3 marksNumerical answer

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 2

Question 19 on its own page

Question 20

+3 marksNumerical answer

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 192

Question 20 on its own page