Sequential model shortcut in PyTorch - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of a simple PyTorch Sequential model

What is the output shape of the tensor after passing through this PyTorch Sequential model?

PyTorch

import torch
import torch.nn as nn

model = nn.Sequential(
    nn.Linear(10, 5),
    nn.ReLU(),
    nn.Linear(5, 2)
)

x = torch.randn(3, 10)
out = model(x)
print(out.shape)

A torch.Size([5, 2])

B torch.Size([3, 2])

C torch.Size([10, 2])

D torch.Size([3, 5])

💡 Hint

Remember the input batch size and the output size of the last Linear layer.

The input tensor has shape (3, 10). The first Linear layer outputs (3, 5), then ReLU keeps the shape, and the last Linear layer outputs (3, 2). So the final output shape is (3, 2).

❓ Model Choice

intermediate

Choosing the correct Sequential model for image flattening

Which PyTorch Sequential model correctly flattens a 28x28 grayscale image and outputs 10 class scores?

A nn.Sequential(nn.Flatten(), nn.Linear(28*28, 10))

B nn.Sequential(nn.Flatten(), nn.Linear(10, 28*28))

C nn.Sequential(nn.Conv2d(1, 10, 3), nn.Flatten())

D nn.Sequential(nn.Linear(28*28, 10), nn.Flatten())

💡 Hint

Flatten should come before Linear to convert 2D image to 1D vector.

Option A first flattens the 28x28 image into a vector of length 784, then applies a Linear layer to output 10 scores. Option A tries to flatten after Linear which is invalid. Option A uses Conv2d which changes shape differently. Option A outputs wrong size.

❓ Hyperparameter

advanced

Effect of changing hidden layer size in Sequential model

You have a Sequential model: nn.Sequential(nn.Linear(20, 50), nn.ReLU(), nn.Linear(50, 5)). What happens if you change the hidden layer size from 50 to 10?

A The input size must be changed to 10.

B The model will have more parameters and train slower.

C The output size changes from 5 to 10.

D The model has fewer parameters and may train faster but could underfit.

💡 Hint

Hidden layer size controls model capacity and parameter count.

Reducing hidden layer size from 50 to 10 reduces parameters, which can speed training but may reduce model capacity, risking underfitting. Output size remains 5, input size stays 20.

🔧 Debug

advanced

Identify the error in this Sequential model code

What error will this PyTorch code raise?

PyTorch

import torch.nn as nn
model = nn.Sequential(
    nn.Linear(10, 20),
    nn.ReLU(),
    nn.Linear(15, 5)
)

A TypeError because ReLU is not callable

B SyntaxError due to missing comma

C RuntimeError due to mismatched input size in second Linear layer

D No error, code runs fine

💡 Hint

Check the output size of first Linear and input size of second Linear.

The first Linear outputs size 20, but the second Linear expects input size 15, causing a runtime size mismatch error.

❓ Metrics

expert

Interpreting training loss and accuracy from Sequential model training

After training a PyTorch Sequential model for classification, you observe the following metrics: training loss decreases steadily, but training accuracy stays around 50%. What is the most likely explanation?

A The model is not learning useful features; possibly wrong labels or model too simple.

B The optimizer is not updating weights due to zero learning rate.

C The loss function is incorrect and does not reflect accuracy.

D The batch size is too large causing unstable training.

💡 Hint

Loss decreasing but accuracy stuck means model predictions improve but not enough to classify correctly.

If loss decreases but accuracy stays low, the model may be learning something but not enough to classify well. This can happen if labels are wrong or model capacity is insufficient.