First PyTorch computation - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

What is the output of this PyTorch tensor operation?

Consider the following PyTorch code that creates two tensors and adds them. What is the output tensor?

PyTorch

import torch
x = torch.tensor([1, 2, 3])
y = torch.tensor([4, 5, 6])
z = x + y
print(z)

A tensor([5, 7, 9])

B tensor([4, 7, 9])

C tensor([1, 2, 3, 4, 5, 6])

D tensor([3, 3, 3])

💡 Hint

Remember that adding two tensors of the same shape adds their elements one by one.

Adding two tensors element-wise sums each corresponding element. So [1+4, 2+5, 3+6] = [5, 7, 9].

❓ Model Choice

intermediate

Choosing the right PyTorch tensor for GPU computation

You want to perform fast matrix multiplication on a GPU using PyTorch. Which tensor creation method ensures the tensor is on the GPU?

A torch.tensor([[1, 2], [3, 4]], dtype=torch.float64)

B torch.tensor([[1, 2], [3, 4]])

C torch.tensor([[1, 2], [3, 4]], device='cuda')

D torch.tensor([[1, 2], [3, 4]], requires_grad=True)

💡 Hint

To use GPU, the tensor must be created or moved to the CUDA device.

Specifying device='cuda' creates the tensor directly on the GPU, enabling fast GPU computations.

❓ Metrics

advanced

Interpreting training loss and accuracy in PyTorch

During training a classification model in PyTorch, you observe the following after one epoch: training loss = 0.8, training accuracy = 0.65. What does this mean?

A The model is 80% accurate and 65% loss means it is underfitting.

B The model is making correct predictions 65% of the time, and the loss value indicates how far predictions are from true labels.

C Loss and accuracy are unrelated; only accuracy matters for model quality.

D A loss of 0.8 means the model predictions are perfect, and accuracy of 0.65 is low.

💡 Hint

Loss measures prediction error; accuracy measures correct predictions percentage.

Loss quantifies how wrong the model predictions are; accuracy shows the fraction of correct predictions. Here, 65% accuracy means 65% correct predictions, and 0.8 loss indicates moderate error.

🔧 Debug

advanced

Why does this PyTorch code raise an error?

What error does this PyTorch code raise and why? import torch x = torch.tensor([1, 2, 3]) y = torch.tensor([4, 5]) z = x + y

PyTorch

import torch
x = torch.tensor([1, 2, 3])
y = torch.tensor([4, 5])
z = x + y

A RuntimeError: The tensors have different shapes and cannot be broadcast for addition.

B SyntaxError: Missing parentheses in print statement.

C TypeError: Unsupported operand types for +: 'Tensor' and 'Tensor'.

D No error; output is tensor([5, 7, 3]).

💡 Hint

Check if the two tensors have the same shape or compatible shapes for addition.

PyTorch requires tensors to have the same shape or broadcastable shapes for element-wise operations. Here, shapes (3,) and (2,) are incompatible, causing a RuntimeError.

🧠 Conceptual

expert

Understanding autograd in PyTorch

Which statement best describes how PyTorch's autograd system works during backpropagation?

A Autograd requires the user to write explicit derivative functions for each operation.

B Autograd manually updates weights during training without needing gradients.

C Autograd only works with CPU tensors and ignores GPU tensors.

D Autograd records operations on tensors with requires_grad=True to build a computation graph for automatic differentiation.

💡 Hint

Think about how PyTorch tracks operations to compute gradients automatically.

PyTorch's autograd tracks operations on tensors that require gradients, building a graph to compute derivatives automatically during backpropagation.