PyTorch ecosystem overview - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding PyTorch's Core Components

Which of the following is NOT a core component of the PyTorch ecosystem?

A NumPy for tensor operations within PyTorch

B TensorBoard for visualization and debugging

C TorchScript for model serialization and optimization

D TorchVision for computer vision datasets and models

💡 Hint

Think about which library is external and not part of PyTorch itself.

NumPy is a separate library for numerical computing in Python. While PyTorch tensors can interoperate with NumPy arrays, NumPy is not a core component of the PyTorch ecosystem.

❓ Predict Output

intermediate

Output of PyTorch Model Training Loop

What will be the printed output after running this PyTorch training loop snippet?

PyTorch

import torch
import torch.nn as nn

model = nn.Linear(2, 1)
criterion = nn.MSELoss()
optimizer = torch.optim.SGD(model.parameters(), lr=0.1)

inputs = torch.tensor([[1.0, 2.0]])
targets = torch.tensor([[1.0]])

optimizer.zero_grad()
outputs = model(inputs)
loss = criterion(outputs, targets)
loss.backward()
optimizer.step()

print(round(loss.item(), 4))

A 0.75

B 0.5

C 0.25

D 0.0

💡 Hint

The loss is the mean squared error between model output and target before the optimizer step.

The initial weights are random, so the output will not match the target exactly. The MSE loss will be a positive value around 0.5 for this simple input and target.

❓ Model Choice

advanced

Choosing the Right PyTorch Ecosystem Library

You want to train a natural language processing model using pre-built datasets and tokenizers. Which PyTorch ecosystem library should you use?

A TorchServe

B TorchAudio

C TorchText

D TorchVision

💡 Hint

Think about which library focuses on text data.

TorchText provides tools for text processing, datasets, and tokenization for NLP tasks within the PyTorch ecosystem.

❓ Hyperparameter

advanced

Effect of Learning Rate in PyTorch Optimizer

If you increase the learning rate in a PyTorch optimizer too much, what is the most likely effect on training?

A Training will always reach a better minimum

B Training will ignore the learning rate and behave the same

C Training will converge faster and more smoothly

D Training may become unstable and loss may oscillate or diverge

💡 Hint

Think about what happens if steps are too large when trying to find a minimum.

A very high learning rate can cause the model to overshoot minima, causing unstable training and possibly divergence.

❓ Metrics

expert

Calculating Accuracy from PyTorch Model Outputs

Given a batch of model outputs as logits and true labels, which code snippet correctly computes the classification accuracy in PyTorch?

PyTorch

outputs = torch.tensor([[2.0, 1.0, 0.1], [0.5, 2.5, 0.3], [1.2, 0.7, 2.1]])
labels = torch.tensor([0, 1, 2])

A

preds = outputs.argmax(dim=1)
accuracy = (preds == labels).float().mean().item()

B

preds = outputs.max(dim=0)[1]
accuracy = (preds == labels).float().mean().item()

C

preds = outputs.argmax(dim=1)
accuracy = (preds == labels).sum().item()

D

preds = outputs.max(dim=1)[0]
accuracy = (preds == labels).float().mean().item()

💡 Hint

Check how to get predicted class indices and how to compute mean accuracy.

Using argmax over dimension 1 gives predicted class indices per sample. Comparing with labels and taking the mean of correct predictions gives accuracy.