__getitem__ and __len__ in PyTorch - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is the purpose of the __getitem__ method in a PyTorch dataset?

The __getitem__ method lets you get one data sample by its index. It returns the input and label for that index, so PyTorch can use it during training or testing.

beginner

Why do we need to define __len__ in a PyTorch dataset?

The __len__ method tells PyTorch how many samples are in the dataset. This helps PyTorch know when to stop an epoch and how to shuffle data.

intermediate

How does __getitem__ relate to the DataLoader in PyTorch?

The DataLoader calls __getitem__ to get each sample when making batches. It uses the index to fetch data one by one or in parallel.

intermediate

What happens if __len__ returns the wrong number in a PyTorch dataset?

If __len__ is wrong, PyTorch might miss data or try to get samples that don't exist, causing errors or incomplete training.

beginner

Show a simple example of a PyTorch dataset class with <code>__getitem__</code> and <code>__len__</code> methods.

```python
import torch
from torch.utils.data import Dataset

class SimpleDataset(Dataset):
    def __init__(self, data, labels):
        self.data = data
        self.labels = labels

    def __len__(self):
        return len(self.data)

    def __getitem__(self, idx):
        x = self.data[idx]
        y = self.labels[idx]
        return x, y

# Example usage:
data = torch.tensor([[1,2],[3,4],[5,6]])
labels = torch.tensor([0,1,0])
dataset = SimpleDataset(data, labels)
print(len(dataset))  # Output: 3
print(dataset[1])    # Output: (tensor([3, 4]), tensor(1))
```

What does the __getitem__ method return in a PyTorch dataset?

A The model's prediction

B A single data sample and its label

C The batch size

D The total number of samples

__getitem__ returns one data sample and its label for the given index.

Why is __len__ important in a PyTorch dataset?

A It trains the model

B It loads the data from disk

C It normalizes the data

D It tells how many samples are in the dataset

__len__ tells PyTorch the size of the dataset.

If __len__ returns 100, what does that mean?

A The model has 100 layers

B The batch size is 100

C There are 100 samples in the dataset

D The dataset is empty

The number returned by __len__ is the number of samples.

What happens if you try to access an index outside the range in __getitem__?

A You get an error

B You get a random sample

C PyTorch automatically fixes it

D Nothing happens

Accessing an invalid index causes an error.

Which PyTorch class do you usually inherit to create a custom dataset with __getitem__ and __len__?

A torch.utils.data.Dataset

B torch.nn.Module

C torch.optim.Optimizer

D torch.Tensor

Custom datasets inherit from torch.utils.data.Dataset.

Explain in your own words why __getitem__ and __len__ are essential for PyTorch datasets.

Think about how PyTorch gets data samples and knows when to stop.

Describe what could go wrong if __len__ returns a smaller or larger number than the actual dataset size.

Consider what happens when PyTorch tries to access data beyond the dataset.

Practice

1. What does the __len__ method do in a PyTorch dataset class?

easy

A. It saves the dataset to disk.

B. It returns one data item by index.

C. It returns the total number of items in the dataset.

D. It trains the model on the dataset.

Solution

1. Step 1: Understand the role of __len__ in datasets

   The __len__ method tells PyTorch how many items are in the dataset so it knows the dataset size.

2. Step 2: Differentiate from __getitem__

   Unlike __getitem__ which returns one item, __len__ returns the total count of items.

3. Final Answer:

   It returns the total number of items in the dataset. -> Option C

4. Quick Check:

   __len__ = total items [OK]

Hint: Remember: length means total count, not single item [OK]

Common Mistakes:

• Confusing __len__ with __getitem__
• Thinking __len__ returns data samples
• Assuming __len__ trains the model

2. Which of the following is the correct way to define the __getitem__ method in a PyTorch dataset class?

easy

A. def __getitem__(self, index, batch): return self.data[index:batch]

B. def __getitem__(self, index): return self.data[index]

C. def __getitem__(self): return self.data

D. def __getitem__(index): return self.data[index]

Solution

1. Step 1: Check method signature requirements

   The __getitem__ method must accept self and an index parameter to return one item.

2. Step 2: Validate return statement

   Returning self.data[index] correctly fetches the item at the given index.

3. Final Answer:

   def __getitem__(self, index): return self.data[index] -> Option B

4. Quick Check:

   __getitem__ needs self and index [OK]

Hint: Always include self and index parameters in __getitem__ [OK]

Common Mistakes:

• Missing self parameter
• Adding extra parameters not supported
• Returning whole data instead of one item

3. Given this dataset class, what will print(len(dataset)) output?

class MyDataset:
    def __init__(self):
        self.data = [10, 20, 30, 40]
    def __len__(self):
        return len(self.data)
    def __getitem__(self, idx):
        return self.data[idx]

dataset = MyDataset()

medium

A. 4

B. 3

C. Error

D. 40

Solution

1. Step 1: Check the data list length

   The list self.data contains 4 elements: 10, 20, 30, 40.

2. Step 2: Understand __len__ return value

   The __len__ method returns len(self.data), which is 4.

3. Final Answer:

   4 -> Option A

4. Quick Check:

   len(data list) = 4 [OK]

Hint: len(dataset) calls __len__, returns data list length [OK]

Common Mistakes:

• Confusing length with last element value
• Expecting __getitem__ output instead
• Forgetting to call len() properly

4. Identify the error in this PyTorch dataset class:

class Dataset:
    def __init__(self, data):
        self.data = data
    def __len__(self):
        return self.data
    def __getitem__(self, idx):
        return self.data[idx]

medium

A. __len__ should return an integer, not the data list

B. __getitem__ is missing the self parameter

C. __init__ should not accept data as argument

D. __getitem__ should return the whole data list

Solution

1. Step 1: Check the return type of __len__

   The __len__ method returns self.data, which is a list, not an integer.

2. Step 2: Understand expected __len__ behavior

   __len__ must return the number of items (an integer), so it should return len(self.data).

3. Final Answer:

   __len__ should return an integer, not the data list -> Option A

4. Quick Check:

   __len__ returns int length, not list [OK]

Hint: __len__ must return count, not the data itself [OK]

Common Mistakes:

• Returning data list instead of its length
• Forgetting to use len() inside __len__
• Misdefining __getitem__ parameters

5. You want to create a PyTorch dataset that returns the square of numbers from 0 to 9. Which implementation correctly uses __getitem__ and __len__?

hard

A. class SquareDataset:
def __len__(self):
return 9
def __getitem__(self, idx):
return idx ** 2

B. class SquareDataset:
def __len__(self):
return 10
def __getitem__(self, idx):
return idx + idx

C. class SquareDataset:
def __len__(self):
return 10
def __getitem__(self):
return idx * idx

D. class SquareDataset:
def __len__(self):
return 10
def __getitem__(self, idx):
return idx * idx

Solution

1. Step 1: Check the length method

   Since numbers 0 to 9 means 10 items, __len__ should return 10.

2. Step 2: Verify __getitem__ returns square

   The method must accept idx and return idx * idx to give squares.

3. Final Answer:

   class SquareDataset: def __len__(self): return 10 def __getitem__(self, idx): return idx * idx -> Option D

4. Quick Check:

   Length 10, item = idx squared [OK]

Hint: Length matches count; __getitem__ returns square [OK]

Common Mistakes:

• Returning wrong length (9 instead of 10)
• Missing idx parameter in __getitem__
• Returning sum instead of square