Indexing and slicing in PyTorch - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is indexing in PyTorch tensors?

Indexing means selecting a single element from a tensor using its position, like picking one item from a list by its number.

beginner

How does slicing work in PyTorch tensors?

Slicing means selecting a range of elements from a tensor, similar to cutting a piece from a loaf of bread by specifying start and end positions.

beginner

What does the colon ':' symbol mean in tensor slicing?

The ':' means 'take all elements' along that dimension, like saying 'give me everything in this row or column'.

intermediate

How do you select the first 3 rows and last 2 columns of a 2D tensor in PyTorch?

Use slicing like tensor[:3, -2:] which means 'rows from start to 3 (not including 3)' and 'last 2 columns'.

intermediate

What happens if you use a negative index in PyTorch tensor indexing?

Negative index counts from the end, so -1 means the last element, -2 means second last, just like counting backwards in a list.

In PyTorch, what does tensor[2] select?

A The second element along the first dimension

B The third element along the first dimension

C All elements except the third

D The last element

Indexing starts at 0, so tensor[2] selects the element at position 2, which is the third element.

What does tensor[:, 1:4] do?

A Selects all rows and columns 1 to 3

B Selects rows 1 to 4 and all columns

C Selects only the first row and columns 1 to 4

D Selects all rows and columns 1 to 4

The ':' means all rows, and '1:4' means columns from 1 up to but not including 4.

How do you select the last element of a tensor?

A tensor[-1]

B tensor[0]

C tensor[1]

D tensor[last]

Negative indexing with -1 selects the last element.

What does tensor[1:5:2] mean?

A Select elements from index 1 to 5, stepping by 2

B Select elements from index 1 to 5, stepping by 1

C Select elements from index 2 to 5, stepping by 2

D Select elements from index 1 to 4, stepping by 2

The slice 1:5:2 means start at 1, go up to but not including 5, stepping by 2.

If tensor is 3D, what does tensor[:, 0, :] select?

A All elements in the first slice of the first dimension

B All elements in the last slice of the third dimension

C All elements in the first slice of the second dimension

D Only the first element of the tensor

The ':' means all elements in that dimension, '0' selects the first element in the second dimension.

Explain how to use indexing and slicing to select parts of a PyTorch tensor.

Think about how you pick items from a list or a grid.

Describe how negative indices work in PyTorch tensor indexing and slicing.

Imagine counting backwards from the end of a line.