Gradient access (.grad) in PyTorch - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to access the gradient of tensor x after backward pass.

PyTorch

import torch
x = torch.tensor(2.0, requires_grad=True)
y = x * x
loss = y
loss.backward()
print(x[1])

Drag options to blanks, or click blank then click option'

A.grad

B.data

C.item()

D.requires_grad

Attempts:

3 left

💡 Hint

Common Mistakes

Using .data instead of .grad to access gradients.

Trying to access gradient before calling backward().

The .grad attribute holds the gradient of the tensor after backward() is called.

2fill in blank

medium

Complete the code to zero out the gradients of tensor x.

PyTorch

import torch
x = torch.tensor(3.0, requires_grad=True)
y = x * 3
loss = y
loss.backward()
x.grad[1]()

Drag options to blanks, or click blank then click option'

Aclear

Bzero_

Creset

Dzero

Attempts:

3 left

💡 Hint

Common Mistakes

Using zero() instead of zero_(), which does not modify in-place.

Trying to assign None to x.grad instead of zeroing.

The zero_() method resets the gradient tensor to zero in-place.

3fill in blank

hard

Fix the error in accessing the gradient of tensor z after backward pass.

PyTorch

import torch
z = torch.tensor(4.0, requires_grad=True)
out = z ** 3
out.backward()
print(z[1])

Drag options to blanks, or click blank then click option'

A.data

B.grad_fn

C.grad.data

D.grad

Attempts:

3 left

💡 Hint

Common Mistakes

Using .grad_fn which gives the function that created the tensor, not the gradient.

Trying to access .data which is the raw tensor data, not the gradient.

The correct way to access the gradient is z.grad after backward().

4fill in blank

hard

Fill both blanks to compute the gradient of y = x^2 and print it.

PyTorch

import torch
x = torch.tensor(5.0, requires_grad=True)
y = x[1]2
loss = y
loss.backward()
print(x[2])

Drag options to blanks, or click blank then click option'

A**

B.grad

C.data

D*

Attempts:

3 left

💡 Hint

Common Mistakes

Using * instead of ** for power operation.

Accessing .data instead of .grad for gradients.

Use ** for power and .grad to access the gradient after backward().

5fill in blank

hard

Fill all three blanks to compute gradients for z = a * b and print gradients of a and b.

PyTorch

import torch
a = torch.tensor(2.0, requires_grad=True)
b = torch.tensor(3.0, requires_grad=True)
z = a[1]b
z.backward()
print(a[2])
print(b[3])

Drag options to blanks, or click blank then click option'

A*

B.grad

D+

Attempts:

3 left

💡 Hint

Common Mistakes

Using + instead of * for multiplication.

Trying to access .data instead of .grad for gradients.

Use * for multiplication and .grad to access gradients of both tensors.