Model parameters inspection in PyTorch - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to print the names of all parameters in a PyTorch model.

PyTorch

for name, param in model.[1]():
    print(name)

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

Anamed_parameters

Bparameters

Cstate_dict

Dchildren

Attempts:

3 left

💡 Hint

Common Mistakes

Using parameters() which returns only parameter values without names.

Using state_dict() which returns a dictionary, not an iterator.

The named_parameters() method returns both the parameter names and their values, which allows us to print the names.

2fill in blank

medium

Complete the code to count how many parameters in the model require gradients.

PyTorch

count = sum(p.requires_grad for p in model.[1]())
print(count)

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

Anamed_parameters

Bparameters

Cchildren

Dmodules

Attempts:

3 left

💡 Hint

Common Mistakes

Using named_parameters() when only parameters are needed.

Using children() which returns submodules, not parameters.

The parameters() method returns all parameters, and we check their requires_grad attribute to count those that require gradients.

3fill in blank

hard

Fix the error in the code to print the shape of each parameter tensor.

PyTorch

for name, param in model.named_parameters():
    print(name, param.[1])

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

Ashape

Bdim

Csize

Dlength

Attempts:

3 left

💡 Hint

Common Mistakes

Using size without parentheses, which prints the method object instead of the shape.

Using dim which returns the number of dimensions, not the shape.

Using length which is not a tensor property.

In PyTorch, shape is a property/attribute that returns the shape of a tensor as a torch.Size tuple. size is a method that requires parentheses (size()); without them, it refers to the method object itself.

4fill in blank

hard

Fill the blank to create a dictionary of parameter names and their flattened sizes.

PyTorch

param_sizes = {name: param.[1]() for name, param in model.named_parameters()}

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

Asize

Bdim

Cshape

Dnumel

Attempts:

3 left

💡 Hint

Common Mistakes

Using size() or shape which return the shape tuple, not the flattened count.

Using dim which returns number of dimensions, not size.

numel() returns the total number of elements (flattened size) in the tensor.

5fill in blank

hard

Fill all three blanks to filter parameters that require gradients and create a dictionary of their names and shapes.

PyTorch

grad_params = {name: param.[1]() for name, param in model.named_parameters() if param.[2] and param.[3]

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

Asize

Brequires_grad

Cis_leaf

Dgrad

Attempts:

3 left

💡 Hint

Common Mistakes

Using grad which is None before backward pass.

Not filtering by is_leaf which may include intermediate tensors.

We get the shape with size(). We filter parameters that require gradients (requires_grad) and are leaf tensors (is_leaf) to ensure they are trainable parameters.