Train/val/test split in PyTorch - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to split the dataset into training and testing sets using PyTorch.

PyTorch

from torch.utils.data import random_split

dataset_size = len(dataset)
train_size = int(0.8 * dataset_size)
test_size = dataset_size - train_size
train_dataset, test_dataset = random_split(dataset, [[1], test_size])

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

Atrain_size

Btest_size

Cdataset_size

D0.8

Attempts:

3 left

💡 Hint

Common Mistakes

Using the test size instead of train size for the first split.

Passing a float instead of an integer for the split size.

The random_split function requires the sizes of the splits as a list. Here, train_size is the number of samples for training.

2fill in blank

medium

Complete the code to create a validation set from the training dataset using PyTorch.

PyTorch

val_size = int(0.1 * len(train_dataset))
train_size = len(train_dataset) - [1]
train_dataset, val_dataset = random_split(train_dataset, [train_size, val_size])

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

Alen(train_dataset)

Btrain_size

C0.1

Dval_size

Attempts:

3 left

💡 Hint

Common Mistakes

Using the new training size instead of validation size in the subtraction.

Not converting the validation size to an integer.

The validation size is subtracted from the training dataset length to get the new training size. The val_size variable holds the number of validation samples.

3fill in blank

hard

Fix the error in the code to correctly split the dataset into train, validation, and test sets.

PyTorch

train_size = int(0.7 * len(dataset))
val_size = int(0.2 * len(dataset))
test_size = int(0.1 * len(dataset))
train_dataset, val_dataset, test_dataset = random_split(dataset, [train_size, val_size, [1]])

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

Atest_size

Bval_size

Ctrain_size

Dlen(dataset) - (train_size + val_size)

Attempts:

3 left

💡 Hint

Common Mistakes

Using validation size twice in the split sizes.

Calculating test size incorrectly.

The third split size should be the test size to match the intended 70/20/10 split.

4fill in blank

hard

Fill both blanks to create DataLoader objects for training and validation datasets with batch size 32.

PyTorch

from torch.utils.data import DataLoader

train_loader = DataLoader(train_dataset, batch_size=[1], shuffle=[2])
val_loader = DataLoader(val_dataset, batch_size=32, shuffle=False)

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

A32

B16

CTrue

DFalse

Attempts:

3 left

💡 Hint

Common Mistakes

Not shuffling training data.

Using wrong batch size.

The batch size for training is 32, and shuffling should be enabled (True) to mix data during training.

5fill in blank

hard

Fill all three blanks to create a dictionary with dataset sizes for train, validation, and test sets.

PyTorch

dataset_sizes = {
    'train': len([1]),
    'val': len([2]),
    'test': len([3])
}

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

Atrain_dataset

Bval_dataset

Ctest_dataset

Ddataset

Attempts:

3 left

💡 Hint

Common Mistakes

Using the original dataset length instead of split datasets.

Mixing up dataset variables.

The dictionary stores the lengths of the train, validation, and test datasets respectively.