Why custom data pipelines handle real data in PyTorch - Test Your Understanding

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the PyTorch data loader module.

PyTorch

from torch.utils.data import [1]

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

ADataLoader

BDataset

CTensor

Dnn

Attempts:

3 left

💡 Hint

Common Mistakes

Importing Dataset instead of DataLoader

Importing nn which is for neural networks

The DataLoader class is used to load data in batches for training in PyTorch.

2fill in blank

medium

Complete the code to define a custom dataset class inheriting from PyTorch's Dataset.

PyTorch

class CustomDataset([1]):
    def __init__(self, data):
        self.data = data
    def __len__(self):
        return len(self.data)
    def __getitem__(self, idx):
        return self.data[idx]

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

ADataLoader

BTensor

CDataset

DModule

Attempts:

3 left

💡 Hint

Common Mistakes

Inheriting from DataLoader instead of Dataset

Using Module which is for models

The custom dataset must inherit from Dataset to work with PyTorch data loaders.

3fill in blank

hard

Fix the error in the data loading loop to get batches from the data loader.

PyTorch

for [1] in data_loader:
    inputs, labels = batch
    print(inputs.shape, labels.shape)

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

Abatch

Bdata

Cinputs

Dlabels

Attempts:

3 left

💡 Hint

Common Mistakes

Using inputs or labels as loop variable

Using data which is not defined here

The loop variable should be batch to unpack inputs and labels correctly.

4fill in blank

hard

Fill both blanks to create a data loader with batch size 32 and shuffling enabled.

PyTorch

data_loader = DataLoader(dataset, batch_size=[1], shuffle=[2])

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

A32

BTrue

CFalse

D64

Attempts:

3 left

💡 Hint

Common Mistakes

Using batch size 64 when 32 is asked

Setting shuffle to False which disables random order

Batch size should be 32 and shuffling should be True to randomize data order.

5fill in blank

hard

Fill all three blanks to transform raw data into tensors and normalize it in the custom dataset.

PyTorch

import torch
from torchvision import transforms

class CustomDataset(Dataset):
    def __init__(self, data):
        self.transform = transforms.Compose([
            transforms.ToTensor(),
            transforms.Normalize(mean=[1], std=[2])
        ])
        self.data = data
    def __getitem__(self, idx):
        sample = self.data[idx]
        sample = self.transform(sample)
        return sample
    def __len__(self):
        return len(self.data)

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

A[0.5, 0.5, 0.5]

B[0.229, 0.224, 0.225]

C[0.485, 0.456, 0.406]

Attempts:

3 left

💡 Hint

Common Mistakes

Swapping mean and std values

Using incorrect normalization values

The mean and std values are standard for normalizing images in PyTorch. Mean is [0.485, 0.456, 0.406] and std is [0.229, 0.224, 0.225]. The first blank is mean, second is std.