🐛 Describe the bug

It is intuitive to access training and/or target data in torchvision.datasets.cifar.CIFAR10 class with torchvision.datasets.cifar.CIFAR10.data or torchvision.datasets.cifar.CIFAR10.target.

ISSUE when transform applied these values are different from values acquired using indices i.e. train_dataset[0] done by__getitem__()

import torch
import torchvision.datasets as datasets
import torchvision.transforms as transforms


train_dataset =  datasets.CIFAR10(root='./data', train=True, download=True, transform=train_transforms)
train_transforms = transforms.Compose([
    transforms.RandomHorizontalFlip(p=0.5),
    transforms.ToTensor(),
    transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
])
with_data=train_dataset.data[0,0,0,:]
with_idx,_=train_dataset[0][:,0,0].numpy()

print(with_data == with_idx)

Versions

PyTorch version: 2.3.0+cpu
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A

Python version: 3.11.5 |
Is CUDA available: False
Is XNNPACK available: True

[pip3] numpy==1.26.2
[pip3] torch==2.3.0
[pip3] torchsummary==1.5.1
[pip3] torchvision==0.18.0
[conda] mkl 2021.4.0 pypi_0 pypi
[conda] numpy 1.26.2 pypi_0 pypi
[conda] torch 2.3.0 pypi_0 pypi
[conda] torchsummary 1.5.1 pypi_0 pypi
[conda] torchvision 0.18.0 pypi_0 pypi

Tasks

Beta Give feedback

No tasks being tracked yet.

Hey @CircuitsOfAIM, this seems like it should be in https://github.com/pytorch/vision. Do you mind opening an issue there? Thanks!