PyTorch DataLoader Returning List Instead of Tuple

vision
1

Hi,

My issue is similar to this solved one, but I am wondering about why the DataLoader returns a list instead of a tuple.

Code for reproducing (PyTorch 2.0.1):

from typing import Any, Optional

import torch 
from torch.utils.data import Dataset, DataLoader 

class ImageDataset(Dataset): 
    def __init__(self, imgs: torch.Tensor, targets: torch.Tensor, img_transform: Optional[Any] = None) -> None:
        """Image dataset class.""" 
        super().__init__()
        self.imgs = imgs
        self.targets = targets
        self.img_transform = img_transform
    
    def __len__(self) -> int:
        assert len(self.imgs) == len(self.targets), "Number of images and targets must be equal"
        return len(self.imgs)

    def __getitem__(self, idx: int) -> Tuple[torch.Tensor, torch.Tensor]:
        img = self.imgs[idx]
        if self.img_transform is not None:
            img = self.img_transform(img)

        target = self.targets[idx]

        return img, target


if __name__ == "__main__": 
    imgs = torch.rand(256, 3, 32, 32)
    targets = torch.randint(0, 10, (256,))
    dataset = ImageDataset(imgs, targets)
    dataset[0]

The type of dataset[0] is tuple, just as I would expect, but putting the dataset into a DataLoader yields:

data_loader = DataLoader(dataset, batch_size=1, shuffle=False)
for idx, data in enumerate(data_loader):
    print(data)
    break

Why is the type of data a list now? I would still expect a tuple…

Best,
Imahn

2

The default collate_fn seems to return lists for Sequences as seen here. If you want to return a tuple you might need to implement a custom collate_fn.