Which index the dataloader is chosing

data
1

I am using torch.utils.data.DataLoader. like train_features, train_labels = next(iter(train_dataloader))
I want to determine which index the trainer is chosen?
for batch_idx, data in enumerate(trainloader, 0): can index from 0, but if we shuffle, can we know the indices for this batch?

2

You could return the passed index from the Dataset.__getitem__ method and use it in your DataLoader loop.

1 Like
3

Hello,

In :
https://pytorch.org/tutorials/beginner/basics/data_tutorial.html

when using:

train_features, train_labels = next(iter(train_dataloader))

I believe that train_features takes the values ​​from train_dataloader.data and train_labels takes the values ​​from train_dataloader.targets. However, the class has several other tensors. How do I know which tensor is actually being caught, and why is this tensor being caught instead of the others?

1 Like
4

The Dataset.__getitem__ method is responsible for the loading and processing of the samples and will return these. The DataLoader will just batch these. If you want to check which tensors are processed in which way, check the Dataset implementation.

2 Likes