03 Pytorch Computer Vision

[ZaidKhan-git]

I am confused how does next(iter(train_dataloader)) returns train_features_batch and train_labels_batch ?
Also if dataloader already returns an iterable, then why are we making it iterable again ? I am bit confused here.

[tt-a1i]

DataLoader is iterable, but it isn't the iterator itself.

iter(train_dataloader) creates an iterator, then next(...) asks that iterator for one batch. Since each dataset item contains a feature and a label, the batch also contains two values, which can be unpacked like this:

train_features_batch, train_labels_batch = next(iter(train_dataloader))

A for loop does the same work automatically:

for train_features_batch, train_labels_batch in train_dataloader:
    break

You only need to call iter() yourself when you want to manually fetch a batch with next().