Understanding the Dataset Class

Part 1 (2019)
1

I am having trouble understanding the Dataset class. I understand the the magic methods__getitem__ and__len__ allow us to index and get the length of the Dataset object. However, I’m not sure how this is implemented.

I’ve tried to create a list for example:

ls = [1,2,38,9]

and then

l = Dataset(ls)

but I get

Traceback (most recent call last):
File “<pyshell#14>”, line 1, in
ls.Dataset()
AttributeError: ‘list’ object has no attribute ‘Dataset’

I have also tried

l = Dataset([1,2,6,7])

but get

Traceback (most recent call last):
File “<pyshell#20>”, line 1, in
l = Dataset([1,2,6,7])
TypeError: object() takes no parameters

Any help anyone could provide on how exactly this implementation works would be greatly appreciated.

2

Dataset is part of PyTorch, an abstract class. You will need to subclass it to make MyOwnDataset, and implement the magic methods yourself. There are lots of examples and tutorials online.