I have seen in the Dataset pytorch tutorial (Writing Custom Datasets, DataLoaders and Transforms — PyTorch Tutorials 2.2.0+cu121 documentation) that super is not called when inheriting from Dataset.
Does anyone know why this has not been done?
I have seen in the Dataset pytorch tutorial (Writing Custom Datasets, DataLoaders and Transforms — PyTorch Tutorials 2.2.0+cu121 documentation) that super is not called when inheriting from Dataset.
Does anyone know why this has not been done?
torch.utils.data.Dataset
doesn’t define its own __init__
so by calling super().__init__()
you are not performing any meaningful actions. In comparison, calling super().__init__()
when subclassing torch.nn.Module
is required to initialize the internal data structures for storing buffers, modules, etc.
So when extending the Dataset class we should not use super().__init__()
in our init function of the new class that extends Dataset?
You can, but it’s not necessary. The end-result is the same.