CODEWITHSUNDEEP
pythonmachine-learningpytorchtensor
qwesix
qwesix

Reputation: 13

Passing 3 dimensional and one dimensional features to neural network with PyTorch Dataloader

I have examples with size 2x8x8 as tensor and I'm using PyTorch Dataloader for them. But now I want to add an additional 1 dim tensor with size 1 (a single number) as input too.

So I have two input parameters for the neural network, a multidimensional for convolutional layers and an additional one that I will concatenate later.

Probably I could use two dataloader, for every tensor shape one, but than I could not shuffle them.

How can I use a single PyTorch Dataloader for this two different input tensors?

Upvotes: 0

Views: 572

Answers (1)

trialNerror
trialNerror

Reputation: 3563

This is not about the dataloader, this should be done in the your dataset. Implement your own dataset by making it inherit from torch.util.data.Dataset (you need to implement __len__ and __getitem__). Make your __getitem__method return both your tensors, and you should be fine.

You can follow this tutorial if you need.

Upvotes: 1

Related Questions