Way to change torch tensor from 28x28 to 32x32

Is there any way to reshape tensor shape.

I ran into this problem using TensorDataset and Dataloader. I can’t use transforms here.
Any directly method without changing to numpy?

Thanks for ideas.


Can torch.nn.functional.interpolate this solve your issue?


1 Like

Yay, thanks.
As solution i have converted
[20,784] i.e [20,1,28,28] to [20,1,32,32]
using interpolate as
xb = F.interpolate(xb, size=32,32) #The resize operation on tensor.

1 Like