Incorrect device handling in weights in BinaryAUROC

[e-pet]

🐛 Describe the bug

This works:

from torcheval.metrics import BinaryAUROC
import torch
auroc = BinaryAUROC()
input = torch.tensor([0.1, 0.2, 0.3, 0.4], device='cuda')
target = torch.tensor([0, 0, 1, 1], device='cuda')
auroc.update(input, target)
auroc.compute()

This does not (identical except for the added weight):

auroc = BinaryAUROC()
input = torch.tensor([0.1, 0.2, 0.3, 0.4], device='cuda')
target = torch.tensor([0, 0, 1, 1], device='cuda')
weight = torch.tensor([0.1, 0.2, 0.3, 0.4], device='cuda')
auroc.update(input, target, weight=weight)
auroc.compute()

"RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!"

I believe the issue is that BinaryAUROC.update() (

torcheval/torcheval/metrics/classification/auroc.py

Line 97 in bb421a7

def update(

)

has input = input.to(self.device) and target = target.to(self.device) but not weight = weight .to(self.device). So I believe this should be a very simple 1-2 line fix. (I did not investigate whether other metrics might have the same issue.)

Versions

irrelevant