GeneralizedDiceScore fails with 2D tensors and input_format="index"

[fguiotte]

🐛 Bug

The generalized dice score does not work with 2D tensors of shape (N, L) and input_format="index".

To Reproduce

import torch
from torchmetrics.segmentation import GeneralizedDiceScore

batch_size = 16
num_classes = 8
L = 100

y = torch.randint(num_classes, (batch_size, L))
pred = torch.randint(num_classes, (batch_size, L))

dice = GeneralizedDiceScore(num_classes=num_classes, input_format="index")

dice(y, pred)

ValueError: Expected both `preds` and `target` to have at least 3 dimensions, but got 2.

Expected behavior

According to the documentation, the dice score should work on tensors of shape (N, ...) with input_format="index":

torchmetrics/src/torchmetrics/segmentation/generalized_dice.py

Lines 54 to 61 in 0a64b3f

	- ``preds`` (:class:`~torch.Tensor`): An one-hot boolean tensor of shape ``(N, C, ...)`` with ``N`` being
	the number of samples and ``C`` the number of classes. Alternatively, an integer tensor of shape ``(N, ...)``
	can be provided, where the integer values correspond to the class index. The input type can be controlled
	with the ``input_format`` argument.
	- ``target`` (:class:`~torch.Tensor`): An one-hot boolean tensor of shape ``(N, C, ...)`` with ``N`` being
	the number of samples and ``C`` the number of classes. Alternatively, an integer tensor of shape ``(N, ...)``
	can be provided, where the integer values correspond to the class index. The input type can be controlled
	with the ``input_format`` argument.

Environment

• TorchMetrics version 1.5.1

Additional context

The exception is coming from dice functional:

torchmetrics/src/torchmetrics/functional/segmentation/generalized_dice.py

Lines 55 to 56 in 0a64b3f

	if preds.ndim < 3:
	raise ValueError(f"Expected both `preds` and `target` to have at least 3 dimensions, but got {preds.ndim}.")

[github-actions]

Hi! thanks for your contribution!, great first issue!