validation_step description differs from implementation

[botcs]

Hi,

this is just a tiny issue that I came across while adding a custom dataset, that's not in ImageFolder format.
In solo/methods/base.py L553 the validation_step says the following:

    """Validation step for pytorch lightning. It does all the shared operations, such as
    forwarding a batch of images, computing logits and computing metrics.

    Args:
        batch (List[torch.Tensor]):a batch of data in the format of [img_indexes, X, Y].
        batch_idx (int): index of the batch.

    Returns:
        Dict[str, Any]: dict with the batch_size (used for averaging), the classification loss
            and accuracies.
    """

However the batch is only split to X, Y according to L568:
X, targets = batch

As I said, it's a tiny thing, but not sure why or where does the library modify the dataset/dataloader to have a different return value depending on train/val split.
Sorry for the rookie question.

Thanks,
Csabi

[vturrisi]

Hey! I think we just messed up the docstring and forgot to update it a long time ago. I would say that for sure we have other inconsistencies in the documentation that we plan on progressively updating. I'll try to open a small PR to fix this one.