[BUG]: Error while assuming fixed number of dimensions in QuantileLoss.loss

[bobbyx27]

WHERE
In pytorch_forecasting.metrics.quantile
In QuantileLoss.loss (shown below)

WHAT
torch.cat with dim=2 does not work as tensors are expected to have unknown number of dimensions.
It works fine with dim=-1.

(By the way I'm not sure about the criterion's scaling "2 *" logic. I don't find it in other references)
(And specifying the quantile dimension is the last one of y_pred/losses should be added to documentation)

CODE

def loss(self, y_pred: torch.Tensor, target: torch.Tensor) -> torch.Tensor:
      # calculate quantile loss
      losses = []
      for i, q in enumerate(self.quantiles):
            errors = target - y_pred[..., i]
            losses.append(torch.max((q - 1) * errors, q * errors).unsqueeze(-1))
      losses = 2 * torch.cat(losses, dim=2)

    return losses
'''