Pearson Correlation Coefficient raises error when 2D tensor but single task

[shenoynikhil]

🐛 Bug

I have a regression based modelling repository where the predictions can be multi-output or single-output based on configuration. My network outputs [n_samples, n_tasks] where n_task varies according to the task. If n_task is 1 then trying, torchmetrics.functional.pearson_corrcoef(predictions, targets) gives the error,

ValueError: Expected argument `num_outputs` to match the second dimension of input, but got 1 and 1

Changing the output shape for a single task specifically just so as to fit the metric function does not seem like a good solution. I think a simple change should be able to fix it.
My current workout around,

import torchmetrics.functional as Fm

# predictions are [n, 1] for single task/output
Fm.pearson_corrcoef(predictions, targets) if predictions.shape[1] > 1 else Fm.pearson_corrcoef(predictions[:, 0], targets[:, 0])

There are other metrics that handle this,

        metrics = {
            "mse": Fm.mean_squared_error(predictions, targets, squared=True),
            "rmse": Fm.mean_squared_error(predictions, targets, squared=False),
            "mae": Fm.mean_absolute_error(predictions, targets),
            "r2": Fm.r2_score(predictions, targets, multioutput="raw_values"),
            "mape": Fm.mean_absolute_percentage_error(predictions, targets),
            # TODO: Raise issue on torchmetrics
            "pcc": (
                Fm.pearson_corrcoef(predictions, targets) if predictions.shape[1] > 1 else
                Fm.pearson_corrcoef(predictions[:, 0], targets[:, 0])
            ),
        }

To Reproduce

Steps to reproduce the behavior...

Code sample

Expected behavior

Environment

• TorchMetrics version (and how you installed TM, e.g. conda, pip, build from source):
• Python & PyTorch Version (e.g., 1.0):
• Any other relevant information such as OS (e.g., Linux):

Additional context

[github-actions]

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