Iteration over a metric never ends

[dakinggg]

🐛 Bug

If you try to iterate over a metric (by mistake in my case), it just spins forever, making it appear that the program is hung. I'm not sure internally to torchmetrics exactly why this behavior occurs and if it is desired, but it would be nice if it errored out or something, unless there is a use for iterating over a metric that I am not aware of?

from torchmetrics.classification import Accuracy
acc = Accuracy()
for i, item in enumerate(acc):
    print(i, item)

The above code will spin forever, printing something like this

8144 CompositionalMetric(
  <lambda>(
    Accuracy(),
    None
  )
)

Expected behavior

I would expect an error to be raised if I try to iterate over a metric.

Environment

• TorchMetrics version (and how you installed TM, e.g. conda, pip, build from source): 0.10.2