Why does F1, Recall, Precision, and Accuracy are outputting the same thing in my implementation?

[FeryET]

I have created a very barebone pl.LightningModule and in that I have create a MetricCollection for each of my train, val and test stages. I have uploaded my example in this gist: The aformentiond gist

As you can see there is something fishy going on. Is log_dict appending the same values to all keys?

[FeryET]

Apparently someone has reproduced this recently:

https://stackoverflow.com/questions/69139618/torchmetrics-does-not-work-with-pytorchlightning