How to log a torchmetrics that return multiple values (average=None)

[mnslarcher]

Hi,

I would like to log both the global and the classwise F1 score (or other metrics).

self.log("F1", self.f1, prog_bar=True)

works only if:
self.f1 = F1(num_classes=NUM_CLASSES, average="micro", mdmc_average="global")

but not if:
self.f1 = F1(num_classes=NUM_CLASSES, average=None, mdmc_average="global")

In this second case it complains that:
ValueError: only one element tensors can be converted to Python scalars

There is any smart way to log also the second case or I have to write a for loop and log the value for each class?

What I have to do to ensure that if I use a for loop then the classwise metric is aggretated correctly?

Thanks,
Mario

[mnslarcher]

Sorry I see now that it was already asked here: #7373