In 5.3 video 2 how is train_acc callable like a function

[bonosa]

if it's an attribute?

https://github.com/Lightning-AI/dl-fundamentals/blob/main/unit05-lightning/5.3-torchmetrics/5.3-part2-lightning-torchmetrics.py

chatGPT says it's not a regular attribute but an instance of the Accuracy class from torchmetrics. Is that correct?

[rasbt]

It's an object we instantiate from the Accuracy class, and then we assign it to an attribute of the LightningModel. The train_acc object then has a .forward() method that is executed when we call train_acc(...).
I have a blog article that might be helpful to understand what's happening: https://sebastianraschka.com/blog/2022/torchmetrics.html

[bonosa]

the article cleared my doubts. Thanks!