Manu Joseph V

Yes. It is an error in the code. It is already fixed on the main branch. Can you check if that works for you? I'll push a release with all...

This is definitely a good use-case and is currently not supported. It would required a bit of re-factoring and re-writing the preprocessing pipeline and including additional dependencies like `dask`. Once...

What @TuozhenLiu suggested fixes this as of now. Long term fix is in progress and will be pushed as part of the new release. It stems from the PyTorch Lightning...

Looks like a PyTorch Lightning bug. Are you running with a profiler?

I think you should raise a bug, along with a reproducible example, to PyTorch Lightning... meanwhile, probably disable the profiler and try running it.

Will definitely test after relaxing the dependency and eliminate that in the next release.. Thanks for pointing out..

Only some of the parameters from `Trainer` were propagated to PyTorch Lightning. I'm changing that in the next release. Latest code in GitHub already allows for this option using `trainer_kwargs`...

It's difficult to debug an issue unless I can replicate it at my end. Unless you can prepare a way I can replicate it at my end, I doubt I'll...

This looks like a version issue... What is the version of `torchmetrics` you have installed?

Looks like an issue with PyTorch Lightning and torchmetrics version incompatibility... Until the lightning team fixes this (or I change the versions of both), downgrading to `torchmetrics>=0.7.0,