Convert PyTorch loss to torchmetric Metric?

[charlesbmi]

I'm interested in using a Metric like Poisson negative log likelihood loss, which is already implemented in PyTorch as a Module.

Is there a helper-class to wrap existing PyTorch losses in a torchmetric Metric?

In particular, I would like to store and log the best epoch-accumulated loss so far (similar to this example with Accuracy: https://github.com/ashleve/lightning-hydra-template/blob/main/src/models/mnist_model.py#L93-L96). I looked at the MeanSquaredError source code as an example and know that I can re-implement a Metric, but seems like it could be nice to be able to easily wrap a PyTorch loss in a Metric class. Thanks!