Loading (any) HuggingFace dataset using MLCroissant

[camelia-tfds]

I noticed that, at least when loading datasets using "/workspaces/croissant/python/mlcroissant/mlcroissant/scripts/load.py", we are pointing to the JSON metadata of datasets which are under one of the subdirectories of "/workspaces/croissant/datasets" (0.8 or 1.0 or 1.1).
In those directories there is a certain number of datasets, but they do not seem to cover everything that one can find in HuggingFace (https://huggingface.co/datasets?modality=modality:image&sort=trending).

How does one load a HuggingFace dataset locally if it is not referred to in the codebase?
Is there a condition that any dataset which is to be loaded using mlcroissant has to have metadata under "/workspaces/croissant/datasets" or is there a different way to pass dataset metadata paths to our script?

Thanks in advance for any guidance!

[ccl-core]

Have you tried using the croissant url as jsonld in https://github.com/mlcommons/croissant/blob/f441ad7e832435e80530a1f01f62e93216dd04db/python/mlcroissant/mlcroissant/scripts/load.py#L122C19-L122C26 , e.g. https://huggingface.co/api/datasets/ylecun/mnist/croissant ?

I believe the jsonld argument to the Dataset class should work also with an url.