Include cardData in list_models and list_datasets

[muellerzr]

To support listing the carbon emissions as well as any other metadata that can be returned with cardData:True, this PR now includes that as a parameter to both list_models and list_datasets.

Note: By default this is True because timing wise I did not see a difference between with vs without cardData, and returning the metadata for each model and dataset only increases the understandability of each one IMO.

cc @julien-c @LysandreJik

[LysandreJik]

Thanks for your PR!

Would it be possible to add a bit of docs regarding how this attribute can be leveraged, for example with the co2_eq_emissions?

https://huggingface.co/docs/hub/searching-the-hub

I observed the same, putting cardData to True or False does not change the time required to list all models, so it's fine for me to put it enabled by default.

[julien-c]

I observed the same, putting cardData to True or False does not change the time required to list all models, so it's fine for me to put it enabled by default.

can you compare the underlying API calls? I think you're not seeing too much bc we have good bandwidth, but the underlying calls outputs from /api/models would have very different content sizes imo.

[muellerzr]

@julien-c @LysandreJik should be good now.

I made the cardData none by default, but the explicit docstring tells what extra data you can get back from it.

[julien-c]

lgtm!

[LysandreJik]

Other than the tests, LGTM

[muellerzr]

Merging PR. For transparency:

        self.assertTrue(all([not hasattr(dataset, "cardData") for dataset in datasets]))

This will always fail because it is hard coded into the DatasetResult to set cardData to None, even if it's not present.

@LysandreJik let me know if you'd like me to clean that up a bit with some setattr's instead of hard-coding inputs.