Update InferenceClient docstring to reflect that token=False is no longer accepted

[abidlabs]

Just wondering -- is there still any way not to pass in the local hf token?

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[hanouticelina]

@abidlabs in practice, HF Inference API still supports passing token=False. We removed that to simplify the logic by avoiding the handle of boolean token for external providers.

[julien-c]

but the HF Inference API requires a token, right? (or if not, it will very very soon)

[Wauplin]

I noticed our internal logic for token=False doesn't really work at the moment (local token is still passed). I'll fix this in a follow-up PR.

And yes with the new inference providers, even HF Inference requires a token now.

[Wauplin]

Hey! Sorry, took me some time to get back to this PR. I've pushed a new commit so that passing token=False raises an error. This is a breaking change of a broken behavior so ok to have it IMO. Previously, passing token=False resulted on the token being sent, which is absolutely the opposite of user's intention. As a reminder, authentication is now required for all HF Inference calls so it just doesn't make sense to pass token=False.

I also took the opportunity to remove some tests that were failing due to authentication issues. They were tests around get_model_status and list_deployed_models which are deprecated anyway.

[Wauplin]

@hanouticelina @julien-c mind re-reviewing it?

[hanouticelina]

thanks!