HuggingFace Hub

[srush]

It seems like HuggingFace is only supported for local models. It would be nice to support calling model on the Hub with the InferenceAPI. I think this is relatively straightforward to do, you just give the model name and the type of inference (text-generation or feature-extraction)