Cannot automatically convert prediction head of model class RobertaAdapterModel to flex head.

[swoldemichael]

I was trying to try out my adapter model on the serverless inference API on Hugging Face, but the prediction head can not be converted to flex head. This would be good for prototyping. Perhaps, consider conversion for all models supported by adapters.

[calpt]

Thanks for bringing this up, I think the inference API has not been updated to the new library yet, we will look into this!

[calpt]

This is being fixed in huggingface/api-inference-community#422.

[calpt]

Should be fixed now with merge of the PR mentioned.