task-mismatch
its any-text, if the argument is their isnt the category, I dont see how you cant fix that?
Hi @TimeLordRaps -
Thanks for reaching out to us! Could you please share more details on what you mean by any-text and what behavior you are expecting when the category doesn’t exist?
like this model takes in "any" modality of the standardized forms, text, images, and audio, and outputs text, granted I may have missed an audio decoder, the categorical error is more so on the side of huggingface, and requires a more ubiquitous solution of automatically cleaning the task specification hierarchy that has been a mess since its introduction. I know we just had a major version, so maybe for the next one a modality generalized form could be better. I was hoping that since this model is from google, they could apply pressure on huggingface to get the task specification and option on the models page sidebar to be more concrete, while being easier to use. I hope that clears things up.