Hi, thanks for sharing the wonderful library.
I would like to evaluate QwenVL on RefCoco dataset. I see the dataset_id can be swift/refcoco. Yet, the evaluation protocol (link) doesn't support RefCoco.
What is the minimal way to eval VLM on RefCoco? Additionally, could you tell us how to evaluate the model my custom VL dataset?
Thanks in advance.