Shouldn't this be handled by moving these entries to MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES?
MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES
Yes, please see: https://github.com/huggingface/transformers/pull/40695
· Sign up or log in to comment