Fix 1 -> The originally imported library AutoModel does not work for this model. AutoModelForCausalLM is the correct one.
Fix 2 -> The original code uses an object detection token (<OD>) for abnormality grounding. The correct token to be used for abnormality grounding is <CAPTION_TO_PHRASE_GROUNDING> along with the abnormality and the definition as input.
