meta-llama

/

Llama-3.2-11B-Vision-Instruct

like 943

Follow

Meta Llama 8,422

In the paper, I could find "We insert normalized (xmin, ymin, xmax, ymax) coordinates directly into the text, demarcated by special tokens." However, there's no special tokens for grounding in tokenizer_config.json. Can you give me some example to use Llama for visual grounding tasks?