PaDT-MLLM

/

PaDT_Pro_3B

like 2

This PR enhances the model card for "Patch-as-Decodable-Token" by:

• Updating the pipeline_tag from image-text-to-text to any-to-any. This more accurately reflects the model's capabilities to generate both textual and diverse visual outputs (detection, segmentation, grounding predictions), as described in the paper abstract.
• Adding library_name: transformers. The model is compatible with the Hugging Face Transformers library, as evidenced by the transformers.AutoProcessor import in the quick start code snippet and the transformers_version in config.json. This enables the interactive inference widget and automated usage examples on the Hub.
• Integrating the "Why PaDT Succeeds?" section from the GitHub README to provide more comprehensive information about the model's core design and advantages.

Please review and merge if these changes align with the model's functionality and best practices for model cards.