KDliang

/

crashchat

like 0

This PR adds a model card for CrashChat, a Multimodal Large Language Model for multitask traffic crash video analysis.

The following metadata is added:

• pipeline_tag: video-text-to-text: The model analyzes crash videos and generates text (descriptions, localization tags, reasoning).
• library_name: transformers: The model configuration includes auto_map and references transformers_version 4.46.3.

The README includes an introduction to the model, a description of the multitask learning strategy, a summary of the model zoo, and the official citation.