Wayne-King

/

SRUM_BAGEL_7B_MoT

like 1

This PR significantly enhances the model card for SRUM by incorporating more detailed information and improving discoverability on the Hugging Face Hub.

Key updates include:

• Metadata: Added library_name: transformers based on evidence of Qwen2Tokenizer and Qwen2ForCausalLM usage, enabling automated transformers code snippets.
• Paper Abstract: Included the full abstract of the paper, providing a comprehensive overview of the model's self-rewarding framework and its impact on unified multimodal models.
• Comprehensive Links: Consolidated and added explicit links to the Hugging Face paper page, project website, GitHub repository, Hugging Face Datasets, and Hugging Face Spaces demo, making related artifacts easily accessible.
• Visual Enhancements: Integrated the primary model logo (srum_log_2.png) for better visual context.
• Quick Start Guide: Included the "Quick Start" section from the GitHub repository, providing clear instructions and code snippets for setting up the environment and downloading model checkpoints, directly addressing practical usage.
• News Section: Added the "News" section from the GitHub README to provide updates from the authors.
• Content Preservation: Maintained all valuable existing sections, such as the author list, inference hyperparameters, benchmarks, and citation, ensuring no loss of current information.

These changes aim to make the SRUM model card more informative, user-friendly, and aligned with Hugging Face Hub best practices.