lglg666

/

SongGeneration-base

like 1

This PR significantly enhances the model card for the LeVo (SongGeneration) model by:

• Updating Metadata:
  • Setting pipeline_tag: text-to-audio to improve discoverability on the Hugging Face Hub.
  • Changing license from unknown to apache-2.0 as indicated by the project's LICENSE file in the GitHub repository.
• Adding Key Links: Including direct links to the Hugging Face paper page (https://huggingface.co/papers/2506.07520), the project page (https://levo-demo.github.io), and the GitHub repository (https://github.com/tencent-ailab/songgeneration).
• Integrating Comprehensive Content: Populating the model card with detailed information extracted directly from the official GitHub README, including:
  • An overview of the model and its architecture.
  • The abstract of the paper.
  • News and updates from the project.
  • A table of available model versions with download links.
  • Complete installation instructions (from scratch and Docker).
  • Detailed inference instructions with bash usage examples, excluding any Python snippets not found in the original GitHub README to ensure accuracy and avoid misleading users.
  • An extensive input guide for lyrics and descriptions.
  • Information on the Gradio UI.
  • Evaluation performance tables for Chinese and English.
  • Citation details in BibTeX format.
  • A link to the project's license file.
  • Contact information.
• Ensuring Proper Rendering: All relative image paths and the license file link have been converted to absolute URLs pointing to the GitHub raw content, ensuring they display correctly on the Hugging Face Hub.

This update provides a much richer and more informative resource for users interested in the LeVo song generation model.