Improve model card: Add pipeline tag, license, paper/project/code links, abstract, and detailed usage for LeVo (SongGeneration)
#1
by
nielsr
HF Staff
- opened
This PR significantly enhances the model card for the LeVo (SongGeneration) model by:
- Updating the
licensefromunknowntoapache-2.0, based on the strong indication in the GitHub repository and consensus among colleagues. - Adding the
pipeline_tag: text-to-audio, making the model discoverable under the text-to-audio category on Hugging Face. - Including the full paper abstract and a detailed overview from the GitHub README.
- Providing direct links to the associated paper (LeVo: High-Quality Song Generation with Multi-Preference Alignment), the project demo page (https://levo-demo.github.io), and the GitHub repository (https://github.com/tencent-ailab/songgeneration).
- Adding the "Model Versions" table, "Installation" instructions (from scratch and Docker), and a comprehensive "Inference" (sample usage) section, including checkpoint downloads, input format, and various flags, all directly sourced from the official GitHub README.
- Integrating the "Evaluation Performance" tables, "Citation," "License," and "Contact" sections from the original GitHub README for a complete model card.
- Ensuring image links point to raw GitHub assets for proper display.
This update makes the model more informative, discoverable, and easier for users to understand and integrate.