Improve model card: Add Tequila paper details, metadata, and update tags
#3
by
nielsr
HF Staff
- opened
This PR improves the model card for the AngelSlim toolkit, focusing on the Tequila quantization method presented in the paper "Tequila: Trapping-free Ternary Quantization for Large Language Models".
Key changes include:
- Added metadata:
pipeline_tag: text-generationfor improved discoverability.library_name: transformersbased onconfig.jsonand usage patterns.license: apache-2.0derived from the project's GitHub repository.
- Updated
tags: Addedquantizationto reflect the Tequila method. - Enhanced content:
- Added a prominent introductory section linking to the Tequila paper (https://huggingface.co/papers/2509.23809) and its specific implementation code (https://github.com/Tencent/AngelSlim/tree/tequila/TernaryQuant).
- The existing comprehensive documentation for the AngelSlim toolkit is retained to provide full context and usage instructions for all supported features, including quantization and speculative decoding.
These updates aim to make the model card more informative and discoverable while accurately representing the capabilities of the AngelSlim project.