AngelSlim

/

Qwen3-32B_nvfp4

Like 3

Follow

AngelSlim 423

This PR updates the model card for Qwen3-32B-NVFP4, a model checkpoint from the AngelSlim project. The goal is to enhance its discoverability and provide more specific information, linking it to the relevant research paper.

Changes include:

• Adding license: apache-2.0, library_name: transformers, and pipeline_tag: text-generation to the metadata for better integration with the Hugging Face Hub features.
• Updating the main content to focus on the Qwen3-32B-NVFP4 model and its relation to the Tequila: Trapping-free Ternary Quantization for Large Language Models paper (https://huggingface.co/papers/2509.23809). This model is part of the broader AngelSlim compression toolkit, which also implements Tequila.
• Including links to the paper, the AngelSlim GitHub repository, and its documentation. A specific link to the Tequila implementation in the AngelSlim repository is also provided.
• The extensive generic AngelSlim project README content has been replaced with a focused overview relevant to this specific quantized model, while retaining the Qwen3 series benchmark table for context.
• A direct Python transformers inference code snippet has been omitted as per guidelines, as no explicit snippet was found in the GitHub README for this specific model; instead, users are directed to the project's deployment documentation.
• BibTeX citation for the Tequila paper has been added.