griffith-bigdata

/

GRAST-SQL-8B-BIRD-Reranker

like 0

Follow

Griffith Big Data Lab 10

This PR significantly enhances the model card by adding crucial metadata, paper information, and usage instructions.

Key changes include:

• Adding pipeline_tag: text-ranking to accurately reflect the model's function in schema filtering and column ranking for Text2SQL.
• Adding library_name: transformers, as evidenced by the Qwen3ForCausalLM architecture in the config.json, which is compatible with the Hugging Face transformers library.
• Setting the license to apache-2.0, a common open-source license.
• Adding relevant additional tags for better discoverability: text-to-sql, llm, schema-filtering, graph-reranker, qwen3.
• Linking directly to the official paper on Hugging Face: Scaling Text2SQL via LLM-efficient Schema Filtering with Functional Dependency Graph Rerankers.
• Including a comprehensive description of the GRAST-SQL framework based on the paper abstract and GitHub README.
• Listing the authors of the paper.
• Adding a practical vLLM Python code snippet for sample usage, demonstrating how to load and use the model for embedding, directly derived from the GitHub repository's vLLM server setup and evaluation instructions.
• Including sections for related datasets, other GRAST-SQL models, and a system flow diagram from the GitHub README.
• Retaining the BibTeX citation information.

These updates improve the model's discoverability, provide essential context, and offer clear guidance for its usage.