OSS Alternative - Discover Top Open Source Alternatives to Popular Software

eosphoros-ai/DB-GPT-Hub

A specialized hub providing models, datasets, and fine-tuning techniques to enhance Large Language Models' performance in Text-to-SQL, Text-to-NLU, and Text-to-GQL tasks.

Core Features

Enhances Text-to-SQL parsing accuracy using LLMs.

Supports Text-to-NLU fine-tuning for improved semantic understanding.

Enables Text-to-GQL fine-tuning for generating graph queries.

Offers evaluated baselines for various LLMs (e.g., Llama2, CodeLlama, Baichuan2).

Integrates LoRA and QLoRA fine-tuning methodologies.

Detailed Introduction

DB-GPT-Hub is a dedicated repository designed to advance the capabilities of Large Language Models (LLMs) in interacting with databases. It centralizes essential models, datasets, and fine-tuning techniques, primarily focused on significantly improving Text-to-SQL parsing accuracy. The project also extends its support to Text-to-NLU for enhanced semantic understanding and Text-to-GQL for generating complex graph queries. By providing comprehensive evaluated baselines and advanced fine-tuning methodologies like LoRA and QLoRA, DB-GPT-Hub empowers developers to optimize LLMs for sophisticated data querying and natural language processing tasks across diverse database paradigms.