OD-FinLLM is a specialized large language model designed for Chinese financial knowledge, fine-tuned from the LLaMA series. It targets users needing accurate financial Q&A and analysis in Chinese, offering improved performance over general-purpose models through a curated dataset and knowledge distillation.

How It Works

OD-FinLLM is built by fine-tuning LLaMA models (LLaMA-2-7B and LLaMA-3.1-8B) on a custom instruction dataset. This dataset comprises publicly available Chinese financial Q&A and web-scraped financial information. A knowledge distillation process transfers capabilities from a larger teacher model (LLaMA-3.1-8B) to a more efficient student model (LLaMA-2-7B), optimizing for financial domain accuracy and reduced computational requirements.

Quick Start & Requirements

• Install dependencies: pip install -r requirements.txt
• Python version: 3.9+
• Model weights available on Huggingface.
• Inference: cd OD-FinLLM && sh run.sh
• Web Demo: python web_demo.py
• Recommended GPU: 24GB+ VRAM (e.g., 3090/4090) for training.

Highlighted Details

• Fine-tuned on LLaMA-2-7B and LLaMA-3.1-8B for Chinese financial Q&A.
• Utilizes a specialized instruction dataset derived from financial Q&A and web scraping.
• Employs knowledge distillation from a larger teacher model to a smaller student model.
• Supports both inference and a web demo interface.

Maintenance & Community

• Developed by the AI team at Firefly.
• Primary contact via GitHub Issues.

Licensing & Compatibility

• License: Apache 2.0
• Disclaimer states resources are strictly for academic research purposes and not for commercial use.

Limitations & Caveats

The project explicitly states that resources are for academic research only and not for commercial use. The accuracy of generated content is not guaranteed due to computational factors and potential precision loss.