Cornucopia-LLaMA-Fin-Chinese

Why we included this project

Anyone building a Chinese-language finance assistant or Q&A tool can start here without training a model from scratch. The repo ships LoRA adapter weights that turn LLaMA-family base models into versions tuned on Chinese financial question-answer data, with adapters built on both the original Meta LLaMA and the Chinese-LLaMA base. It also documents the instruction-tuning pipeline, including how the training data was assembled from public and scraped financial Q&A sources covering insurance, wealth management, stocks, funds, loans, and credit cards. Inference and fine-tuning scripts plus sample prompts are included, so the project works both as a drop-in finance-aware model and as a reference for reproducing a similar domain fine-tune on your own data.