Chinese-LLaMA-Alpaca
中文LLaMA&Alpaca大语言模型+本地CPU/GPU训练部署 (Chinese LLaMA & Alpaca LLMs)
What is Chinese-LLaMA-Alpaca?
中文LLaMA&Alpaca大语言模型+本地CPU/GPU训练部署 (Chinese LLaMA & Alpaca LLMs)
Overview of Chinese-LLaMA-Alpaca
What it is
Chinese-LLaMA-Alpaca is an open-source project that provides Chinese LLaMA base models and Alpaca instruction-tuned large language models. It expands the Chinese vocabulary of the original LLaMA and performs secondary pre-training with Chinese data to improve Chinese semantic understanding. The Alpaca models are fine-tuned with Chinese instruction data to enhance instruction comprehension and execution. The project supports local CPU/GPU training and deployment.
Key features
- Expanded Chinese vocabulary for original LLaMA, improving Chinese encoding/decoding efficiency
- Open-sourced Chinese LLaMA pre-trained models and Chinese Alpaca instruction-tuned models
- Open-sourced pre-training and instruction fine-tuning scripts for further model training
- Supports local CPU/GPU quantization and deployment on personal computers
- Compatible with ecosystems such as 🤗transformers, llama.cpp, text-generation-webui, LlamaChat, LangChain, privateGPT
Use cases
- Chinese LLaMA: text continuation (given preceding content, generate following text)
- Chinese Alpaca: instruction understanding (question answering, writing, suggestions, etc.), multi-turn context understanding (chat, etc.)
- Not suitable for: LLaMA for instruction understanding or multi-turn chat; Alpaca for unrestricted free text generation
Getting started
See the repository README for model download, LoRA weight merging, and local deployment instructions.
License and pricing
Licensed under Apache-2.0. Open source and free.
Based on the project's public README and repository metadata.