Chinese Llama 2 7b

🚀 Chinese Llama 2 7B

A fully open - source and commercially available Chinese version of the Llama2 model, along with a Chinese - English SFT dataset. It strictly follows the llama - 2 - chat input format and is compatible with all optimizations for the original llama - 2 - chat model.

🚀 Quick Start

✨ Features

• Full Open - Source: The model and dataset are completely open - source, allowing for free commercial use.
• Format Compatibility: It adheres to the llama - 2 - chat input format, making it compatible with all optimizations for the original model.
• Multilingual Support: Supports both Chinese and English, with a Chinese - English SFT dataset containing 10 million data points.

📦 Installation

No specific installation steps are provided in the original document.

💻 Usage Examples

Basic Usage

from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer

model_path = "LinkSoul/Chinese-Llama-2-7b"

tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)
model = AutoModelForCausalLM.from_pretrained(model_path).half().cuda()
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)

instruction = """[INST] <<SYS>>\nYou are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe.  Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.

            If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.\n<</SYS>>\n\n{} [/INST]"""

prompt = instruction.format("用英文回答，什么是夫妻肺片？")
generate_ids = model.generate(tokenizer(prompt, return_tensors='pt').input_ids.cuda(), max_new_tokens=4096, streamer=streamer)

📚 Documentation

Online Demo

Talk is cheap, Show you the Demo.

• Demo Address / HuggingFace Spaces
• One - click Start on Colab // In preparation

Resource Download

• Model Download: Chinese Llama2 Chat Model
• 4 - bit Quantization: Chinese Llama2 4bit Chat Model

We used a Chinese - English SFT dataset with 10 million data points.

• Dataset: https://huggingface.co/datasets/LinkSoul/instruction_merge_set
• Training and Inference Code: https://github.com/LinkSoul-AI/Chinese-Llama-2-7b

🔗 Related Projects

• Llama2

📄 License

Apache - 2.0 license

💬 WeChat Group

Welcome to join the WeChat group