--- language: - zh tags: - chatglm - pytorch - Text-Generation license: apache-2.0 widget: - text: |- 介绍下北京 答: library_name: peft pipeline_tag: text-generation base_model: THUDM/chatglm-6b inference: false --- # Chinese QA LoRA Model ChatGLM中文问答LoRA模型 `chatglm-6b-belle-zh-lora` evaluate test data: The overall performance of chatglm-6b-belle-zh-lora on QA **test**: |input_text|predict| |:-- |:--- | |介绍下北京|北京是中国的首都,是中国的政治、文化和经济中心。它位于中国的北部,拥有悠久的历史和文化遗产,如故宫、长城和天坛等。北京也是中国最著名的城市之一,有着丰富的美食、文化和娱乐活动。北京还有许多旅游景点,如颐和园、圆明园和天坛等。北京是一个充满活力和机会的城市,吸引了来自全国和世界各地的人们前来探索。| 在中文开放测试集中的表现优异,继承了两方面的优势:1)微调的底座是ChatGLM-6B模型,中文的表现优于LLAMA,2)微调使用的是高质量100万条中文ChatGPT指令Belle数据集,微调后的模型对话效果优于原始ChatGLM-6B。 ## Usage 本项目开源在textgen项目:[textgen](https://github.com/shibing624/textgen),可支持ChatGLM模型,通过如下命令调用: Install package: ```shell pip install -U textgen ``` ```python from textgen import ChatGlmModel model = ChatGlmModel("chatglm", "THUDM/chatglm-6b", peft_name="shibing624/chatglm-6b-belle-zh-lora") r = model.predict(["介绍下北京\n答:"]) print(r) # ['北京是中国的首都,是中国的政治、文化和经济中心。...'] ``` ## Usage (HuggingFace Transformers) Without [textgen](https://github.com/shibing624/textgen), you can use the model like this: First, you pass your input through the transformer model, then you get the generated sentence. Install package: ``` pip install transformers ``` ```python import sys from peft import PeftModel from transformers import AutoModel, AutoTokenizer sys.path.append('..') model = AutoModel.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True, device_map='auto') model = PeftModel.from_pretrained(model, "shibing624/chatglm-6b-belle-zh-lora") model = model.half().cuda() # fp16 tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True) sents = ['介绍下北京\n答:',] for s in sents: response = model.chat(tokenizer, s, max_length=128, eos_token_id=tokenizer.eos_token_id) print(response) ``` output: ```shell 介绍下北京 北京是中国的首都,是中国的政治、文化和经济中心。它位于中国的北部,拥有悠久的历史和文化遗产,如故宫、长城和天坛等。北京也是中国最著名的城市之一,有着丰富的美食、文化和娱乐活动。北京还有许多旅游景点,如颐和园、圆明园和天坛等。北京是一个充满活力和机会的城市,吸引了来自全国和世界各地的人们前来探索。 ``` 模型文件组成: ``` chatglm-6b-belle-zh-lora ├── adapter_config.json └── adapter_model.bin ``` ### 训练数据集 1. 50万条中文ChatGPT指令Belle数据集:[BelleGroup/train_0.5M_CN](https://huggingface.co/datasets/BelleGroup/train_0.5M_CN) 2. 100万条中文ChatGPT指令Belle数据集:[BelleGroup/train_1M_CN](https://huggingface.co/datasets/BelleGroup/train_1M_CN) 3. 5万条英文ChatGPT指令Alpaca数据集:[50k English Stanford Alpaca dataset](https://github.com/tatsu-lab/stanford_alpaca#data-release) 4. 2万条中文ChatGPT指令Alpaca数据集:[shibing624/alpaca-zh](https://huggingface.co/datasets/shibing624/alpaca-zh) 5. 69万条中文指令Guanaco数据集(Belle50万条+Guanaco19万条):[Chinese-Vicuna/guanaco_belle_merge_v1.0](https://huggingface.co/datasets/Chinese-Vicuna/guanaco_belle_merge_v1.0) 如果需要训练ChatGLM模型,请参考[https://github.com/shibing624/textgen](https://github.com/shibing624/textgen) ## Citation ```latex @software{textgen, author = {Xu Ming}, title = {textgen: Implementation of language model finetune}, year = {2021}, url = {https://github.com/shibing624/textgen}, } ```