- 前端:gradio
- 后端:LLM + RAG
- 环境安装
pip install -r requirements.txt
pip install torch==2.2.1
- 运行
CUDA_VISIBLE_DEVICES=0 python chat.py --llm_path Qwen1.5-7B-Chat --rerank_path bge-reranker-v2-m3
- 访问 http://127.0.0.1:7860/ 即可
# 从阿里云将文档问答系统的镜像拉取到本地
docker pull registry.cn-hangzhou.aliyuncs.com/zhanglonghui/docqa:v1
# docker运行
# 下载到本地的Qwen1.5-7B-Chat路径
qwen_model_path=""
# 下载到本地的bge-reranker-v2-m3路径
rerank_model_path=""
# gpu
device="0"
docker run --gpus device=${device} -v ${qwen_model_path}:/usr/src/app/Qwen1.5-7B-Chat -v ${rerank_model_path}:/usr/src/app/bge-reranker-v2-m3 -p 7860:7860 docqa
# 通过端口7860对demo进行访问
http://localhost:7860
# 可能的报错及解决方案:
# docker run ... 如下报错时,参考教程 https://zhuanlan.zhihu.com/p/373493938
# docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]].