【大语言模型_1】VLLM部署Qwen模型

1、模型下载：

              魔塔社区：魔搭社区
              huggingface：https://huggingface.co/Qwen

2、安装python环境

             1、python官网安装python 【推荐要3.8以上版本】
             2、安装vllm模块


3、启动模型

      

1. CUDA_VISIBLE_DEVICES=0,1 /root/vendor/Python3.10.12/bin/python3.10 -m vllm.entrypoints.openai.api_server --host 0.0.0.0 --port 25010 --served-model-name mymodel --model //root/qwen2.5/qwen2.5-coder-7b-instruct/ --tensor-parallel-size 2 --max-model-len 8096

复制代码

出现以下内容代表运行乐成

1. INFO 09-20 15:22:59 model_runner.py:1335] Graph capturing finished in 11 secs.
2. (VllmWorkerProcess pid=101403) INFO 09-20 15:22:59 model_runner.py:1335] Graph capturing finished in 11 secs.
3. INFO 09-20 15:22:59 api_server.py:224] vLLM to use /tmp/tmplc42ak3s as PROMETHEUS_MULTIPROC_DIR
4. WARNING 09-20 15:22:59 serving_embedding.py:190] embedding_mode is False. Embedding API will not work.
5. INFO 09-20 15:22:59 launcher.py:20] Available routes are:
6. INFO 09-20 15:22:59 launcher.py:28] Route: /openapi.json, Methods: HEAD, GET
7. INFO 09-20 15:22:59 launcher.py:28] Route: /docs, Methods: HEAD, GET
8. INFO 09-20 15:22:59 launcher.py:28] Route: /docs/oauth2-redirect, Methods: HEAD, GET
9. INFO 09-20 15:22:59 launcher.py:28] Route: /redoc, Methods: HEAD, GET
10. INFO 09-20 15:22:59 launcher.py:28] Route: /health, Methods: GET
11. INFO 09-20 15:22:59 launcher.py:28] Route: /tokenize, Methods: POST
12. INFO 09-20 15:22:59 launcher.py:28] Route: /detokenize, Methods: POST
13. INFO 09-20 15:22:59 launcher.py:28] Route: /v1/models, Methods: GET
14. INFO 09-20 15:22:59 launcher.py:28] Route: /version, Methods: GET
15. INFO 09-20 15:22:59 launcher.py:28] Route: /v1/chat/completions, Methods: POST
16. INFO 09-20 15:22:59 launcher.py:28] Route: /v1/completions, Methods: POST
17. INFO 09-20 15:22:59 launcher.py:28] Route: /v1/embeddings, Methods: POST
18. INFO 09-20 15:22:59 launcher.py:33] Launching Uvicorn with --limit_concurrency 32765. To avoid this limit at the expense of performance run with --disable-frontend-multiprocessing
19. INFO:     Started server process [101179]
20. INFO:     Waiting for application startup.
21. INFO:     Application startup complete.
22. INFO:     Uvicorn running on http://0.0.0.0:25010

复制代码

4、使用python脚本调用测试

1. from openai import OpenAI
3. # 初始化客户端
4. client = OpenAI(base_url="http://localhost:25010/v1", api_key="EMPTY")
6. print("欢迎使用Qwen智能问答机器人！输入'退出'以结束对话。")
8. while True:
9.     # 获取用户输入
10.     print("您: ", end='', flush=True)
11.     user_input = input()
12.     if user_input.lower() in ['退出', '再见', '拜拜']:
13.         print("qwen: 再见！期待下次与您交谈。")
14.         break
16.     # 构造消息列表
17.     messages = [
18.         {"role": "system", "content": "你的角色是名为“qwen”的智能问答机器人"},
19.         {"role": "user", "content": user_input}
20.     ]
22.     try:
23.         # 发送请求并获取回复
24.         chat_completion = client.chat.completions.create(
25.             model="mymodel",
26.             messages=messages,
27.             #stop=[ "。"],
28.             stop=["<|endoftext|>", "<|im_end|>", "<|im_start|>"],
29.             stream = False,
30.         )
32.         # 打印模型回复
33.         print("qwen:", chat_completion.choices[0].message.content)
35.     except Exception as e:
36.         print("出现错误: {e}",e)
37.         print("请稍后再试或检查您的网络连接及API配置。")

复制代码

免责声明：如果侵犯了您的权益，请联系站长，我们会及时删除侵权内容，谢谢合作！更多信息从访问主页：qidao123.com:ToB企服之家，中国第一个企服评测及商务社交产业平台。