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
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
出现以下内容代表运行成功
INFO 09-20 15:22:59 model_runner.py:1335] Graph capturing finished in 11 secs.
(VllmWorkerProcess pid=101403) INFO 09-20 15:22:59 model_runner.py:1335] Graph capturing finished in 11 secs.
INFO 09-20 15:22:59 api_server.py:224] vLLM to use /tmp/tmplc42ak3s as PROMETHEUS_MULTIPROC_DIR
WARNING 09-20 15:22:59 serving_embedding.py:190] embedding_mode is False. Embedding API will not work.
INFO 09-20 15:22:59 launcher.py:20] Available routes are:
INFO 09-20 15:22:59 launcher.py:28] Route: /openapi.json, Methods: HEAD, GET
INFO 09-20 15:22:59 launcher.py:28] Route: /docs, Methods: HEAD, GET
INFO 09-20 15:22:59 launcher.py:28] Route: /docs/oauth2-redirect, Methods: HEAD, GET
INFO 09-20 15:22:59 launcher.py:28] Route: /redoc, Methods: HEAD, GET
INFO 09-20 15:22:59 launcher.py:28] Route: /health, Methods: GET
INFO 09-20 15:22:59 launcher.py:28] Route: /tokenize, Methods: POST
INFO 09-20 15:22:59 launcher.py:28] Route: /detokenize, Methods: POST
INFO 09-20 15:22:59 launcher.py:28] Route: /v1/models, Methods: GET
INFO 09-20 15:22:59 launcher.py:28] Route: /version, Methods: GET
INFO 09-20 15:22:59 launcher.py:28] Route: /v1/chat/completions, Methods: POST
INFO 09-20 15:22:59 launcher.py:28] Route: /v1/completions, Methods: POST
INFO 09-20 15:22:59 launcher.py:28] Route: /v1/embeddings, Methods: POST
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
INFO: Started server process [101179]
INFO: Waiting for application startup.
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:25010
4、利用python脚本调用测试
python
from openai import OpenAI
# 初始化客户端
client = OpenAI(base_url="http://localhost:25010/v1", api_key="EMPTY")
print("欢迎使用Qwen智能问答机器人!输入'退出'以结束对话。")
while True:
# 获取用户输入
print("您: ", end='', flush=True)
user_input = input()
if user_input.lower() in ['退出', '再见', '拜拜']:
print("qwen: 再见!期待下次与您交谈。")
break
# 构造消息列表
messages = [
{"role": "system", "content": "你的角色是名为"qwen"的智能问答机器人"},
{"role": "user", "content": user_input}
]
try:
# 发送请求并获取回复
chat_completion = client.chat.completions.create(
model="mymodel",
messages=messages,
#stop=[ "。"],
stop=["<|endoftext|>", "<|im_end|>", "<|im_start|>"],
stream = False,
)
# 打印模型回复
print("qwen:", chat_completion.choices[0].message.content)
except Exception as e:
print("出现错误: {e}",e)
print("请稍后再试或检查您的网络连接及API配置。")