MINDIE镜像下载，选择对应镜像
https://www.hiascend.com/developer/ascendhub/detail/af85b724a7e5469ebd7ea13c3439d48f
可以选择Atlas 300I-Duo对应镜像

docker pull --platform=arm64 swr.cn-south-1.myhuaweicloud.com/ascendhub/mindie:2.1.RC2-300I-Duo-py311-openeuler24.03-lts

启动docker容器

docker run -it -d --net=host --shm-size=500g 
--privileged 
--name mindie-qwen-7b 
--device=/dev/davinci_manager 
--device=/dev/hisi_hdc 
--device=/dev/devmm_svm 
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver:ro 
-v /usr/local/sbin:/usr/local/sbin:ro 
-v /mnt/data/Qwen2.5-7B-Instruct:/data_1 #挂载模型路径
02efebd6ce64  #镜像ID

修改配置

docker exec -it mindie-qwen-7b bash
vim /usr/local/Ascend/mindie/latest/mindie-service/conf/config.json

需要修改

{
"ServerConfig" :
    {
"ipAddress" : "0.0.0.0",
"port" : 1025,
"managementPort" : 1026,
"metricsPort" : 1027,
"allowAllZeroIpListening" : true,
"httpsEnabled" : false,
}，
"BackendConfig": {
    "npuDeviceIds": [[0, 1]],  // 指定使用的NPU设备ID，根据实际卡数填写
    "ModelDeployConfig": {
      "truncation": false,
      "ModelConfig": [
        {
          "modelName": "qwen-vl-7b",           // 模型名称，调用时使用
          "modelWeightPath": "/容器内模型挂载路径", // 容器内模型权重挂载路径
          "worldSize": 2                       // 使用的NPU卡数，与npuDeviceIds对应
        }
      ]
    }
}

修改完配置启动模型服务

cd /usr/local/Ascend/mindie/latest/mindie-service/bin
./mindieservice_daemon        # 前台启动，方便查看日志
# 或
nohup ./mindieservice_daemon > output.log 2>&1 &  # 后台启动，日志写入文件

测试调用

curl http://192.168.5.100:1025/generate -d '{"inputs":"What is deep learning?","max_new_tokens":32,"stream":false,"do_sample":true,"temperature":0.7,"top_p":0.9,"top_k":20,"model":"Qwen2.5-7B-Instruct"}'