在ACS上使用PPU PIP服务示例-真武 PPU 云服务(ppu)-阿里云帮助中心

本文为您介绍在DinD、Buildah、普通PPU容器等场景下在ACS上使用PPU PIP服务示例。

前提条件

熟悉如何在ACS上使用PTG提供的PPU PIP服务和阿里云提供的PPU PIP服务。具体操作，请参见在ACS上使用PPU PIP服务。
记录ACS集群的Service Account名称，后续的配置步骤会引用该名称。具体操作，请参见在ACS产品中开通PPU PIP免密授权。
熟悉配置镜像鉴权的Secret、准备模型文件、部署GPU算力等步骤。具体操作，请参见部署推理算力。

使用PPU PIP服务示例

DinD示例

YAML文件定义了预留实例场景，其他场景请修改metadata.labels。

apiVersion: v1
kind: Pod
metadata:
  labels:
    alibabacloud.com/compute-class: gpu-hpn
    alibabacloud.com/gpu-model-series: PPU810E
    alibabacloud.com/compute-qos: default
    alibabacloud.com/hpn-type: "rdma"
  name: acs-dind-demo-fuyi01
spec:
  containers:
  - image: registry.cn-hangzhou.aliyuncs.com/acs-demo-ns/docker:27-dind
    name: main
    resources: #根据需求调整资源
      limits:
        cpu: "16"
        ephemeral-storage: 256Gi
        memory: 128Gi
        alibabacloud.com/ppu: "2"
      requests:
        cpu: "16"
        ephemeral-storage: 256Gi
        memory: 128Gi
        alibabacloud.com/ppu: "2"
    volumeMounts:
    - mountPath: /var/lib/docker
      name: docker
    # 可选 使用buildkit需要设置
    - mountPath: /var/lib/buildkit
      name: buildkit
    securityContext:
      #  声明为特权容器，需要提交工单申请特权能力
      privileged: true
  
  #添加前序步骤创建的ServiceAccount名称
  serviceAccountName: pip-default
  
  volumes:
  - emptyDir: {}
    name: docker
  # 可选 使用buildkit需要设置
  - emptyDir: {}
    name: buildkit

创建配置DinD容器。

#创建DinD容器
kubectl apply -f dind.yaml

#连接进入DinD容器
kubectl exec -it acs-dind-demo-fuyi01 -- sh

#【DinD容器内】鉴权，为了拉取PPU镜像
docker login --username=public_pull@1903015075229209 egslingjun-registry.cn-wulanchabu.cr.aliyuncs.com --password=CnpER062Qo!

#【DinD容器内】镜像拉取到本地
docker pull egslingjun-registry.cn-wulanchabu.cr.aliyuncs.com/egslingjun/training-xpu-pytorch:25.06

#【DinD容器内】启动内层PPU容器
id=$(docker run --network host -tid --privileged \
-e ALIBABA_CLOUD_OIDC_PROVIDER_ARN=$ALIBABA_CLOUD_OIDC_PROVIDER_ARN \
-e ALIBABA_CLOUD_ROLE_ARN=$ALIBABA_CLOUD_ROLE_ARN \
-e ALIBABA_CLOUD_OIDC_TOKEN_FILE=$ALIBABA_CLOUD_OIDC_TOKEN_FILE \
-e ALIBABA_CLOUD_STS_ENDPOINT=$ALIBABA_CLOUD_STS_ENDPOINT \
-v /var/run/secrets/ack.alibabacloud.com/rrsa-tokens/:/var/run/secrets/ack.alibabacloud.com/rrsa-tokens/ \
egslingjun-registry.cn-wulanchabu.cr.aliyuncs.com/egslingjun/training-xpu-pytorch:25.06 bash)

#【DinD容器内】进入内容PPU容器
docker exec -it $id bash

#【内层PPU容器内】安装免密插件
pip install aiext-pypi-plugin -i http://mirrors.cloud.aliyuncs.com/aiext-pypi/aiext-pypi-plugin/simple --trusted-host mirrors.cloud.aliyuncs.com

#【内层PPU容器内】安装一个镜像中不存在的torchdata （注意版本匹配）
pip install torchdata -i https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu126/simple/

# 在内层容器内可以进行必要的apt install / pip install，环境开发过程略

PIP免密安装过程。

root@acs-dind-demo-fuyi01:/workspace# pip install torchdata -i https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu126/simple/
[AIEXT_PYPI_PLUGIN] Plugin has been loaded.
Looking in indexes: https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu126/simple/, http://mirrors.cloud.aliyuncs.com/pypi/simple/
Collecting torchdata
  Downloading https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu126/packages/torchdata/0.11.0%2Bppu1.5.0.ce/torchdata-0.11.0%2Bppu1.5.0.ce-py3-none-any.whl (59 kB)
Requirement already satisfied: urllib3>=1.25 in /opt/ac2/lib/python3.12/site-packages (from torchdata) (2.5.0)
Requirement already satisfied: requests in /opt/ac2/lib/python3.12/site-packages (from torchdata) (2.32.4)
Requirement already satisfied: torch>=2 in /opt/ac2/lib/python3.12/site-packages (from torchdata) (2.6.0+ali.7.post1.ppu1.5.2.cu126)
Requirement already satisfied: filelock in /opt/ac2/lib/python3.12/site-packages (from torch>=2->torchdata) (3.18.0)
Requirement already satisfied: typing-extensions>=4.10.0 in /opt/ac2/lib/python3.12/site-packages (from torch>=2->torchdata) (4.14.0)
Requirement already satisfied: setuptools in /opt/ac2/lib/python3.12/site-packages (from torch>=2->torchdata) (80.9.0)
Requirement already satisfied: sympy==1.13.1 in /opt/ac2/lib/python3.12/site-packages (from torch>=2->torchdata) (1.13.1)
Requirement already satisfied: networkx in /opt/ac2/lib/python3.12/site-packages (from torch>=2->torchdata) (3.5)
Requirement already satisfied: jinja2 in /opt/ac2/lib/python3.12/site-packages (from torch>=2->torchdata) (3.1.4)
Requirement already satisfied: fsspec in /opt/ac2/lib/python3.12/site-packages (from torch>=2->torchdata) (2024.9.0)
Requirement already satisfied: triton==3.2.0 in /opt/ac2/lib/python3.12/site-packages (from torch>=2->torchdata) (3.2.0+ppu1.5.2.cu126)
Requirement already satisfied: mpmath<1.4,>=1.1.0 in /opt/ac2/lib/python3.12/site-packages (from sympy==1.13.1->torch>=2->torchdata) (1.3.0)
Requirement already satisfied: MarkupSafe>=2.0 in /opt/ac2/lib/python3.12/site-packages (from jinja2->torch>=2->torchdata) (3.0.2)
Requirement already satisfied: charset_normalizer<4,>=2 in /opt/ac2/lib/python3.12/site-packages (from requests->torchdata) (3.4.2)
Requirement already satisfied: idna<4,>=2.5 in /opt/ac2/lib/python3.12/site-packages (from requests->torchdata) (3.10)
Requirement already satisfied: certifi>=2017.4.17 in /opt/ac2/lib/python3.12/site-packages (from requests->torchdata) (2025.6.15)
Installing collected packages: torchdata

root@acs-dind-demo-fuyi01:/workspace# pip list | grep torchdata
torchdata                                0.11.0+ppu1.5.0.ce

退出并保存DinD容器。

#【内层PPU容器内】从内层PPU容器退出到DinD容器
exit

#【DinD容器内】保存内层PPU容器为镜像
docker commit $id my-new-image:new-id

Buildah示例

YAML文件定义了预留实例场景，其他场景请修改metadata.labels。

kind: Pod
apiVersion: v1
metadata:
  name: acs-buildah-demo-fuyi
  labels:
    alibabacloud.com/compute-class: gpu-hpn
    alibabacloud.com/gpu-model-series: PPU810E
    alibabacloud.com/compute-qos: default
    alibabacloud.com/hpn-type: "rdma"
spec:
  restartPolicy: OnFailure
  volumes:
  - emptyDir: {}
    name: buildah
  containers:
  - name: builder
    image: registry-cn-hangzhou.ack.aliyuncs.com/ack-demo/buildah:sampleBuild
    command:
    - sh
    - -c
    - sleep infinity 
    imagePullPolicy: Always
    volumeMounts:
    - mountPath: /var/lib/containers/storage
      name: buildah
    resources: #根据需求调整资源
      limits:
        cpu: "16"
        memory: "128Gi"
        ephemeral-storage: 500Gi
        alibabacloud.com/ppu: 2
      requests:
        cpu: "16"
        memory: "128Gi"
        ephemeral-storage: 500Gi
        alibabacloud.com/ppu: 2

  #添加前序步骤创建的ServiceAccount名称
  serviceAccountName: pip-default

创建配置Buildah容器。

# 创建Buildah容器
kubectl apply -f buildah.yaml

# 进入Buildah容器
kubectl exec -it acs-buildah-demo-fuyi -- sh


#在buildah容器里边，可以使用buildah一比一替换docker的命令
# 【普通PPU容器内】登录，便于拉取公网镜像
buildah login --username=public_pull@1903015075229209 egslingjun-registry.cn-wulanchabu.cr.aliyuncs.com --password=CnpER062Qo!

#【普通PPU容器内】拉取Base镜像到本地，便于构建过程
buildah pull egslingjun-registry.cn-wulanchabu.cr.aliyuncs.com/egslingjun/training-xpu-pytorch-c:xp-fuyao-ppu1.4.0_hotfix2-cuda11.4-torch1.10-py38-24.12.27-squash

#【普通PPU容器内】使用如下Dockerfile构建新镜像（只是演示使用免密pip）
buildah --storage-driver=overlay build --no-cache \
--build-arg ALIBABA_CLOUD_OIDC_PROVIDER_ARN=$ALIBABA_CLOUD_OIDC_PROVIDER_ARN \
--build-arg ALIBABA_CLOUD_ROLE_ARN=$ALIBABA_CLOUD_ROLE_ARN \
--build-arg ALIBABA_CLOUD_STS_ENDPOINT=$ALIBABA_CLOUD_STS_ENDPOINT \
-t demo:v1 .

PIP免密安装（省略了其他Dockerfile指令）。

# 使用ACS提供的标准镜像，如下镜像尺寸较小
FROM egslingjun-registry.cn-wulanchabu.cr.aliyuncs.com/egslingjun/training-xpu-pytorch-c:xp-fuyao-ppu1.4.0_hotfix2-cuda11.4-torch1.10-py38-24.12.27-squash

# 其他Dockerfile步骤省略

# 只示意使用阿里云提供的PIP软件安装相关的步骤
## 1. 设置免密相关Env
ARG ALIBABA_CLOUD_STS_ENDPOINT
ARG ALIBABA_CLOUD_OIDC_PROVIDER_ARN
ARG ALIBABA_CLOUD_ROLE_ARN
ENV ALIBABA_CLOUD_STS_ENDPOINT=${ALIBABA_CLOUD_STS_ENDPOINT}
ENV ALIBABA_CLOUD_OIDC_PROVIDER_ARN=${ALIBABA_CLOUD_OIDC_PROVIDER_ARN}
ENV ALIBABA_CLOUD_ROLE_ARN=${ALIBABA_CLOUD_ROLE_ARN}
ENV ALIBABA_CLOUD_OIDC_TOKEN_FILE=/mount/rrsa-tokens/token

## 2. 首先安装免密插件aiext-pypi-plugin，安装插件限制（暂不支持以python -m pip执行请求）

RUN pip install aiext-pypi-plugin -i http://mirrors.cloud.aliyuncs.com/aiext-pypi/aiext-pypi-plugin/simple --trusted-host mirrors.cloud.aliyuncs.com

## 3. 然后每一条pip install前都要增加--mount选项，就可以正常pip install了（找了2个镜像中未装的包）：
RUN --mount=type=bind,source=/var/run/secrets/ack.alibabacloud.com/rrsa-tokens,target=/mount/rrsa-tokens \
    pip install cumm-cu114 -i https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu114/simple/
RUN --mount=type=bind,source=/var/run/secrets/ack.alibabacloud.com/rrsa-tokens,target=/mount/rrsa-tokens \
    pip install spconv-cu114 -i https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu114/simple/

## 4. PIP安完成后建议清理环境
RUN pip uninstall -y aiext-pypi-plugin

构建过程输出如下所示：

sh-5.2# buildah --storage-driver=overlay build --no-cache \
--build-arg ALIBABA_CLOUD_OIDC_PROVIDER_ARN=$ALIBABA_CLOUD_OIDC_PROVIDER_ARN \
--build-arg ALIBABA_CLOUD_ROLE_ARN=$ALIBABA_CLOUD_ROLE_ARN \
--build-arg ALIBABA_CLOUD_OIDC_TOKEN_FILE=$ALIBABA_CLOUD_OIDC_TOKEN_FILE \
--build-arg ALIBABA_CLOUD_STS_ENDPOINT=$ALIBABA_CLOUD_STS_ENDPOINT \
-t demo:v1 .
STEP 1/13: FROM egslingjun-registry.cn-wulanchabu.cr.aliyuncs.com/egslingjun/training-xpu-pytorch-c:xp-fuyao-ppu1.4.0_hotfix2-cuda11.4-torch1.10-py38-24.12.27-squash
STEP 2/13: ARG ALIBABA_CLOUD_STS_ENDPOINT
STEP 3/13: ARG ALIBABA_CLOUD_OIDC_PROVIDER_ARN
STEP 4/13: ARG ALIBABA_CLOUD_ROLE_ARN
STEP 5/13: ARG ALIBABA_CLOUD_OIDC_TOKEN_FILE
STEP 6/13: ENV ALIBABA_CLOUD_STS_ENDPOINT=sts-vpc.cn-wulanchabu.aliyuncs.com
STEP 7/13: ENV ALIBABA_CLOUD_OIDC_PROVIDER_ARN=acs:ram::1697237910442391:oidc-provider/ack-rrsa-c69a187d1281446df95887e01a6b26c6b
STEP 8/13: ENV ALIBABA_CLOUD_ROLE_ARN=acs:ram::1697237910442391:role/pip-default-c69a187d1281446df95887e01a6b26c6b
STEP 9/13: ENV ALIBABA_CLOUD_OIDC_TOKEN_FILE=/var/run/secrets/ack.alibabacloud.com/rrsa-tokens/token
STEP 10/13: RUN pip install aiext-pypi-plugin -i http://mirrors.cloud.aliyuncs.com/aiext-pypi/aiext-pypi-plugin/simple --trusted-host mirrors.cloud.aliyuncs.com
Looking in indexes: http://mirrors.cloud.aliyuncs.com/aiext-pypi/aiext-pypi-plugin/simple
Collecting aiext-pypi-plugin
  Downloading http://mirrors.cloud.aliyuncs.com/aiext-pypi/aiext-pypi-plugin/packages/aiext_pypi_plugin-0.1.0-py3-none-any.whl (3.0 kB)
Installing collected packages: aiext-pypi-plugin
Successfully installed aiext-pypi-plugin-0.1.0
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
STEP 11/13: RUN --mount=type=bind,source=/var/run/secrets/ack.alibabacloud.com/rrsa-tokens,target=/mount/rrsa-tokens     pip install cumm-cu114 -i https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu114/simple/
[AIEXT_PYPI_PLUGIN] Plugin has been loaded.
[AIEXT_PYPI_PLUGIN] Missing OIDC token file
Looking in indexes: https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu114/simple/
Requirement already satisfied: cumm-cu114 in /opt/conda/lib/python3.8/site-packages (0.5.3)
Requirement already satisfied: numpy in /opt/conda/lib/python3.8/site-packages (from cumm-cu114) (1.21.1)
Requirement already satisfied: pybind11>=2.6.0 in /opt/conda/lib/python3.8/site-packages (from cumm-cu114) (2.7.0)
Requirement already satisfied: pccm>=0.4.2 in /opt/conda/lib/python3.8/site-packages (from cumm-cu114) (0.4.16)
Requirement already satisfied: fire in /opt/conda/lib/python3.8/site-packages (from cumm-cu114) (0.5.0)
Requirement already satisfied: ccimport>=0.3.1 in /opt/conda/lib/python3.8/site-packages (from pccm>=0.4.2->cumm-cu114) (0.4.4)
Requirement already satisfied: lark>=1.0.0 in /opt/conda/lib/python3.8/site-packages (from pccm>=0.4.2->cumm-cu114) (1.2.2)
Requirement already satisfied: portalocker>=2.3.2 in /opt/conda/lib/python3.8/site-packages (from pccm>=0.4.2->cumm-cu114) (2.6.0)
Requirement already satisfied: six in /opt/conda/lib/python3.8/site-packages (from fire->cumm-cu114) (1.16.0)
Requirement already satisfied: termcolor in /opt/conda/lib/python3.8/site-packages (from fire->cumm-cu114) (2.1.1)
Requirement already satisfied: ninja in /opt/conda/lib/python3.8/site-packages (from ccimport>=0.3.1->pccm>=0.4.2->cumm-cu114) (1.11.1.3)
Requirement already satisfied: requests in /opt/conda/lib/python3.8/site-packages (from ccimport>=0.3.1->pccm>=0.4.2->cumm-cu114) (2.26.0)
Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.8/site-packages (from requests->ccimport>=0.3.1->pccm>=0.4.2->cumm-cu114) (1.26.13)
Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.8/site-packages (from requests->ccimport>=0.3.1->pccm>=0.4.2->cumm-cu114) (3.1)
Requirement already satisfied: charset-normalizer~=2.0.0 in /opt/conda/lib/python3.8/site-packages (from requests->ccimport>=0.3.1->pccm>=0.4.2->cumm-cu114) (2.0.0)
Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.8/site-packages (from requests->ccimport>=0.3.1->pccm>=0.4.2->cumm-cu114) (2021.5.30)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
[AIEXT_PYPI_PLUGIN] Missing OIDC token file
STEP 12/13: RUN --mount=type=bind,source=/var/run/secrets/ack.alibabacloud.com/rrsa-tokens,target=/mount/rrsa-tokens     pip install spconv-cu114 -i https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu114/simple/
[AIEXT_PYPI_PLUGIN] Plugin has been loaded.
[AIEXT_PYPI_PLUGIN] Missing OIDC token file
Looking in indexes: https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu114/simple/
Requirement already satisfied: spconv-cu114 in /opt/conda/lib/python3.8/site-packages (2.3.6)
Requirement already satisfied: fire in /opt/conda/lib/python3.8/site-packages (from spconv-cu114) (0.5.0)
Requirement already satisfied: numpy in /opt/conda/lib/python3.8/site-packages (from spconv-cu114) (1.21.1)
Requirement already satisfied: pybind11>=2.6.0 in /opt/conda/lib/python3.8/site-packages (from spconv-cu114) (2.7.0)
Requirement already satisfied: pccm>=0.4.0 in /opt/conda/lib/python3.8/site-packages (from spconv-cu114) (0.4.16)
Requirement already satisfied: ccimport>=0.4.0 in /opt/conda/lib/python3.8/site-packages (from spconv-cu114) (0.4.4)
Requirement already satisfied: requests in /opt/conda/lib/python3.8/site-packages (from ccimport>=0.4.0->spconv-cu114) (2.26.0)
Requirement already satisfied: ninja in /opt/conda/lib/python3.8/site-packages (from ccimport>=0.4.0->spconv-cu114) (1.11.1.3)
Requirement already satisfied: portalocker>=2.3.2 in /opt/conda/lib/python3.8/site-packages (from pccm>=0.4.0->spconv-cu114) (2.6.0)
Requirement already satisfied: lark>=1.0.0 in /opt/conda/lib/python3.8/site-packages (from pccm>=0.4.0->spconv-cu114) (1.2.2)
Requirement already satisfied: six in /opt/conda/lib/python3.8/site-packages (from fire->spconv-cu114) (1.16.0)
Requirement already satisfied: termcolor in /opt/conda/lib/python3.8/site-packages (from fire->spconv-cu114) (2.1.1)
Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.8/site-packages (from requests->ccimport>=0.4.0->spconv-cu114) (2021.5.30)
Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.8/site-packages (from requests->ccimport>=0.4.0->spconv-cu114) (3.1)
Requirement already satisfied: charset-normalizer~=2.0.0 in /opt/conda/lib/python3.8/site-packages (from requests->ccimport>=0.4.0->spconv-cu114) (2.0.0)
Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.8/site-packages (from requests->ccimport>=0.4.0->spconv-cu114) (1.26.13)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
[AIEXT_PYPI_PLUGIN] Missing OIDC token file
STEP 13/13: RUN pip uninstall -y aiext-pypi-plugin
[AIEXT_PYPI_PLUGIN] Plugin has been loaded.
[AIEXT_PYPI_PLUGIN] Missing OIDC token file
Found existing installation: aiext-pypi-plugin 0.1.0
Uninstalling aiext-pypi-plugin-0.1.0:
  Successfully uninstalled aiext-pypi-plugin-0.1.0
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
COMMIT demo:v1
--> 73a79226bf56
Successfully tagged localhost/demo:v1
73a79226bf5653a8dfed3f01c46dddc6c683f2c410b0e9d8f0ca3f5d6e4ad336

查看保存已构建的镜像。

# 查看构建成功的新镜像
buildah --storage-driver=overlay images

# 之后可以按需推送到自己的镜像仓库，略
buildah login -u "{LOGIN_USERNAME}" -p "{LOGIN_PASSWORD}" {target-registry-domain}
buildah --storage-driver=overlay push test:v1 docker://{target-registry-prefix}/{target-image}

普通PPU容器示例

YAML文件定义了普通容器场景，可以使用VPC PPU镜像。

apiVersion: v1
kind: Pod
metadata:
  name: common-ppu-pod
  labels:
    alibabacloud.com/compute-class: gpu-hpn
    alibabacloud.com/gpu-model-series: PPU810E
    alibabacloud.com/compute-qos: default
    alibabacloud.com/hpn-type: "rdma"
spec:
  imagePullSecrets:          #可选使用secret，若普通容器使用vpc镜像不用secret，则公网镜像必须鉴权。
    - name: acs-image-secret #需要和前提条件配置镜像鉴权的Secret的name一致。
  containers:
    - name: demo
      image: acs-registry-vpc.cn-wulanchabu.cr.aliyuncs.com/egslingjun/training-xpu-pytorch:25.06
      command: 
      - sh
      - -c
      - sleep infinity
      resources:
        limits: 
          cpu: 16
          memory: 128G
          alibabacloud.com/ppu: 2
        requests: #结合上述运行模型的command args按需调整
          cpu: 16
          memory: 128G
          alibabacloud.com/ppu: 2
 
  #添加前提条件创建的ServiceAccount名称
  serviceAccountName: pip-default

创建配置普通PPU容器。

#创建普通PPU容器
kubectl apply -f commonppu.yaml

#连接进入普通PPU容器
kubectl exec -it acs-common-ppu-fuyi01 -- bash

#【普通容器内】首先安装免密插件aiext-pypi-plugin，安装插件限制（暂不支持以python -m pip执行请求）
pip install aiext-pypi-plugin -i http://mirrors.cloud.aliyuncs.com/aiext-pypi/aiext-pypi-plugin/simple --trusted-host mirrors.cloud.aliyuncs.com

#【普通容器内】然后就可以正常免密安装相关pip wheel包，找了一个镜像中不存在的包（torchdata）：
pip install torchdata -i https://aiext-pypi.mirrors.aliyuncs.com/pg1-pip/ubuntu_cu126/simple/