MindSpore on volcano

最近更新于 Jul 31, 2021

Mindspore简介

MindSpore是华为公司推出的新一代深度学习框架，是源于全产业的最佳实践，最佳匹配昇腾处理器算力，支持终端、边缘、云全场景灵活部署，开创全新的AI编程范式，降低AI开发门槛。

MindSpore on volcano

在集群中新建mindspore-cpu.yaml如下

  1. apiVersion: batch.volcano.sh/v1alpha1
  2. kind: Job
  3. metadata:
  4.   name: mindspore-cpu
  5. spec:
  6.   minAvailable: 1
  7.   schedulerName: volcano
  8.   policies:
  9.     - event: PodEvicted
  10.       action: RestartJob
  11.   plugins:
  12.     ssh: []
  13.     env: []
  14.     svc: []
  15.   maxRetry: 5
  16.   queue: default
  17.   tasks:
  18.     - replicas: 8
  19.       name: "pod"
  20.       template:
  21.         spec:
  22.           containers:
  23.             - command: ["/bin/bash", "-c", "python /tmp/lenet.py"]
  24.               image: lyd911/mindspore-cpu-example:0.2.0
  25.               imagePullPolicy: IfNotPresent
  26.               name: mindspore-cpu-job
  27.               resources:
  28.                 limits:
  29.                   cpu: "1"
  30.                 requests:
  31.                   cpu: "1"
  32.           restartPolicy: OnFailure

进行部署。

  1. kubectl apply -f mindspore-cpu.yaml

查询集群下作业运行情况。

  1. kubectl get pods