任务配置

参考断点续训的yaml参数说明，同时需增加和修改yaml配置文件。

...
apiVersion: batch.volcano.sh/v1alpha1
kind: Job
metadata:
  name: mindx-dls-test    # 注意和ConfigMap的name的对应关系
  namespace: vcjob        # 根据实际需要选择合适的namespace(ConfigMap, Job需要保持一致)
  labels:
    ring-controller.atlas: ascend-910  # hccl_controller 根据该标签来区分配置Ascend910和非Ascend910的场景
    fault-scheduling: "grace"
    elastic-scheduling: "on"   # 需添加""号
  annotations:
    minReplicas: "1"
spec:
  minAvailable: 2           # 需要和replicas值相等
...
  maxRetry: 0
...
         lifecycle:  # 使用临终遗言功能需要添加加粗代码
           preStop:
             exec:
               command: ["/bin/bash", "-c", "cd /job/code/resnet/scripts; bash pre_stop.sh"]
          resources:
            requests:
              huawei.com/Ascend910: 8                                                # 服务器间分布式训练设置为8，单机训练时根据任务进行调整
            limits:
              huawei.com/Ascend910: 8                                                # 数值与请求数量保持一致
...