config_adl.yml

authorName: default
experimentName: example_pytorch_cifar10
trialConcurrency: 1
maxExecDuration: 100h
maxTrialNum: 10
nniManagerIp: {replace_with_your_ip}
trainingServicePlatform: adl
searchSpacePath: search_space_adl.json
logCollection: http
#choice: true, false
useAnnotation: false
tuner:
  #choice: TPE, Random, Anneal, Evolution, BatchTuner, MetisTuner
  #SMAC (SMAC should be installed through nnictl)
  builtinTunerName: TPE
  classArgs:
    #choice: maximize, minimize
    optimize_mode: maximize
trial:
  namespace: default
  command: python3 /cifar10/main_adl.py
  codeDir: /cifar10
  gpuNum: 1
  # the user needs to have a docker image built by the adl.Dockerfile
  # the docker image should be pushed to a registry for the cluster to pull
  # in our example we provide a docker image from our public docker hub
  image: petuum/nni:cifar-example 
  # optional: 
  # the user needs to provide the secret if the image is pulled from a private registry
  # imagePullSecrets:
  #  - name: {secret}
  adaptive: true
  checkpoint:
    # the user needs to determine the storageClass in Kubenetes
    # For example, for MicroK8s, ‘microk8s-hostpath’ can be used
    storageClass: microk8s-hostpath 
    storageSize: 1Gi
  cpuNum: 1
  memorySize: 1Gi