Add an example of Goptuna suggestion service

[c-bata]

What this PR does / why we need it:
Add an example of Goptuna suggestion service.

Which issue(s) this PR fixes (optional, in fixes #<issue number>(, fixes #<issue_number>, ...) format, will close the issue(s) when PR gets merged):
Part of #1152

Special notes for your reviewer:

1. This experiment is almost the same with tpe-example. I just change the name and algorithm name.
2. Please note that Goptuna's CMA-ES does not support categorical search space. So --optimizer param is sampled by Goptuna independent sampler (default: random sampler).

Release note:

NONE

[kubeflow-bot]

This change is 

[c-bata]

I'm in two minds which example should I add, mxnet-mnist or Himmelblau-function.

Himmelblau function

https://en.wikipedia.org/wiki/Himmelblau%27s_function

This function is quite simple and we can evaluate so fast. But it's not a realistic problem and we need to add CI setting to build docker image.

import argparse
import logging

logging.basicConfig(filename='/var/log/katib.log', level=logging.DEBUG)


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("--x1", dest='x1', type=float)
    parser.add_argument("--x2", dest='x2', type=float)

    args = parser.parse_args()
    print(args.x1, args.x2)

    evaluation = (args.x1 ** 2 + args.x2 - 11.0) ** 2 + (args.x1 + args.x2 ** 2 - 7.0) ** 2

    metric_msg = f'{{metricName: evaluation, metricValue: {evaluation:.4f}}};'
    logging.info(metric_msg)
    print(metric_msg)


if __name__ == '__main__':
    main()

FROM python:3-alpine

ADD main.py /usr/src/main.py
WORKDIR /usr/src

experiment

apiVersion: "kubeflow.org/v1alpha3"
kind: Experiment
metadata:
  namespace: kubeflow
  labels:
    controller-tools.k8s.io: "1.0"
  name: cmaes-example
spec:
  objective:
    type: minimize
    goal: 0.0001
    objectiveMetricName: evaluation
  algorithm:
    algorithmName: cmaes
  metricsCollectorSpec:
    source:
      filter:
        metricsFormat:
          - "{metricName: ([\\w|-]+), metricValue: ((-?\\d+)(\\.\\d+)?)}"
      fileSystemPath:
        path: "/var/log/katib.log"
        kind: File
    collector:
      kind: File
  parallelTrialCount: 2
  maxTrialCount: 250
  maxFailedTrialCount: 3
  parameters:
    - name: x1
      parameterType: double
      feasibleSpace:
        min: "-10"
        max: "10"
    - name: x2
      parameterType: iint
      feasibleSpace:
        min: "-10"
        max: "10"
        step: "2"
  trialTemplate:
    goTemplate:
        rawTemplate: |-
          apiVersion: batch/v1
          kind: Job
          metadata:
            name: {{.Trial}}
            namespace: {{.NameSpace}}
          spec:
            template:
              spec:
                containers:
                - name: {{.Trial}}
                  image: docker.io/cbata/hello-katib-quadratic-function
                  command:
                  - "python3"
                  - "main.py"
                  {{- with .HyperParameters}}
                  {{- range .}}
                  - "--{{.Name}} {{.Value}}"
                  {{- end}}
                  {{- end}}
                restartPolicy: Never

[c-bata]

/assign @andreyvelich

[gaocegege]

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: gaocegege

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [gaocegege]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[andreyvelich]

@c-bata I think, let's keep mxnet for now. If we decide that users needs another example, we can submit Himmelblau function training container.

[c-bata]

I see. Thanks!