feat: Add Hyperparameter tuning Workflow Template

db/go/20201225172926_add_hyperparameter_tuning_example_workflow_template.go

@@ -0,0 +1,36 @@
+package migration
+
+import (
+	"database/sql"
+	"github.com/pressly/goose"
+	"path/filepath"
+)
+
+const hyperparameterTuningTemplateName = "Hyperparameter Tuning Example"
+
+func initialize20201225172926() {
+	if _, ok := initializedMigrations[20201225172926]; !ok {
+		goose.AddMigration(Up20201225172926, Down20201225172926)
+		initializedMigrations[20201225172926] = true
+	}
+}
+
+// Up20201225172926 adds Hyperparameter Tuning Workflow Template
+func Up20201225172926(tx *sql.Tx) error {
+	// This code is executed when the migration is applied.
+	return createWorkflowTemplate(
+		filepath.Join("workflows", "hyperparam_tuning", "20201225172926.yaml"),
+		hyperparameterTuningTemplateName,
+		map[string]string{
+			"framework":  "pytorch",
+			"tuner":      "TPE",
+			"created-by": "system",
+		},
+	)
+}
+
+// Down20201225172926 archives Hyperparameter Tuning Workflow Template
+func Down20201225172926(tx *sql.Tx) error {
+	// This code is executed when the migration is rolled back.
+	return archiveWorkflowTemplate(hyperparameterTuningTemplateName)
+}

db/go/db.go

@@ -83,6 +83,7 @@ func Initialize() {
 	initialize20201221195937()
 	initialize20201223062947()
 	initialize20201223202929()
+	initialize20201225172926()
 	initialize20201229205644()
 
 	if err := client.DB.Close(); err != nil {

db/go/util.go

@@ -48,6 +48,51 @@ func updateWorkspaceTemplateManifest(filename, templateName string) error {
 	return nil
 }
 
+// createWorkflowTemplate will create the workflow template given by {{templateName}} with the contents
+// given by {{filename}}
+// It will do so for all namespaces.
+func createWorkflowTemplate(filename, templateName string, labels map[string]string) error {
+	client, err := getClient()
+	if err != nil {
+		return err
+	}
+	defer client.DB.Close()
+
+	namespaces, err := client.ListOnepanelEnabledNamespaces()
+	if err != nil {
+		return err
+	}
+
+	manifest, err := readDataFile(filename)
+	if err != nil {
+		return err
+	}
+
+	uid, err := uid2.GenerateUID(templateName, 30)
+	if err != nil {
+		return err
+	}
+
+	for _, namespace := range namespaces {
+		workflowTemplate := &v1.WorkflowTemplate{
+			UID:      uid,
+			Name:     templateName,
+			Manifest: manifest,
+			Labels:   labels,
+		}
+
+		workflowTemplate.Manifest, err = ReplaceRuntimeVariablesInManifest(client, namespace.Name, workflowTemplate.Manifest)
+		if err != nil {
+			return err
+		}
+		if _, err := client.CreateWorkflowTemplate(namespace.Name, workflowTemplate); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
 // updateWorkflowTemplateManifest will update the workflow template given by {{templateName}} with the contents
 // given by {{filename}}
 // It will do so for all namespaces.
@@ -93,3 +138,30 @@ func updateWorkflowTemplateManifest(filename, templateName string, labels map[st
 
 	return nil
 }
+
+// archiveWorkflowTemplate removes a Workflow Template by a given templateName
+func archiveWorkflowTemplate(templateName string) error {
+	client, err := getClient()
+	if err != nil {
+		return err
+	}
+	defer client.DB.Close()
+
+	namespaces, err := client.ListOnepanelEnabledNamespaces()
+	if err != nil {
+		return err
+	}
+
+	uid, err := uid2.GenerateUID(templateName, 30)
+	if err != nil {
+		return err
+	}
+
+	for _, namespace := range namespaces {
+		if _, err := client.ArchiveWorkflowTemplate(namespace.Name, uid); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}

db/yaml/workflows/hyperparameter-tuning/20201225172926.yaml

@@ -0,0 +1,185 @@
+# source: https://github.com/onepanelio/templates/blob/master/workflows/nni-hyperparameter-tuning/mnist/
+entrypoint: main
+arguments:
+  parameters:
+    - name: source
+      value: https://github.com/onepanelio/templates
+    - name: revision
+      value: master
+    - name: config
+      displayName: Configuration
+      required: true
+      hint: NNI configuration
+      type: textarea.textarea
+      value: |-
+        authorName: Onepanel, Inc.
+        experimentName: MNIST TF v2.x
+        trialConcurrency: 1
+        maxExecDuration: 1h
+        maxTrialNum: 10
+        trainingServicePlatform: local
+        searchSpacePath: search_space.json
+        useAnnotation: false
+        tuner:
+          # gpuIndices: '0'           # uncomment and update to the GPU indices to assign this tuner
+          builtinTunerName: TPE       # choices: TPE, Random, Anneal, Evolution, BatchTuner, MetisTuner, GPTuner
+          classArgs:
+            optimize_mode: maximize   # choices: maximize, minimize
+        trial:
+          command: python main.py --output /mnt/output
+          codeDir: .
+          # gpuNum: 1                 # uncomment and update to number of GPUs
+    - name: search-space
+      displayName: Search space configuration
+      required: true
+      type: textarea.textarea
+      value: |-
+        {
+          "dropout_rate": { "_type": "uniform", "_value": [0.5, 0.9] },
+          "conv_size": { "_type": "choice", "_value": [2, 3, 5, 7] },
+          "hidden_size": { "_type": "choice", "_value": [124, 512, 1024] },
+          "batch_size": { "_type": "choice", "_value": [16, 32] },
+          "learning_rate": { "_type": "choice", "_value": [0.0001, 0.001, 0.01, 0.1] },
+          "epochs": { "_type": "choice", "_value": [10] }
+        }
+    - displayName: Node pool
+      hint: Name of node pool or group to run this workflow task
+      type: select.nodepool
+      name: sys-node-pool
+      value: {{.DefaultNodePoolOption}}
+      required: true
+
+volumeClaimTemplates:
+  - metadata:
+      name: hyperparamtuning-data
+    spec:
+      accessModes: [ "ReadWriteOnce" ]
+      resources:
+        requests:
+          storage: 20Gi
+  - metadata:
+      name: hyperparamtuning-output
+    spec:
+      accessModes: [ "ReadWriteOnce" ]
+      resources:
+        requests:
+          storage: 20Gi
+
+templates:
+  - name: main
+    dag:
+      tasks:
+        - name: hyperparameter-tuning
+          template: hyperparameter-tuning
+        - name: workflow-metrics-writer
+          template: workflow-metrics-writer
+          dependencies: [hyperparameter-tuning]
+          arguments:
+            # Use sys-metrics artifact output from hyperparameter-tuning Task
+            artifacts:
+              - name: best-metrics
+                from: "{{tasks.hyperparameter-tuning.outputs.artifacts.sys-metrics}}"
+  - name: hyperparameter-tuning
+    inputs:
+      artifacts:
+        - name: src
+          git:
+            repo: '{{workflow.parameters.source}}'
+            revision: '{{workflow.parameters.revision}}'
+          path: /mnt/data/src
+        - name: config
+          path: /mnt/data/src/workflows/hyperparameter-tuning/mnist/config.yaml
+          raw:
+            data: '{{workflow.parameters.config}}'
+        - name: search-space
+          path: /mnt/data/src/workflows/hyperparameter-tuning/mnist/search_space.json
+          raw:
+            data: '{{workflow.parameters.search-space}}'
+    outputs:
+      artifacts:
+        - name: output
+          path: /mnt/output
+          optional: true
+          archive:
+            none: {}
+    container:
+      image: onepanel/dl:0.17.0
+      args:
+        - --config
+        - /mnt/data/src/workflows/hyperparameter-tuning/mnist/config.yaml
+      workingDir: /mnt
+      volumeMounts:
+        - name: hyperparamtuning-data
+          mountPath: /mnt/data
+        - name: hyperparamtuning-output
+          mountPath: /mnt/output
+    nodeSelector:
+      beta.kubernetes.io/instance-type: '{{workflow.parameters.sys-node-pool}}'
+    sidecars:
+      - name: nni-web-ui
+        image: 'onepanel/nni-web-ui:0.17.0'
+        env:
+          - name: ONEPANEL_INTERACTIVE_SIDECAR
+            value: 'true'
+        ports:
+          - containerPort: 9000
+            name: nni
+      - name: tensorboard
+        image: 'tensorflow/tensorflow:2.3.0'
+        command:
+          - sh
+          - '-c'
+        env:
+          - name: ONEPANEL_INTERACTIVE_SIDECAR
+            value: 'true'
+        args:
+          # Read logs from /mnt/output/tensorboard - /mnt/output is auto-mounted from volumeMounts
+          - tensorboard --logdir /mnt/output/tensorboard
+        ports:
+          - containerPort: 6006
+            name: tensorboard
+  - name: workflow-metrics-writer
+    inputs:
+      artifacts:
+        - name: best-metrics
+          path: /tmp/sys-metrics.json
+    script:
+      image: onepanel/python-sdk:v0.16.0
+      command: [python, '-u']
+      source: |
+        import os
+        import json
+
+        import onepanel.core.api
+        from onepanel.core.api.models.metric import Metric
+        from onepanel.core.api.rest import ApiException
+        from onepanel.core.api.models import Parameter
+
+        # Load Task A metrics
+        with open('/tmp/sys-metrics.json') as f:
+            metrics = json.load(f)
+
+        with open('/var/run/secrets/kubernetes.io/serviceaccount/token') as f:
+            token = f.read()
+
+        # Configure API authorization
+        configuration = onepanel.core.api.Configuration(
+            host = os.getenv('ONEPANEL_API_URL'),
+            api_key = {
+                'authorization': token
+            }
+        )
+        configuration.api_key_prefix['authorization'] = 'Bearer'
+
+        # Call SDK method to save metrics
+        with onepanel.core.api.ApiClient(configuration) as api_client:
+            api_instance = onepanel.core.api.WorkflowServiceApi(api_client)
+            namespace = '{{workflow.namespace}}'
+            uid = '{{workflow.name}}'
+            body = onepanel.core.api.AddWorkflowExecutionsMetricsRequest()
+            body.metrics = metrics
+            try:
+                api_response = api_instance.add_workflow_execution_metrics(namespace, uid, body)
+                print('Metrics added.')
+            except ApiException as e:
+                print("Exception when calling WorkflowServiceApi->add_workflow_execution_metrics: %s\n" % e)