Experiments/katib_minikube

.gitignore

@@ -131,5 +131,6 @@ dmypy.json
 exp__*
 experiments/simple_raytune/benchmark__RaytuneBenchmark
 experiments/optuna_minikube/benchmark__OptunaMinikubeBenchmark
+experiments/katib_minikube/benchmark__KatibBenchmark
 
 data/

experiments/katib_minikube/experiment_template.yaml

@@ -0,0 +1,52 @@
+---
+apiVersion: kubeflow.org/v1beta1
+kind: Experiment
+metadata:
+  namespace: kubeflow
+  name: $study_name
+spec:
+  objective:
+    type: maximize
+    goal: 0.8
+    objectiveMetricName: precision
+  algorithm:
+    algorithmName: grid
+  parallelTrialCount: $worker_num
+  maxTrialCount: $jobs_num
+  maxFailedTrialCount: $worker_num
+  parameters:
+    - name: lr
+      parameterType: double
+      feasibleSpace:
+        min: "0.001"
+        max: "0.02"
+        step: "0.001"
+  trialTemplate:
+    primaryContainerName: training-container
+    trialParameters:
+      - name: learningRate
+        description: Learning rate for the training model
+        reference: lr
+    trialSpec:
+      apiVersion: batch/v1
+      kind: Job
+      spec:
+        template:
+          spec:
+            containers:
+              - name: training-container
+                image: $worker_image
+                imagePullPolicy: IfNotPresent
+                command:
+                  - "python3"
+                  - "experiments/katib_minikube/$worker_image/task.py"
+                  - "--batch-size=64"
+                  - "--lr=$trialParameters"
+                env:
+                  - name: STUDY_NAME
+                    value: "$study_name"
+                  - name: DB_CONN
+                    value: "postgresql://postgresadmin:admin123@postgres:5432/postgresdb"
+                  - name: "METRICS_STORAGE_HOST"
+                    value: "$metrics_ip"
+            restartPolicy: Never

experiments/katib_minikube/grid.yaml

@@ -0,0 +1,52 @@
+---
+apiVersion: kubeflow.org/v1beta1
+kind: Experiment
+metadata:
+  namespace: kubeflow
+  name: katib-study-32
+spec:
+  objective:
+    type: maximize
+    goal: 0.8
+    objectiveMetricName: precision
+  algorithm:
+    algorithmName: grid
+  parallelTrialCount: 2
+  maxTrialCount: 2
+  maxFailedTrialCount: 2
+  parameters:
+    - name: lr
+      parameterType: double
+      feasibleSpace:
+        min: "0.001"
+        max: "0.02"
+        step: "0.001"
+  trialTemplate:
+    primaryContainerName: training-container
+    trialParameters:
+      - name: learningRate
+        description: Learning rate for the training model
+        reference: lr
+    trialSpec:
+      apiVersion: batch/v1
+      kind: Job
+      spec:
+        template:
+          spec:
+            containers:
+              - name: training-container
+                image: mnist_task
+                imagePullPolicy: IfNotPresent
+                command:
+                  - "python3"
+                  - "experiments/katib_minikube/mnist_task/task.py"
+                  - "--batch-size=64"
+                  - "--lr=${trialParameters.learningRate}"
+                env:
+                  - name: STUDY_NAME
+                    value: "katib-study-32"
+                  - name: DB_CONN
+                    value: "postgresql://postgresadmin:admin123@postgres:5432/postgresdb"
+                  - name: "METRICS_STORAGE_HOST"
+                    value: "134.101.4.161"
+            restartPolicy: Never