Example with collecting timestamp of the metrics

examples/v1alpha3/bayesianoptimization-example.yaml

@@ -11,7 +11,7 @@ spec:
     goal: 0.99
     objectiveMetricName: Validation-accuracy
     additionalMetricNames:
-      - accuracy
+      - Train-accuracy
   algorithm:
     algorithmName: bayesianoptimization
     algorithmSettings:
@@ -51,10 +51,10 @@ spec:
               spec:
                 containers:
                 - name: {{.Trial}}
-                  image: docker.io/kubeflowkatib/mxnet-mnist-example
+                  image: docker.io/kubeflowkatib/mxnet-mnist
                   command:
-                  - "python"
-                  - "/mxnet/example/image-classification/train_mnist.py"
+                  - "python3"
+                  - "/opt/mxnet-mnist/mnist.py"
                   - "--batch-size=64"
                   {{- with .HyperParameters}}
                   {{- range .}}

examples/v1alpha3/grid-example.yaml

@@ -11,7 +11,7 @@ spec:
     goal: 0.99
     objectiveMetricName: Validation-accuracy
     additionalMetricNames:
-      - accuracy
+      - Train-accuracy
   algorithm:
     algorithmName: grid
   parallelTrialCount: 3
@@ -32,8 +32,8 @@ spec:
     - name: --num-epochs
       parameterType: int
       feasibleSpace:
-        min: "20"
-        max: "40"
+        min: "10"
+        max: "15"
     # Grid doesn't support categorical, refer to https://chocolate.readthedocs.io/api/sample.html#chocolate.Grid
     # - name: --optimizer
     #   parameterType: categorical
@@ -55,10 +55,10 @@ spec:
               spec:
                 containers:
                 - name: {{.Trial}}
-                  image: docker.io/kubeflowkatib/mxnet-mnist-example
+                  image: docker.io/kubeflowkatib/mxnet-mnist
                   command:
-                  - "python"
-                  - "/mxnet/example/image-classification/train_mnist.py"
+                  - "python3"
+                  - "/opt/mxnet-mnist/mnist.py"
                   - "--batch-size=64"
                   {{- with .HyperParameters}}
                   {{- range .}}

examples/v1alpha3/hyperband-example.yaml

@@ -11,7 +11,7 @@ spec:
     goal: 0.99
     objectiveMetricName: Validation-accuracy
     additionalMetricNames:
-      - accuracy
+      - Train-accuracy
   algorithm:
     algorithmName: hyperband
     algorithmSettings:
@@ -58,10 +58,10 @@ spec:
               spec:
                 containers:
                 - name: {{.Trial}}
-                  image: kubeflowkatib/mxnet-mnist-example
+                  image: docker.io/kubeflowkatib/mxnet-mnist
                   command:
-                  - "python"
-                  - "/mxnet/example/image-classification/train_mnist.py"
+                  - "python3"
+                  - "/opt/mxnet-mnist/mnist.py"
                   - "--batch-size=64"
                   {{- with .HyperParameters}}
                   {{- range .}}

examples/v1alpha3/mxnet-mnist/Dockerfile

@@ -0,0 +1,13 @@
+FROM ubuntu:16.04
+
+RUN apt-get update && \
+    apt-get install -y wget python3-dev gcc && \
+    wget https://bootstrap.pypa.io/get-pip.py && \
+    python3 get-pip.py
+
+RUN pip3 install mxnet
+
+ADD . /opt/mxnet-mnist
+WORKDIR /opt/mxnet-mnist
+
+ENTRYPOINT ["python3", "/opt/mxnet-mnist/mnist.py"]

examples/v1alpha3/mxnet-mnist/README.md

@@ -0,0 +1,6 @@
+# Mxnet image classification example
+This is Mxnet image classification training container with recording time of the metrics.
+
+It uses only simple multilayer perceptron network (mlp).
+
+If you want to read more about this example, visit official [incubator-mxnet](https://github.com/apache/incubator-mxnet/tree/master/example/image-classification) github repository.