update kfp 'github issue summarization' example (kubeflow#823)

* checkpointing

* more updates to keep gh summ pipelines example current
cleanup & update; remove obsolete pipelines
create 'preemptible' version of hosted kfp pipeline
notebook update, readme update

* in notebook, add kernel restart after pip install
minor pipeline cleanup
add archive version of pipeline

* fixed namespace glitch, cleaned up css positioning issue

github_issue_summarization/pipelines/README.md

@@ -4,6 +4,8 @@
 This Kubeflow Pipelines example shows how to build a web app that summarizes GitHub issues using Kubeflow Pipelines to train and serve a model.
 The pipeline trains a [Tensor2Tensor](https://github.com/tensorflow/tensor2tensor/) model on GitHub issue data, learning to predict issue titles from issue bodies. It then exports the trained model and deploys the exported model using  [Tensorflow Serving](https://github.com/tensorflow/serving). The final step in the pipeline launches a web app, which interacts with the TF-Serving instance in order to get model predictions.
 
+The example is designed to run on a Hosted KFP installation, installed via the [Cloud Console](https://console.cloud.google.com/ai-platform/pipelines/clusters) or via ['standalone' installation](https://www.kubeflow.org/docs/pipelines/installation/standalone-deployment/) instructions, but would also be straightforward to run on a Kubeflow installation with minor changes.
+
 You can follow this example as a codelab: [g.co/codelabs/kfp-gis](https://g.co/codelabs/kfp-gis).
 
 <!-- Or, you can run it as a [Cloud shell Tutorial](https://console.cloud.google.com/?cloudshell=true&cloudshell_git_repo=https://github.com/kubeflow/examples&working_dir=github_issue_summarization/pipelines&cloudshell_tutorial=tutorial.md).  The source for the Cloud Shell tutorial is [here](tutorial.md). -->

github_issue_summarization/pipelines/components/kubeflow-resources/containers/tf-serving-gh/Dockerfile

@@ -12,15 +12,28 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-FROM ubuntu:16.04
+FROM tensorflow/tensorflow:1.15.0-py3
 
 RUN apt-get update -y
 
 RUN apt-get install --no-install-recommends -y -q ca-certificates python-dev python-setuptools wget unzip
 
 RUN easy_install pip
 
-RUN pip install pyyaml==3.12 six==1.11.0 requests==2.18.4 tensorflow==1.11.0
+RUN pip install pyyaml==3.12 six requests==2.18.4
+
+# RUN apt-get update \
+#   && apt-get install -y python3-pip python3-dev wget unzip \
+#   && cd /usr/local/bin \
+#   && ln -s /usr/bin/python3 python \
+#   && pip3 install --upgrade pip
+
+# RUN apt-get install -y wget unzip git
+
+# RUN pip install --upgrade pip
+# RUN pip install urllib3 certifi retrying
+# RUN pip install google-cloud-storage
+
 
 RUN wget -nv https://dl.google.com/dl/cloudsdk/release/google-cloud-sdk.zip && \
     unzip -qq google-cloud-sdk.zip -d tools && \
@@ -34,16 +47,9 @@ RUN wget -nv https://dl.google.com/dl/cloudsdk/release/google-cloud-sdk.zip && \
     tools/google-cloud-sdk/bin/gcloud config set component_manager/disable_update_check true && \
     touch /tools/google-cloud-sdk/lib/third_party/google.py
 
-RUN wget -nv https://github.com/ksonnet/ksonnet/releases/download/v0.11.0/ks_0.11.0_linux_amd64.tar.gz && \
-    tar -xvzf ks_0.11.0_linux_amd64.tar.gz && \
-    mkdir -p /tools/ks/bin && \
-    cp ./ks_0.11.0_linux_amd64/ks /tools/ks/bin && \
-    rm ks_0.11.0_linux_amd64.tar.gz && \
-    rm -r ks_0.11.0_linux_amd64
 
-ENV PATH $PATH:/tools/google-cloud-sdk/bin:/tools/ks/bin
+ENV PATH $PATH:/tools/google-cloud-sdk/bin
 
 ADD build /ml
 
 ENTRYPOINT ["python", "/ml/deploy-tf-serve.py"]
-

github_issue_summarization/pipelines/components/kubeflow-resources/tf-serving-gh/deploy-tf-serve.py

@@ -40,9 +40,10 @@ def main():
                            'If not set, assuming this runs in a GKE container and current ' +
                            'cluster is used.')
   parser.add_argument('--zone', type=str, help='zone of the kubeflow cluster.')
+  parser.add_argument('--namespace', type=str, default='default')
   args = parser.parse_args()
 
-  KUBEFLOW_NAMESPACE = 'kubeflow'
+  # KUBEFLOW_NAMESPACE = 'kubeflow'
 
   # Make sure model dir exists before proceeding
   retries = 0
@@ -90,7 +91,7 @@ def main():
     with open(target_file, "w") as target:
       data = f.read()
       changed = data.replace('MODEL_NAME', args.model_name)
-      changed1 = changed.replace('KUBEFLOW_NAMESPACE', KUBEFLOW_NAMESPACE)
+      changed1 = changed.replace('KUBEFLOW_NAMESPACE', args.namespace)
       changed2 = changed1.replace('MODEL_PATH', args.model_path)
       target.write(changed2)
 

github_issue_summarization/pipelines/components/kubeflow-resources/tf-serving-gh/tf-serve-template.yaml

@@ -28,7 +28,7 @@ spec:
     app: MODEL_NAME
   type: ClusterIP
 ---
-apiVersion: extensions/v1beta1
+apiVersion: apps/v1
 kind: Deployment
 metadata:
   labels:
@@ -37,16 +37,19 @@ metadata:
   namespace: KUBEFLOW_NAMESPACE
 spec:
   replicas: 1
+  selector:
+    matchLabels:
+      app: MODEL_NAME
   template:
     metadata:
       labels:
         app: MODEL_NAME
         version: v1
     spec:
-      volumes:
-      - name: gcp-credentials-user-gcp-sa
-        secret:
-          secretName: user-gcp-sa
+      # volumes:
+      # - name: gcp-credentials-user-gcp-sa
+      #   secret:
+      #     secretName: user-gcp-sa
       containers:
       - args:
         - --port=9000
@@ -56,15 +59,15 @@ spec:
         command:
         - /usr/bin/tensorflow_model_server
         image: tensorflow/serving
-        env:
-        - name: GOOGLE_APPLICATION_CREDENTIALS
-          value: /secret/gcp-credentials/user-gcp-sa.json
-        - name: CLOUDSDK_AUTH_CREDENTIAL_FILE_OVERRIDE
-          value: /secret/gcp-credentials/user-gcp-sa.json
-        volumeMounts:
-        - mountPath: /secret/gcp-credentials
-          name: gcp-credentials-user-gcp-sa
-        imagePullPolicy: IfNotPresent
+        # env:
+        # - name: GOOGLE_APPLICATION_CREDENTIALS
+        #   value: /secret/gcp-credentials/user-gcp-sa.json
+        # - name: CLOUDSDK_AUTH_CREDENTIAL_FILE_OVERRIDE
+        #   value: /secret/gcp-credentials/user-gcp-sa.json
+        # volumeMounts:
+        # - mountPath: /secret/gcp-credentials
+        #   name: gcp-credentials-user-gcp-sa
+        imagePullPolicy: Always
         livenessProbe:
           initialDelaySeconds: 30
           periodSeconds: 30

github_issue_summarization/pipelines/components/t2t/containers/t2t_app/Dockerfile

@@ -21,9 +21,9 @@ RUN apt-get install --no-install-recommends -y -q ca-certificates python-dev pyt
 
 RUN easy_install pip
 
-RUN pip install tensorflow-probability==0.5
+RUN pip install tensorflow-probability==0.7
 RUN pip install tensor2tensor==1.11.0
-RUN pip install tensorflow-serving-api
+RUN pip install tensorflow-serving-api==1.14.0
 RUN pip install gunicorn
 RUN pip install pyyaml==3.12 six==1.11.0
 RUN pip install pandas

github_issue_summarization/pipelines/components/t2t/containers/webapp-launcher/Dockerfile

@@ -12,15 +12,20 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-FROM ubuntu:16.04
+FROM ubuntu:18.04
 
-RUN apt-get update -y
+RUN apt-get update \
+  && apt-get install -y python3-pip python3-dev wget unzip \
+  && cd /usr/local/bin \
+  && ln -s /usr/bin/python3 python \
+  && pip3 install --upgrade pip
 
-RUN apt-get install --no-install-recommends -y -q ca-certificates python-dev python-setuptools wget unzip
+# RUN apt-get install -y wget unzip git
 
-RUN easy_install pip
+RUN pip install --upgrade pip
+RUN pip install urllib3 certifi retrying
+RUN pip install google-cloud-storage
 
-RUN pip install pyyaml==3.12 six==1.11.0 requests==2.18.4 tensorflow==1.12.0
 
 RUN wget -nv https://dl.google.com/dl/cloudsdk/release/google-cloud-sdk.zip && \
     unzip -qq google-cloud-sdk.zip -d tools && \
@@ -34,14 +39,8 @@ RUN wget -nv https://dl.google.com/dl/cloudsdk/release/google-cloud-sdk.zip && \
     tools/google-cloud-sdk/bin/gcloud config set component_manager/disable_update_check true && \
     touch /tools/google-cloud-sdk/lib/third_party/google.py
 
-RUN wget -nv https://github.com/ksonnet/ksonnet/releases/download/v0.11.0/ks_0.11.0_linux_amd64.tar.gz && \
-    tar -xvzf ks_0.11.0_linux_amd64.tar.gz && \
-    mkdir -p /tools/ks/bin && \
-    cp ./ks_0.11.0_linux_amd64/ks /tools/ks/bin && \
-    rm ks_0.11.0_linux_amd64.tar.gz && \
-    rm -r ks_0.11.0_linux_amd64
 
-ENV PATH $PATH:/tools/google-cloud-sdk/bin:/tools/ks/bin
+ENV PATH $PATH:/tools/google-cloud-sdk/bin
 
 ADD build /ml
 

github_issue_summarization/pipelines/components/t2t/t2t-app/app/templates/index.html

@@ -12,8 +12,8 @@
   </head>
   <body class="text-center">
     <form class="form-signin" action="summary" method="post">
-      <img class="mb-4" src="https://assets-cdn.github.com/images/modules/logos_page/GitHub-Mark.png" alt="" width="72" height="72">
-      <h1 style="margin-top: -20px;">Github Issue Summarization</h1>
+      <!-- <img class="mb-4" src="https://assets-cdn.github.com/images/modules/logos_page/GitHub-Mark.png" alt="" width="72" height="72"> -->
+      <h1 style="margin-top: 20px;">Github Issue Summarization</h1>
       <p style="margin-left: 20%; margin-right: 20%;">This app takes as input a Github issue body and predicts a title for it. Behind the scenes it uses a
         <a href="https://github.com/tensorflow/tensor2tensor" target="_blank">Tensor2Tensor</a> TensorFlow model, served via <a href="https://github.com/tensorflow/serving/" target="_blank">TF-Serving </a>.</p>
       <p>(Thanks to <a href="https://github.com/hamelsmu" target="_blank">Hamel Husain</a> for the original concept and source data.)</p>

github_issue_summarization/pipelines/components/t2t/webapp-launcher/deploy-webapp.py

@@ -36,11 +36,12 @@ def main():
                            'If not set, assuming this runs in a GKE container and current ' +
                            'cluster is used.')
   parser.add_argument('--zone', type=str, help='zone of the kubeflow cluster.')
+  parser.add_argument('--namespace', type=str, default='default')
   args = parser.parse_args()
 
-  KUBEFLOW_NAMESPACE = 'kubeflow'
+  # KUBEFLOW_NAMESPACE = 'kubeflow'
 
-  print("using model name: %s and namespace: %s" % (args.model_name, KUBEFLOW_NAMESPACE))
+  print("using model name: %s and namespace: %s" % (args.model_name, args.namespace))
 
   logging.getLogger().setLevel(logging.INFO)
   args_dict = vars(args)
@@ -70,7 +71,7 @@ def main():
     with open(target_file, "w") as target:
       data = f.read()
       changed = data.replace('MODEL_NAME', args.model_name)
-      changed1 = changed.replace('KUBEFLOW_NAMESPACE', KUBEFLOW_NAMESPACE).replace(
+      changed1 = changed.replace('KUBEFLOW_NAMESPACE', args.namespace).replace(
         'GITHUB_TOKEN', args.github_token).replace(
         'DATA_DIR', 'gs://aju-dev-demos-codelabs/kubecon/t2t_data_gh_all/')
       target.write(changed1)

github_issue_summarization/pipelines/components/t2t/webapp-launcher/t2tapp-template.yaml

@@ -1,27 +1,3 @@
-apiVersion: networking.istio.io/v1alpha3
-kind: VirtualService
-metadata:
-  name: webapp-MODEL_NAME
-spec:
-  gateways:
-  - kubeflow-gateway
-  hosts:
-  - '*'
-  http:
-  - match:
-    - uri:
-        prefix: /webapp/
-    rewrite:
-      uri: /
-    route:
-    - destination:
-        host: MODEL_NAME-webappsvc.KUBEFLOW_NAMESPACE.svc.cluster.local
-        port:
-          number: 80
-    timeout: 300s
-
----
-
 apiVersion: v1
 kind: Service
 metadata:
@@ -50,33 +26,36 @@ spec:
 
 ---
 
-apiVersion: extensions/v1beta1
+apiVersion: apps/v1
 kind: Deployment
 metadata:
   name: MODEL_NAME-webapp
 spec:
   replicas: 1
+  selector:
+    matchLabels:
+      app: ghsumm
   template:
     metadata:
       labels:
         app: ghsumm
         role: frontend
     spec:
-      volumes:
-      - name: gcp-credentials-user-gcp-sa
-        secret:
-          secretName: user-gcp-sa
+      # volumes:
+      # - name: gcp-credentials-user-gcp-sa
+      #   secret:
+      #     secretName: user-gcp-sa
       containers:
       - name: MODEL_NAME-webapp
-        image: gcr.io/google-samples/ml-pipeline-t2tapp:v3ap
-        env:
-        - name: GOOGLE_APPLICATION_CREDENTIALS
-          value: /secret/gcp-credentials/user-gcp-sa.json
-        - name: CLOUDSDK_AUTH_CREDENTIAL_FILE_OVERRIDE
-          value: /secret/gcp-credentials/user-gcp-sa.json
-        volumeMounts:
-        - mountPath: /secret/gcp-credentials
-          name: gcp-credentials-user-gcp-sa
+        image: gcr.io/google-samples/ml-pipeline-t2tapp:vap9
+        # env:
+        # - name: GOOGLE_APPLICATION_CREDENTIALS
+        #   value: /secret/gcp-credentials/user-gcp-sa.json
+        # - name: CLOUDSDK_AUTH_CREDENTIAL_FILE_OVERRIDE
+        #   value: /secret/gcp-credentials/user-gcp-sa.json
+        # volumeMounts:
+        # - mountPath: /secret/gcp-credentials
+        #   name: gcp-credentials-user-gcp-sa
         # resources:
         #   limits:
         #     nvidia.com/gpu: 1

github_issue_summarization/pipelines/example_pipelines/gh_summ_hosted_kfp.py

@@ -0,0 +1,88 @@
+# Copyright 2019 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import kfp.dsl as dsl
+import kfp.gcp as gcp
+import kfp.components as comp
+
+
+COPY_ACTION = 'copy_data'
+TRAIN_ACTION = 'train'
+DATASET = 'dataset'
+MODEL = 'model'
+
+copydata_op = comp.load_component_from_url(
+  'https://raw.githubusercontent.com/kubeflow/examples/master/github_issue_summarization/pipelines/components/t2t/datacopy_component.yaml'  # pylint: disable=line-too-long
+  )
+
+train_op = comp.load_component_from_url(
+  'https://raw.githubusercontent.com/kubeflow/examples/master/github_issue_summarization/pipelines/components/t2t/train_component.yaml' # pylint: disable=line-too-long
+  )
+
+
+@dsl.pipeline(
+  name='Github issue summarization',
+  description='Demonstrate Tensor2Tensor-based training and TF-Serving'
+)
+def gh_summ(  #pylint: disable=unused-argument
+  train_steps: 'Integer' = 2019300,
+  project: str = 'YOUR_PROJECT_HERE',
+  github_token: str = 'YOUR_GITHUB_TOKEN_HERE',
+  working_dir: 'GCSPath' = 'gs://YOUR_GCS_DIR_HERE',
+  checkpoint_dir: 'GCSPath' = 'gs://aju-dev-demos-codelabs/kubecon/model_output_tbase.bak2019000/',
+  deploy_webapp: str = 'true',
+  data_dir: 'GCSPath' = 'gs://aju-dev-demos-codelabs/kubecon/t2t_data_gh_all/'
+  ):
+
+
+  copydata = copydata_op(
+    data_dir=data_dir,
+    checkpoint_dir=checkpoint_dir,
+    model_dir='%s/%s/model_output' % (working_dir, dsl.RUN_ID_PLACEHOLDER),
+    action=COPY_ACTION,
+    )
+
+
+  train = train_op(
+    data_dir=data_dir,
+    model_dir=copydata.outputs['copy_output_path'],
+    action=TRAIN_ACTION, train_steps=train_steps,
+    deploy_webapp=deploy_webapp
+    )
+
+  serve = dsl.ContainerOp(
+      name='serve',
+      image='gcr.io/google-samples/ml-pipeline-kubeflow-tfserve:v6',
+      arguments=["--model_name", 'ghsumm-%s' % (dsl.RUN_ID_PLACEHOLDER,),
+          "--model_path", train.outputs['train_output_path'], "--namespace", 'default'
+          ]
+      )
+
+  train.set_gpu_limit(1)
+
+  with dsl.Condition(train.outputs['launch_server'] == 'true'):
+    webapp = dsl.ContainerOp(
+        name='webapp',
+        image='gcr.io/google-samples/ml-pipeline-webapp-launcher:v1',
+        arguments=["--model_name", 'ghsumm-%s' % (dsl.RUN_ID_PLACEHOLDER,),
+            "--github_token", github_token, "--namespace", 'default']
+
+        )
+    webapp.after(serve)
+
+
+if __name__ == '__main__':
+  import kfp.compiler as compiler
+  compiler.Compiler().compile(gh_summ, __file__ + '.tar.gz')